The relationship between school-related gender-based violence and absenteeism: Evidence from 14 Southern and Eastern African countries

Children in sub-Saharan African countries face higher exposure to gender-based violence (GBV) compared to their counterparts in other world regions (United Nations Children’s Fund [UNICEF], 2014). When GBV occurs in schools, it severely endangers access to education. Using the third round of data of the Southern and Eastern Africa Consortium for Monitoring Educational Quality (SACMEQ) from over 60,000 Grade 6 learners across 14 countries, we examined the relationship between GBV in primary schools and learners’ absenteeism. Findings indicate that sexual harassment perpetrated by teachers significantly increases learners’ absenteeism. In contrast, effects were less clear when the perpetrator was a fellow learner. Effects found are similar in magnitude for girls and boys. Violence prevention education programmes and stricter punishment for offenders are needed to establish a safer school environment and overcome harassment-related barriers to learning.


Introduction
With respect to school environments which should be gender-sensitive and safe, school-related gender-based violence (SRGBV) severely infringes basic child rights and stands in the way of a child's access to education. SRGBV can be defined as "acts or threats of sexual, physical, or psychological violence occurring in and around schools, perpetrated as a result of gender norms and stereotypes, and enforced by unequal power dynamics" (United Nations Educational, Scientific and Cultural Organization [UNESCO] & United Nations Girls' Education Initiative [UNGEI], 2015:2). Gender-based violence (GBV) reflects the deeply rooted social and cultural contexts that cause inequality and power imbalances between different genders and between adults and children (UNGEI & UNESCO, 2013). While GBV is present in all societies at the beginning of the 21st century, it occurs more often in societies where females are accorded a lower social rank and where poverty is more prevalent (Jones, Moore, Villar-Marquez & Broadbent, 2008;UNICEF, 2014).
Manifold detrimental impacts of SRGBV have been reported on physical and mental health, as well as on educational outcomes. Victims of SRGBV often suffer from anxiety, depression, and post-traumatic stress disorder. They may further be exposed to severe physical injuries, unwanted pregnancies, and sexually transmitted diseases such as HIV/AIDS. Moreover, SRGBV victims are more likely to skip classes, stay at home, or drop out from school altogether, negatively impacting on their learning outcomes and capabilities (Leach, Dunne & Salvi, 2014;Psaki, Mensch & Soler-Hampejsek, 2017;UNESCO & UNGEI, 2015).
In many African countries, GBV and discrimination are often taken for granted in school life (Leach, Fiscian, Kadzamira, Lemani & Machakanja, 2003;Lee, Rhee & Rudolf, 2019). Even though it has been acknowledged that SRGBV inhibits learners' attendance, thereby widening the gap between school intake and completion rates, GBV-related practices continue to persist in educational institutions while schools often neglect the situation (Harber, 2014;Jones et al., 2008;Leach et al., 2014;Plan, 2008). SRGBV victims are not only older learners, but the issue is already significant in primary schools. According to the 2007 SAQMEQ III of primary schools in 15 African education systems, on average 41% of school heads reported having issues with learner-to-learner sexual harassment in their schools. Moreover, 16% of school heads reported having issues with teacher-to-learner sexual harassment. The actual figure is likely to be even higher, given that sexual harassment tends to be under-reported due to cultural norms, shame, and failure to identify what harassment is and what it is not (UNICEF, 2014). Schools often implicitly legitimise and maintain destructive gender norms through the tacit approval of existing violent conditions (UNGEI & UNESCO, 2013). Furthermore, teachers often neglect the problem, justify the abuse, or become perpetrators themselves. Cases have been reported in which teachers use their power to force learners to have sexual relations, raising the reason that female learners' outfits or behaviours are provocative and teachers are in sexual need due to the long distance from home (Harber, 2014;Plan, 2008).
Given the significance of SRGBV, particularly in African countries, the objectives of our study were threefold: (1) to examine the existence and prevalence of sexual harassment in primary schools across 14 Southern and Eastern African countries; (2) to estimate the effect of sexual harassment on learners' absenteeism from school; and (3) to examine whether this effect differs depending on the perpetrator (fellow learner vs. teacher).
This study contributes to the literature on SRGBV in several ways. Firstly, the SRGBV literature notoriously suffers from a lack of data. Despite the growing recognition of both the prevalence of SRGBV and its significance for learning outcomes, the issue is still largely underresearched and under-reported (UNGEI & UNESCO, 2013). Most past studies were conducted for a single country, used qualitative or descriptive methods, and sometimes only relied on anecdotal evidence (Abuya, Onsomu, Moore & Sagwe, 2012;Bisika, Ntata & Konyani, 2009;Ncontsa & Shumba, 2013). Hence, the main contribution of this study is to help fill this gap by providing quantitative evidence from nationally representative surveys for 14 Southern and Eastern African countries. In particular, statistical evidence from large and multi-country datasets is extremely rare in the SRGBV literature. To the best of our knowledge, our study presents the largest scope at which the effect of SRGBV on absenteeism has been rigorously tested to date. Secondly, by using multivariate regression techniques, our study will be able to control for many confounding factors that drive absenteeism. Earlier studies, in contrast, often failed to control for the influence of socioeconomic and other environmental factors. This will allow us to isolate the effect of SRGBV on absenteeism more precisely. Thirdly, while several studies focussed on SRGBV and absenteeism, most studies investigated learners' reluctance to attend rather than actual attendance/absence from school (Hill & Kearl, 2011;Lipson, 2001). In our study we measured actual absence from school based on learners' self-reports. Fourthly, existing studies have often dealt with violence against children and school violence more generally, studying gender aspects only on the side (e.g. UNESCO & UNGEI, 2015). In contrast, in this article we focus exclusively on sexual harassment perpetrated by teachers and fellow learners in order to provide a more in-depth analysis of the issue of SRGBV. Finally, with this study we hope to contribute with tangible evidence to guide SRGBV-related policies in Southern and Eastern African countries and beyond.

Conceptual framework
Children have the right to live free from fear and abuse, as well as to receive basic education to enable them to lead an informed and cultivated life. In 1989, the United Nations (UN) Convention on the Rights of the Child acknowledged a child's right to education (Article 28) and called for the protection of children against all forms of abuse and neglect (Articles 19 and 34). Inspired by the UN Convention, a year later, in 1990, the African Charter on the Rights and Welfare of the Child ([ACRWC], African Union, 1990) stipulated in Article 11 that "[e]very child shall have the right to education." In Article 16 it further calls for the protection of children against all types of abuse and torture, including "psychological and mental injury or abuse" and "maltreatment including sexual abuse." Both GBV and school absenteeism had been officially acknowledged to represent major infringements of basic child rights. In a recent review on child rights in Africa, UNICEF (2020) concludes that the ACRWC has helped to enhance awareness of child rights, trigger legislative change, and improve accountability during the past 30 years.
Why child rights matter for human development has been further conceptualised by Amartya Sen and Martha Nussbaum (Nussbaum & Sen, 1993). Education and the freedom from abuse are important prerequisites in order to achieve higher states of human capabilities. Sen and Nussbaum define capabilities as what a person is capable "to be and to do" in life. Among the central human functional capabilities defined by Nussbaum (2001), bodily integrity, freedom from fear and anxiety, and access to an adequate education play key roles. Sen and Nussbaum have further stressed how various functional capabilities are interrelated. For example, both authors have argued that in societies that strongly discriminate based on gender, females' access to education and income are undermined, limiting females' capability space. Motivated by both the child rights approach and the capability approach, with the empirical analysis in this article we aim to shed new light on the prevalence of GBV and its impact on access to education across Southern and Eastern Africa.

Past evidence
Research on school violence has been growing over the past decades in both developed and developing countries. While most studies have focussed on the issue of bullying and violence in general, the number of studies that have dealt with the gender issue in school violence is rather limited. Most existing studies have employed qualitative methods, and quantitative evidence from large-scale data sets is particularly rare. Even when such quantitative data were available (particularly in OECD countries), analyses have often relied on "superficial analysis" of the data (Leach et al., 2014:11). Existing studies on school-related GBV are usually scarcer for the Global North compared to developing countries (Research Triangle Institute [RTI] International, 2016;UNGEI & UNESCO, 2013). Some countries have collected data on GBV at schools but have not analysed the effects or consequences of such violence on education outcomes (Girlguiding, 2017;Ranchod & Boezak, 2016). A notable exception is the American Association of University Women (AAUW), which has consistently conducted research on secondary schools in the United States of America, with sample sizes of between 1,600 and 2,000. Their findings indicate that sexual harassment caused learners to stay home from school or skip classes (Hill & Kearl, 2011;Lipson, 2001). Other studies from advanced countries found associations between sexual harassment and reduced school adjustment (Felix & McMahon, 2006) and feelings of discomfort and unsafety in schools (Gådin, 2012;Ormerod, Collinsworth & Perry 2008;Timmerman, 2003).
In the developing world, especially in sub-Saharan Africa, most existing analyses have employed smaller case studies and it is particularly hard to find statistical evidence from large samples (Dunne, Humphreys & Leach, 2006;Leach et al., 2014). Using one of the larger samples for Malawi in 2004, Bisika et al. (2009) found that of 1,496 teenage females in their sample, 43.9% reported that someone had touched their breasts, buttocks or genitals inappropriately and without their permission. More than 60% of these incidents had taken place in the 6 months prior to the interview. The largest proportion of these assaults (48.7%) had taken place on school premises. According to the study, the majority of perpetrators of this kind of GBV were fellow learners (52.2%), while teachers only represented a small portion of perpetrators (3.8%). Moreover, 39% of the harassed girls reported that sexual harassment adversely affected their education and 3% dropped out of school. In a more recent study in Malawi, Psaki et al. (2017) used longitudinal data to examine the relationship between violence in schools and at home and education outcomes for 869 females and 884 males aged 14 to 17. Unlike other studies, the authors did not find significant associations between school violence, absenteeism, and dropout for girls; in contrast, male victims were 50% more likely to be absent from and drop out of school than their female counterparts. For girls, rape and pregnancy are among the most severe consequences of GBV. According to Leach et al. (2003), 44% and 63% of the study participants reported knowing a girl's pregnancy case by a male peer in schools in Malawi and Ghana, respectively. As for the prevalence of rape related to secondary schools in Lesotho, 10.7% (7%) of learners had the perception that rape perpetrated by children (educators) had occurred in the 12 months prior to the interview (De Wet, 2007). In further studies from Zimbabwe, South Africa, Nigeria, Ethiopia, and Kenya it was found that sexual abuse and harassment resulted in victims feeling insecure, lowering their participation, limiting their movement in school, and ultimately absenteeism and dropout (Abuya et al., 2012;Altinyelnken & Le Mat, 2018;Leach, Machakanja & Mandoga, 2000;Ncontsa & Shumba, 2013;Onoyase, 2020). Using the same dataset used in our study, Saito (2013) finds an increase in the perceived prevalence of sexual harassment and other forms of violence at schools across Southern and Eastern African countries between 2000 and 2007. It is unclear, however, whether this represents a real increase or increased reporting due to higher awareness of the issue. In her study she further examined the relationship between sexual harassment and various learning outcomes (including absenteeism). While she found no correlation between a violence index and absenteeism, the analysis was a rather simple crosstabulation and did not control for the numerous other factors affecting absence rates.
In summary, past studies have shown that SRGBV in sub-Saharan Africa is a substantial problem which endangers child rights and access to education. SRGBV is perpetrated by peer learners andto a lesser extentby teachers. In particular, past evidence indicates a clear link to exist between SRGBV, mental health, and access to learning. In the following section we investigate this link for 14 Southern and Eastern African countries, controlling for a broad set of potentially confounding factors, and distinguishing SRGBV effects by the type of perpetrator. Based on the existing literature, through the analysis in this study we examine the following hypotheses: (H1) Learner absenteeism and sexual harassment are common problems at primary schools in Southern and Eastern African countries; (H2) Sexual harassment in schools increases learner absenteeism; (H3) The impact of sexual harassment on absenteeism may differ depending on the perpetrator.

Data and Methodology Data
In this study we used secondary data analysis relying on data collected by the SACMEQ III Project in 2007. SACMEQ III collected data in SACMEQ member countries covering 15 education systems in total: Botswana, Kenya, Lesotho, Malawi, Mauritius, Mozambique, Namibia, Seychelles, South Africa, Swaziland, Tanzania (mainland), Tanzania (Zanzibar), Uganda, Zambia, and Zimbabwe. Although the fourth round of SACMEQ had been implemented between 2012 and 2014, SACMEQ IV data has not been publicly released and national reports were only available from four countries at the time of this study. Clearly, SACMEQ III data are no longer able to provide the latest accuracy in terms of the prevalence of SRGBV and absenteeism. Nevertheless, it can help us shed new light on the relationship between SRGBV and absenteeism which is unlikely to change over time. According to international organisations, the prevalence and significance of the issue has not decreased since 2007 (UNESCO, 2017;UNICEF, 2014).
Data for the SACMEQ III project were collected during the last quarter of 2007 by interviewing learners, teachers, and school heads. SACMEQ III assessed the schooling conditions and the performance levels of both learners and teachers in literacy, numeracy, and basic health knowledge. The desired target population was "all learners at Grade 6 level in 2007 (at the first week of the eighth month of the school year) who were attending registered mainstream primary schools." Among the total targeted population, learners in schools with fewer than 20 learners in Grade 6 or in schools for special educational needs were excluded, which accounted for 3.6% of the total targeted population. Stratification was carried out on the regional level, i.e. national sampling frames of schools were separated into regional lists of schools prior to undertaking the sampling.
SACMEQ III surveyed 2,779 primary schools, 8,026 teachers, and 61,396 learners in Grade 6 following a two-stage clustered sampling design. ii At the first stage, schools (clusters) in the target population were selected based on a probabilityproportional-to-size (PPS) technique. At the second stage, 25 learners were sampled from all the Grade 6 classes in each sampled school (cluster) by computer-generated random numbers. In particular, this random selection of the learner sample at each school level was carried out by trained data collectors. It was not acceptable to permit school principals or classroom teachers to select learners, as those might have had an interest to select their best learners. Sampling design was carried out in accordance with international standards. In fact, the predecessor of SACMEQ in the 1990s was initiated by the International Institute for Educational Planning. Hungi, Makuwa, Ross, Saito, Dolata, Van Cappelle, Paviot and Vellien (2010) provide further details regarding data collection. The final sample used in this study comprised 60,805 learners, 4,303 reading teachers, and 4,259 mathematics teachers from a total of 2,745 schools across 14 countries.

Methodology
Using multiple regression analysis, we employed probit estimation to examine the relationship between GBV in schools and learners' absenteeism. Starting with a simple regression specification, additional sets of independent variables were added in sequence until the full model controls for all observable aspects of a learner's decision to attend. The dependent variable is the binary variable , which indicates the absence from school of learner from class in school in country , and is defined as follows: Equation 1 presents the simplest regression model that includes only harassment variables and country dummies.

Equation 1
Sexual harassment, the SRGBV measure used in this study, is defined as "any unwelcome sexual advance, request for sexual favour, verbal or physical conduct or gesture of a sexual nature, or any other behaviour of a sexual nature that might reasonably be expected or be perceived to cause offense or humiliation to another" (UNSG, 2008:1). In the model of Equation 1, the vector includes a set of dummy variables reflecting the head of the school's perception of sexual harassment targeting learners in school in country at the time of the interview.
Sexual harassment problems within schools as perceived by heads of schools were the main independent variables used in this study. iii Two types of sexual harassment variables were used in this study: harassment perpetrated by learners (s_lharass) and harassment perpetrated by teachers (s_tharass). School heads were asked how often the school had to deal with a number of behaviour of learners and teachers, including sexual harassment. Responses were categorised into three options: "never", "sometimes", and "often." Unfortunately, SACMEQ does not ask learners about their individual experiences of sexual harassment. Only school heads are asked about the average occurrence of sexual harassment in their school. Therefore, the coefficient vector β in Equation 1 should be interpreted as the effect of sexual harassment occurring in school on all learners' average likelihood of absence. This average effect of harassment on all learners' absence will be lower than the effect on only the victim's absence, given that it averages over victims, bystanders, and non-victims. The term is a vector of country dummies and is the error term. Given that we used clustered survey data, standard errors were adjusted for clustering of observations at the school level. The simple regression model in Equation 1 could then be augmented to more complete models until the full model.

Equation 2
where and indicate vectors of individual and household characteristics of learner , respectively. Moreover, refers to a vector of teacher characteristics of both reading and mathematics ("math" hereafter) teachers of class , and is a vector of school characteristics of school . Table 1 provides a detailed description of all variables used in this study. Frequency of vandalism by learners (never = 0; sometimes = 1; often = 2) Theft Frequency of theft by learners (never = 0; sometimes = 1; often = 2) Teacher absence Frequency of teacher absenteeism (never = 0; sometimes = 1; often = 2) Country dummies Dummy variable (1 = yes; 0 = no) for each country Regressions can further be estimated separately for girls and boys. iv While much of the discussion around SRGBV has been focussed on the impact on girls, lately there is growing recognition of violence targeting boys and non-binary gender children (UNICEF, 2014). Some evidence exists that girls are more likely to experience GBV, while boys are more likely to experience corporal punishment (UNESCO, 2017). Besides gender-specific effects of SRGBV on attendance, other gender-related factors exist that can lead to a difference in average school attendance by boys and girls. On the one hand, school absenteeism of females in Grade 6 can be expected to be elevated due to girls' monthly menstrual processes and their responsibilities in the home (Miiro, Rutakumwa, Nakiyingi-Miiro, Nakuya, Musoke, Namakula, Francis, Torondel, Gibson, Ross & Weiss, 2018). On the other hand, boys can be distracted from attending school because they are often expected to work for pay besides school in order to complement family income (Wolf, McCoy & Godfrey, 2016).

Results
In the following section we present the results of this study, starting with descriptive statistics, followed by main regression results, and a heterogeneity analysis by gender. Table 2 presents summary statistics for all variables used in this study. On average, learners were 13 years old, had 4.2 siblings, and their mothers and fathers had 7.3 and 8.1 years of schooling respectively. Eighty-two per cent of learners spoke English at home. The sample was further equally distributed across sex; 51% of the sample was female. Moreover, 41.5% of learners attended a school in an urban area. Teachers were on average 37.4 years old. Across the entire sample, 44.9% of learners were absent from school at least once during the past month (girls 43.8%; boys 46.6%). School heads reported learner-to-learner sexual harassment to be an issue in the schools of 40.5% of learners in the total sample (in 29.6% of cases school heads reported the problem to occur "sometimes"; in 10.9% of cases it was reported it to occur "often"). Teacher-to-learner sexual harassment was reported to be an issue in the schools of 15.5% of learners (6.9% reported "sometimes"; 8.6% "often"). While no internationally comparable statistics exist for sexual harassment in primary schools, it can be expected that these numbers are high given that SACMEQ countries are among the highest in GBV experienced during teen age (UNICEF, 2014:67). Table 3 presents the average occurrence of absenteeism and sexual harassment by country according to SACMEQ III data. The prevalence of both phenomena were substantial, lending support to hypothesis H1. Substantial differences existed in learner absence rates across education systems. Mauritius (74%), Zambia (68%), and Uganda (65%) recorded the highest prevalence of school absenteeism during the past month. While in Mauritius, 74% of all learners were absent at least once during the past month; this was only the case for 16% of learners in Botswana, the country with the lowest absence rate in the region in 2007. These rates confirm earlier literature that found high prevalence of truancy in sub-Saharan African countries (e.g. Onyeaka, Kugbey, Ayanore & Asante, 2020).  (Saito, 2013). Table 3 further presents the average prevalence of sexual harassment issues reported by the heads of primary schools by country. Mauritius (15%; "sometimes" + "often") and Mozambique (19%) reported the lowest incidence of learner-tolearner harassment, while the highest incidence was reported in the Seychelles (62%), in Uganda (58%), in Kenya (49%), and Zambia (48%). Reported teacher-to-learner harassment was the highest in Uganda (37%) and Zambia (21%), while it was the lowest in the Seychelles (0%), in Mauritius (3%), and Botswana (4%). Table 4 shows the results of estimating the models laid out in Equations 1 and 2. Column 1 shows the marginal effects between sexual harassment and absenteeism when only controlled for country fixed effects. v In Column 2, learner-level control variables were added, which is followed by household-level controls in column 3. In column 4, teacher-and school-level variables were further added to the model. Standard errors in parentheses are corrected for clustering at the school level; ***p < 0.01, **p < 0.05, *p < 0.1. Reference categories for harassment are "s_lharass_never" and "s_tharass_never." The reference for country dummies is Botswana.

Regression Results
Keeping everything else constant, estimated coefficients reveal a statistically significant relationship between sexual harassment and learners' absence from school when harassment originates from teachers. However, no significant effect on absenteeism was found when harassment originated from fellow learners. Furthermore, effects of teacher harassment on learner absence were significantly higher when harassment occurred "often" compared to "sometimes." In the full model (4), marginal effects were 0.0332 (p = 0.00035) for "sometimes" and 0.0563 (p = 0.000024) for "often"; both being highly statistically significant. Estimated coefficients indicate that a learner's likelihood of absence during a given month rises by 3.3 percentage points when the school has problems with teacher-tolearner harassment occurring "sometimes", and by 5.6 percentage points when harassment occurs "often." The evidence from Table 4 lends support to hypotheses H2 and H3. It can further be observed that absence from school tends to be lower for female learners, for learners from better-off families, if the learner's mother is more educated, if learners live closer to the school, and if a free meal is offered at school.  Table 4. Estimations include the same set of control variables as the full model in Table 4, Column 4.
In a next step, the full model in Equation 2 was estimated separately for girls and boys to examine potential heterogeneity in effects by gender. The results of these estimations are summarised in Table 5. Interestingly, estimated coefficients suggest that no significant difference existed in how sexual harassment affected girls compared to boys. Both genders were negatively affected by SRGBV, and sexual harassment increased absenteeism for both girls and boys at a similar order of magnitude.

Discussion
Summarising the above results, the analysis in this study points towards a high prevalence of sexual harassment in primary schools in SACMEQ countries. It further shows a strong relationship between teacher-to-learner harassment and school absenteeism, while this effect was not found to be significant for learner-to-learner harassment. Moreover, effects found were not only significant for girls, but similarly important for boys.
GBV, including both physical and verbal abuse, yields impairment of emotional development, mental distress, and low self-esteem (Altinyelken & Le Mat, 2018;Ncontsa & Shumba, 2013). Victims may be discouraged from active participation in class and pursuit of academic excellence because they fear to attract unwanted attention from teachers and peers for an inappropriate purpose. The occurrence of GBV creates an intimidating and stressful learning environment and disengagement from schools. In turn, it hinders learners' attendance and causes dropout (Bisika et al., 2009;Jones et al., 2008;UNESCO & UN Women, 2016). With respect to physical harm, especially for girls, learners may become pregnant which usually signifies the end of a girl's education (Leach et al., 2014). An intimidating learning environment is not only an issue for victims themselves. Bystanders of violence may also suffer from mental distress, such as helplessness, nightmares about being the next target, and guilt for not taking action (Flannery, Wester & Singer, 2004;Janosz, Archambault, Pagani, Pascal, Morin & Bowen, 2008). Rivers, Poteat, Noret and Ashurst (2009) found empirical evidence that observing peer victimisation can negatively impact witnesses' mental health. That is to say, GBV in school adversely affects the whole school environment by inhibiting direct and indirect victims' learning.
The findings in this study suggest a significant impact of sexual harassment on absenteeism, in particular when the former originates from teachers. This is in line with results from Timmerman (2003) for the Netherlands, Psaki et al. (2017) for Malawi, Chitsamatanga and Rembe (2020) for South Africa, Onoyase (2020) for Nigeria, and Johnson Ross and Parkes (2020) for Ethiopia and Zambia, who found that GBV increased absenteeism and school dropout. The findings in our study are contradictory to the findings by Saito (2013), who also used SACMEQ data and did not find a significant correlation between a violence index and absenteeism. However, her analysis was a rather simple crosstabulation and did not control for the numerous other factors affecting absence rates. The distinctive contribution of our study is that it provides quantitative evidence from nationally representative surveys for 14 Southern and Eastern African countries. In particular, statistical evidence from large and multi-country datasets is extremely rare in the SRGBV literature. To the best of our knowledge, our study presents the largest scope at which the effect of SRGBV on absenteeism has been rigorously tested to date.
Several potential interpretations are considered for our results. GBV is rooted in power relations and teachers are generally positioned superior to their learners in schools (Onoyase, 2020;Psaki et al., 2017). Since teachers can commit GBV by wielding their authority (Dunne et al., 2006), teacher-to-learner harassment could involve a more distinct power relation compared to learner-to-learner relations. Learners usually don't have the power to decide whether to consent to a teacher's coercive behaviour (International Women's Human Rights [IWHR] Clinic Georgetown University Law Center, 2018). According to Leach et al. (2014), in sub-Saharan Africa, there are widespread reports that teachers demand sexual favours from girls in exchange for preferential treatment, such as money, good grades in exams, protection from corporal punishment, private tutoring, and promises of marriage. What is more serious is that they threaten learners with punishment, exam failure, or public ridicule if learners don't accept their favours. Teachers also take it for granted to use free learner labour and force girls in particular, to perform personal errands in their homes where they can avoid other people witnessing (Parkes & Heslop, 2011). In research conducted in Cape Town, South Africa, high school girls pointed out male teachers' staff rooms as one of the unsafe places in schools (Mandela, 2012). Furthermore, punishment for the teacher's perpetration is often too light (Chitsamatanga & Rembe, 2020;IWHR Clinic Georgetown University Law Center, 2018). In sub-Saharan Africa, when an occurrence of GBV is revealed, even in the case when a learner is pregnant, teachers are hardly expelled from their careers. At most, the perpetrating teacher is transferred to another school (UNESCO, 2017).
Furthermore, contrary to most previous studies, findings show no significant relationship between sexual harassment by peer learners and absenteeism, which is surprising. This can in part be explained by the nature of the dataset. Since sexual harassment was measured at the schoollevel, not the individual level, estimated effects are averaged across victims, bystanders, and nonvictims, which tends to water down the direct effect.
Regarding gender-specific regression, findings indicate similar negative effects of teachers' harassment on absenteeism for both girls and boys. This is in contrast to several previous studies suggesting stronger impacts on girls than on boys (Gruber & Fineran, 2008Hill & Kearl, 2011). However, it is in line with Psaki et al. (2017), who found only significant impacts of sexual harassment for boys and not girls in rural Malawi. Our results support a recent trend in the GBV literature to look beyond girls as main victims.

Limitations of the Study
This study was not without limitations. Firstly, information on sexual harassment was self-reported by the school head. Considering a tendency that sexual harassment is usually under-reported and that school heads tend to be reluctant to report GBV problems in their schools, the frequency of sexual harassment reported is likely to underestimate the actual magnitude of the problem. Secondly, SRGBV should best be measured at the individual level. SACMEQ III only asked school heads about the average incidence of sexual harassment in their school as a whole. Learners were not asked questions related to GBV. This complicates the estimation of the actual effect of such violence on the victim. Rather, our results estimate the impact of sexual harassment cases on school absenteeism averaged across victims, bystanders, and non-victims. The coefficient estimates in our study can therefore be viewed as the lower bounds of the impact of SRGBV on absenteeism of victims.

Concluding Remarks
Schools should be safe havens where learners can study freely and without fear. However, schoolrelated GBV not only presents a severe infringement of basic child rights and the very role of schools, it further jeopardises children's access to education and thus to basic human capabilities. GBV can lower learners' self-esteem and motivation which leads to disengagement from school and an increase in learners' absenteeism. In this study we investigated the relationship between SRGBV and learners' absence from school for more than 60,000 Grade 6 learners across 14 Southern and Eastern African countries. Forty-one per cent of schools reported to have problems with learner-to-learner sexual harassment, while 16% reported to have problems with teacher-to-learner sexual harassment. South Africa has shown comparatively less teacher-to-learner sexual harassment compared to the region (6%), yet the same prevalence of learner-to-learner harassment remains (41%) although dedicated programmes towards safer and more gender-sensitive schools had been running since 1999 (Republic of South Africa, 2020). Our findings indicate significant effects of sexual harassment perpetrated by teachers on learners' absence from school. In schools where the head of the school reported the issue to occur "sometimes", learners were on average 3.3% more likely to be absent from school at least once during the past month. When the issue occurred "often", this likelihood increased by 5.6%. No significant effects were found for sexual harassment perpetrated by fellow learners. In addition, results found were similar in size for boys and girls.
These results have profound policy implications. Poor attendance easily extends to dropout and the end of education. Failure to complete primary education renders it impossible to pursue further education and to acquire jobrelated skills. Given its importance, it is necessary to protect the child's rights to human dignity, bodily integrity, and to education. A combination of stronger law enforcement and effective awarenessraising programmes in schools is needed in order to create more gender-sensitive learning environments across Southern and Eastern Africa. Succeeding in these endeavours would signify a key step forward in realising the goals of the ACRWC.

Authors' Contributions
Sora Lee wrote the manuscript and cleaned and analysed the data. Robert Rudolf co-wrote the manuscript, analysed the data, designed the tables, and provided advice throughout the entire process.

Notes
i.
This article is based on the master's thesis of Sora Lee. ii.
It should be noted that while the data only allows to evaluate the impact of SRGBV on Grade 6 learners' absenteeism (only Grade 6 learners were interviewed), the occurrence of sexual harassment as reported by the school head is based on the entire school. Focusing on SRGBV in primary schools is another contribution of this article, given that most existing literature is on the issue in secondary school settings. iii.
While absenteeism is self-reported by learners in SACMEQ data, learners were not asked about experiences of SRGBV. Instead, school heads were asked about the latter being an issue for the school as a whole. The SACMEQ questionnaire only includes questions on sexual harassment, not on sexual violence (forced sexual acts, attempts to obtain sexual acts, etc.). iv.
SACMEQ data do not allow the identification of nonbinary gender. v.
Marginal effects show the change in y when the xvariable is raised by one unit (dy/dx