Online and offline sexual harassment associations of anxiety and depression in an adolescent sample

Abstract

Purpose

The aims of this study were to study the prevalence of sexual harassment online and offline, to analyse the associations between subjection to sexual harassment and adolescents’ mental health and analyse if there are any significant differences between girls and boys. We also examine if good peer-relationships interact with the associations between sexual harassment and mental health complaints.

Methods

This cross-sectional study included 594 adolescents, age 12–20. Participants responded to a web survey including the self-assessment scales Revised Children´s Anxiety and Depression Scale and Beck Youth Inventories as well as subjection to online and offline sexual harassment and peer-relational quality. Linear regression analysis was used to study whether symptoms of anxiety and depression correlated to subjection to online and offline sexual harassment and peer-relational quality.

Results

We found that 48.50% of girls and 28.19% of boys reported sexual harassment victimization. Offline was the most frequently reported site of victimization. Online harassment correlated significantly with increased anxiety and depressive symptoms in girls but not boys. Offline harassment as well as online and offline harassment correlated significantly with increasing symptoms for both genders. Participants who reported good peer-relationships had significantly less symptoms.

Conclusions

This study shows that sexual harassment remains a common plague for adolescents, especially for girls. Offline sexual harassment is the most common form of harassment for both genders. For girls, but not for boys, online sexual harassment correlated significantly with anxiety and depressive symptoms. A strong negative correlation between satisfaction to peer-relationships and mental health symptoms was found.

Keywords:

• Adolescents
• offline
• online
• sexual harassment

Introduction

Reports about the occurrence of sexual harassment in different areas of society have succeeded each other in recent years. Sexual harassment can take many forms but is usually divided into physical, verbal, or non-verbal forms and can take place in person (offline) or through media such as the internet or phone (online) [1]. Previous studies often have not separated harassment that takes place online from harassment offline. Given the fact that smartphone access and internet use are nearly universal among Swedish teens [2], there is a need to understand the impact online experiences have on the mental health of adolescents. Previous studies of sexual harassment, online and offline, among adolescents have found that the prevalence is relatively large, but this varies among studies from different countries, the year of measurement, and the definition of sexual harassment. Different studies use different strategies to ask for the respondent experiences; while some provide examples of situations of harassment for the respondent to answer, others ask open questions where the respondents themselves are left to think what situations to include. A previous Swedish study found that 51.6% of girls and 45.1% of boys reported offline sexual victimization whereas online sexual solicitations were reported in 35.5% of girls and 19.9% of boys [3]. In a countrywide Finnish study, 55% of girls and 40% of boys reported sexual victimization [4]. Two cross-sectional studies from Norway found prevalence numbers of 80% among girls and 74% among boys in 2007, and 64% and 62%, respectively, in 2013 [5]. An American study found a lower prevalence at 29% and 20% for girls and boys respectively [6]. Given this variation, local studies are important to accurately estimate the actual prevalence of this problem in different countries.

Some studies have found that sexual harassment victimization correlates with symptoms of anxiety in adolescent girls and boys [5], whereas other studies have found this correlation in girls but not in boys [7]. Cross-sectional data has linked sexual harassment victimization to other forms of mental illness outcomes among adolescents, such as higher levels of self-harm [6,8], substance use and depressive symptoms [6]. Furthermore, longitudinal data shows that participants who report more sexual harassment victimization at one point were more likely to report more emotional distress and symptoms of negative affect on the following measures, indicating a cause-and-effect relation [8,9]. Contrary to the above findings, however, a recent Finnish study found that subjection to sexual harassment was inversely or non-significantly associated with depressive symptoms in adolescents. Although a significant correlation was found with social anxiety [7]. Considering this somewhat surprising result, we found it important to include both depressive and anxiety assessments in the present analysis.

Internet and technological advances have opened new ways for communication and social participation among people [10]. As more social interactions take place online, the potential risk of cyber victimization increases. A national survey in Sweden in 2019 showed that all teenagers use the internet daily, and nearly all possess a smartphone [2]. Parallel to this, the number of teenagers who have reported being the victims of bullying and sexual assaults online has increased in recent years [2], which is a shift in trend from previous years [11]. A number of studies have found that girls are more frequently the victims of sexual harassment both online and offline [12–14]. However, not all studies are aligned. A meta-analysis of the prevalence of unwanted online sexual solicitation from 1990 to 2016 found an overall decrease since 2000, and boys were more frequently the victims [11]. A Swedish study found that online sexual harassment was associated with a higher level of depressive symptoms in girls. The association was non-significant for boys when adjusting for social support and offline sexual harassment. The same study also found that offline sexual harassment predicted online sexual harassment, indicating that victims are often targeted both online and offline [3]. Despite the integration of technology into everyday life [2], few studies have analysed the effects of sexual victimization online separately from sexual victimization offline [3,12]. As victimization online has increased and usage is universal, it is important to examine whether the negative impact on young peoplés mental health is similar to traditional forms of peer victimization offline.

Parallel to doing research on the predictors of mental illness, attention must also be given to possible factors alleviating the symptoms. Among other factors, peer support has been associated with lower levels of depressive and anxiety symptoms in adolescents [15]. Low peer support has been found to increase the odds of suffering depressive symptoms [3]. Therefore, we find it important to examine if experiencing good peer relationships is linked to a lesser likelihood of reporting sexual harassment victimization.

In summary, the aims of this study are to: (1) Study the prevalence of sexual harassment online and offline. (2) Analyse the associations between subjection to sexual harassment and adolescents´ mental health. (3) Explore if there are any differences between girls and boys. (4) Explore if satisfaction with peer-relationships interacts with the associations between sexual harassment and mental health complaints.

Materials and methods

Study setting

Data was collected from a survey conducted by the project Adolescents’- experience of mental illness- Psychometric properties of new Swedish versions of tests (UPOP). The purpose of that project was to develop new instruments for screening and treating mental illness among young patients. Ethical approval was given by the Regional Ethical Review Board in Umeå, Sweden (number 2018/59-31). The study was conducted in a medium-sized town and its surroundings in the northern part of Sweden. Data collection took place during the school year 2018–2019 and was spread over most school months that year.

Procedure

Four schools from areas of different socioeconomic standards were included after permission was granted by their principals. Verbal and written information was given to students by their teachers, written consent was then obtained from the students who chose to volunteer. Informed consent was obtained from the parents of students younger than 15 years of age. Each participant received a login code to access the web survey. Time to fill out the web survey was given during school hours. All questions had to be answered to finish the survey, therefore no missing data was meant to be created. Due to a human mistake, for some students it was possible to complete the survey without all questions answered rendering some missing data. A teacher and research assistant were present to answer questions from the students during their participation. Participants were informed that they could withdraw their participation at any time without giving any reason. A snack was provided during participation and a cinema gift card was given to participants after completing the questionnaire.

Participants

The study population was a convenience sample containing students from grades 7–9, age 12–16, and gymnasium grade 1–3, age 17–20 (M = 15.73, SD = 1.77). A total of 897 students were asked to participate in the study and 617 (69%) agreed to do so. 594 respondents were included in the analyses after the removal of respondents with too much missing data, 367 girls and 227 boys.

Self-assessment measures

Dependent variables

Revised Children’s Anxiety and Depression Scale (RCADS). The scale contains 47 items, of which 37 measure symptoms of anxiety and 10 measure symptoms of depression. Due to a mistake when programming the web survey some questions were not collected. To minimize the impact of this error, the depression scale was not used. Participants with too many missing data points were excluded. Answers were graded on a four-grade Likert scale ranging from never (0) to always (3). Respondents with data missing on more than 10 items for the whole anxiety scale, or more than two items on a separate sub-scale, were not included in the analyses [16]. The RCADS scale has shown a very good internal consistency with Cronbach’s alpha of 0.93 across many different countries and languages [17]. RCADS has also shown good reliability for cross-cultural use [17]. Cronbach´s alpha was 0.95 (95% CI [0.93 − 0.96]) in the present study. RCADS has been shown to be a valid instrument with moderate to good correlations with other anxiety assessment instruments for non-clinical samples [18].

Swedish version of Beck Youth Inventories of Social and Emotional Impairment (BYI). The BYI is an instrument developed for assessing depression, anxiety, anger, disruptive behaviour, and self-concepts in children. Each sub-scale consists of 20 questions, each ranked on a four-grade Likert scale ranging from never (0) to always (3) [19]. Only the depression subscale was used (BYI-D) in the present study. All sub-scales (including BYI-D) have previously shown satisfying internal consistency with Cronbach’s alpha above 0.80 in a Swedish normative sample [20]. Internal consistency was excellent with a Cronbach´s alpha coefficient of 0.94 (CI 95% [0.93 − 0.95]) in the present study. BYI has been found to be a valid instrument in non-clinical samples [20] in comparison with other instruments that measure similar symptoms.

Independent variables

Offline and online sexual harassment. Questions about sexual harassment experiences were inspired by previously used questionnaires [21]. Participants were asked to think back on the previous six months and report any experience of verbal, physical, or non-verbal sexual harassment. Examples were given for each sub-category: (1) Examples of verbal harassment: being called a fag or pussy against your will, or someone commenting on your private life in a sexual manner. (2) Examples of physical harassment: against your will, someone trying to grab you in a sexual manner, pulling on your clothes; for example, underwear or bra-strap. 3. Examples of non-verbal harassment: if someone spreads sexual rumours, sends pictures, messages, or films against your will. Questions for each sub-category were answered by choosing ‘no, never’; ‘yes, at school’; ‘yes, through the internet or phone’; or ‘yes, elsewhere’. The option ‘yes, through the internet or phone’ was defined as online sexual harassment and options ‘yes, at school’ and/or ‘yes, elsewhere’ were defined as offline harassment. Based on the answers from all three questions, each student was grouped into one of four groups (Online, Offline, Online and offline and Not harassed). From these groups, four dichotomous variables for harassment were created (‘online,’ ‘offline,’ ‘online and offline,’ and ‘not harassed’) and used for analysis.

Gender. The questionnaire contained one question where participants declared their legal gender.

Satisfaction with peer relationships: The questionnaire contained a three-question-form about participants’ relation to peers. The questions were: ‘has it been easy for you to make new friends and then stay friends with them,’ ‘do you feel that other peers want to be friends with you,’ and ‘are you satisfied with your relations to your friends?’ Answers were graded on a five-grade Likert scale ranging from ‘never’ to ‘almost always.’ All participants’ answers were graded and averaged, generating a score from one to five, corresponding to their overall peer-relational quality.

Statistical analyses

All statistical analyses were conducted using SPSS, Version 26.0. Gender differences in the prevalence of sexual harassment victimization were analysed using the Chi-square test and Phi-coefficients. Gender differences in mean scores for dependent variables were compared using Independent Sample t-Tests. Effect sizes were small using the definition proposed by Cohen [22], with Phi-Coefficients of 0.10 for small effects, 0.30 for medium effects, and 0.50 for large effects. Linear multiple regressions were used to examine whether sexual harassment victimization and satisfaction with peer -relationships, along with gender and age, could predict levels of depression and anxiety symptoms in our sample. Age made no significant contribution to the model and was, therefore, not included in the final analyses. Intercorrelation analysis between the dependent variables used in the multiple regression was run and was found strong and in the expected direction, r = 0.80; p < 0.01. However, both dependent variables were included due to previous research findings, discussed in the Introduction. To test whether the quality of peer relationships impacted the correlation between dependent and independent variables, interaction analysis was run but found not significant.

Tests of assumptions were calculated. Continuous variables were checked for normality using histograms and scores for skewness and kurtosis. RCADS-, BYI-D, and peer-relation -scores were found skewed, however since the sample size was large, violations of this assumption were tolerated [23]. Residual P-plots and scatterplots indicated that the assumptions of linearity and homoscedasticity were satisfied. Independent variables showed no significant multicollinearity, with no Variance Inflation Factor (VIF) above 2.

Results

Prevalence of sexual harassment

Offline sexual harassment was the most common type of harassment for both genders. It was more common for girls to be victimized both online and offline than for boys. Similarly, fewer girls than boys reported they had never been harassed. No other statistical differences were found. See Table 1.

Table 1. Prevalence of sexual harassment among boys and girls, Chi2-test and total prevalence of sexual harassment.

	Total (n = 594)		Girls (n = 367)		Boys (n = 227)
	n	%	n	%	n	%	χ^2	p	Phi
Online	47	7.91	34	9.26	13	5.73	1.95	0.16	–0.06
Offline	145	24.41	98	26.70	47	20.70	2.42	0.12	–0.07
Online and Offline	50	8.42	46	12.53	4	1.76	19.74	<0.001	–0.19
Not harassed	352	59.26	189	51.50	163	71.81	23.12	<0.001	0.20

Anxiety and depressive symptoms

Girls, on average, reported higher levels of anxiety and depressive -symptoms, assessed with RCADS and BYI-D, than boys. Mean values differed significantly (p < 0.001) between girls and boys for both scales (RCADS: girls [M = 32.46, SD = 18.53]; boys [M = 19.18, SD = 16.08]. BYI-D: girls [M = 17.93, SD = 11.20]; boys [M = 13.09, SD = 10.31].)

Anxiety symptoms

Results from multiple regression analysis showed that harassment variables were positively- and satisfaction with peer-relationships along with gender was negatively correlated with participants’ rating of anxiety symptoms, see Table 2. The independent variables explained 28% of variance in anxiety symptoms (Adj.R² = 0.28; F(5,588) = 45.97). Gender had a significant and relatively large correlation with anxiety symptoms, and therefore, we performed analyses separated by gender. Furthermore, victimization offline alone correlated more than both online harassment and online and offline victimization combined.

Table 2. Results from multiple regression for the total sample.

	Anxiety (RCADS)		Depression (BYI-D)
	Adj.R^2	ß	Adj.R^2	ß
Model	0.28***		0.29**
Gender		–0.29***		–0.15***
Online		0.11**		0.12**
Offline		0.20***		0.20***
Online and Offline		0.18***		0.22***
Satisfaction with peer-relationships		–0.30***		–0.40***

*p < 0.05, **p < 0.01, ***p < 0.001.

Depressive symptoms

Results from multiple regression analysis showed that gender, harassment variables, and peer -relationships also significantly correlated with participants’ rating of depressive symptoms, see Table 2. The results for depressive symptoms were similar to those for anxiety symptoms, although online and offline harassment combined showed the strongest connection to depressive symptoms. In total, 29% of variance in depressive symptom was explained by the model (Adj. R² = 0.29; F (5,588) = 49.34). We found that gender significantly correlated with higher BYI-D -score and, therefore, performed separate analyses for girls and boys.

Gender

See Table 3 for results. Multiple regression analyses showed that the independent variables explained more variance for depressive symptoms (Adj. R² for girls 0.28; F (4,362) = 36.90, for boys 0.21; F (4,222) = 16.00) than for anxiety symptoms (Adj. R² for girls 0.20; F (4,362) = 23.34, for boys 0.14; F (4,222) = 10.11). Furthermore, online sexual harassment correlated positively with symptoms of mental illness for girls but was not significant for boys. Offline sexual harassment alone had a stronger correlation with poor mental health on boys than online and offline harassment combined. To further analyse the effect of gender on the outcome, an interaction analysis between gender and harassment was run but found not significant.

Table 3. Results from multiple regression for girls and boys separately.

	Anxiety (RCADS)				Depression (BYI-D)
	Girls		Boys		Girls		Boys
	Adj.R^2	ß	Adj.R^2	ß	Adj.R^2	ß	Adj.R^2	ß
Model	0.20***		0.14***		0.28***		0.21***
Online		0.14**		0.06		0.15**		0.06
Offline		0.22***		0.21**		0.20***		0.21**
Online and Offline		0.21***		0.14*		0.25***		0.16**
Satisfaction with peer-relationships		–0.34***		–0.28***		–0.43***		–0.37***

*p < 0.05, **p < 0.01, ***p < 0.001.

Satisfaction with peer-relationships

Satisfaction with peer-relationships showed the strongest connection to the outcome in all analyses (see Tables 2 and 3). Participants who rated the quality of their peer-relationships higher had significantly fewer symptoms of anxiety and depression. The effect was especially large for depressive symptoms, compared to anxiety, and for girls compared to boys. To test our hypothesis that peer-relationship satisfaction could significantly influence the effect sexual harassment had on mental health, interaction analyses for peer-relationships and harassment variables were run but found not significant.

Discussion

Prevalence of online and offline sexual harassment

Results from this community sample study confirms that sexual harassment is a common nuisance among Swedish adolescents, especially girls. Despite the widespread use of internet among adolescents, offline remained the most common place for victimization of both girls and boys in the present study. Although the internet potentially allows anyone to be harassed by anyone, it also enables relatively simple ways to protect oneself, especially from unknown people, for example, by blocking incoming calls or certain contacts on social media. The same possibilities for protection offline are difficult to achieve when the perpetrator, for example, is a classmate. However, teaching children about the risks of the internet is nonetheless important as internet use will likely remain an integrated part of adolescent lifestyle in the foreseeable future.

Results from this study confirm results from previous studies that show girls to be more frequently victimized by sexual harassment [3–6,24]. A nation-wide survey in Sweden shows that this pattern lingers in adulthood, with more women than men reporting sexual harassment [25]. To our surprise, age showed no significant association with either of the dependent variables.

As stated in the Introduction, the prevalence numbers of sexual harassment vary within large spans in earlier research. Other studies in Nordic and western countries have found similar or higher prevalence numbers [3,12,24,26], and in most of them, girls were found to be more frequently victimized both online and offline. The present study applied an inclusive strategy that used open questions and provided examples of situations that could count as sexual harassment but left to the respondent to decide what situations to include. Consequently, we hoped to come as close as possible to the real prevalence number. However, the possibility of a response bias cannot be ruled out since participation in the study was optional, and the schools were not randomly selected.

Anxiety and depression linked to sexual harassment in girls and boys

Gender significantly correlated with anxiety and depressive symptoms. Analyses of girls and boys separately showed that online harassment was significant for girls but not for boys, while at the same time no significant difference in the prevalence of online harassment was found. Previous studies have found that girls experience multiple types of harassment more frequently than boys, both online [3,12], and offline [26,27]. This was not captured in our study since the number of incidents of sexual harassment were not recorded, and therefore, all participants were weighted equally, irrespectively of the amount of harassment they had experienced. The association with mental health from offline sexual harassment was almost the same for both girls and boys. So, although offline harassment is more commonly targeted towards girls, the connection to mental health, when it occurs, is similar for both girls and boys.

Satisfaction with peer-relationships

Results from the present study underlines the importance of friends as buffers to the mental illness of adolescents. Participants who reported greater satisfaction and felt themselves to be someone others wanted as a friend reported significantly lower symptoms of both depression and anxiety. Whether this is mediated because teens who have many friends are possibly victims to a lesser extent or, as has been found before [24,28], the acts of harassment are less distressing for victims who have better peer -support, is beyond the scope of this article. Surprisingly and contrary to our hypothesis, no significant interaction effect was seen between peer-relationships and sexual harassment. Possibly, the positive effect rendered by good peer-relationships is general and not specifically seen when an adolescent is subjected to sexual harassment. The correlation between satisfaction with peer-relationship and less symptoms of mental illness, however, was large for both anxiety and depressive symptoms and for both genders, thus supporting strategies for creating inclusive environments for adolescents in preventing mental illness.

Clinical relevance

Since this study points to the fact that sexual harassment correlates with symptoms of anxiety and depression in a convenience school sample, further studies related to a clinical population would be of interest. In the future, it is important for clinical work to know more about the ways in which boys and girls are exposed to sexual abuse online and how they engage in sexual risk behaviours online and offline. Moreover, increased knowledge about online sexual victimization is needed for victims, caregivers, as well as for care providers. The knowledge can be used in the assessment of mental illness in clinics, and as a knowledge base in trauma therapy. In clinics, there is a need to ask more about online environments, otherwise teens might not feel that the professionals can understand them.

Methodological aspects and limitations

No claim of causality can be made due to the cross-sectional design used in this study. Although the connection between harassment and poorer mental health has been established through both cross-sectional and longitudinal data, studies have also found the relationship to be reciprocal [9,21,29]. This suggests that adolescents with mental health problems could also be more vulnerable to harassment than others [9].

All variables significantly correlated with the outcome for the total sample. However, due to small sample sizes and large confidence intervals in the groups of boys who reported online only (n = 13), and online and offline (n = 4) harassment, no strong conclusions can be drawn from these groups. However, strong significant relationships were found even though the gender groups were small.

Participants were not geographically stratified and did not fully match the Swedish pediatric population, for example, the unbalanced gender ratio limited generalizability. As mental health is associated with socioeconomics, another limitation was a quantified measure of socioeconomics.

The instruments used to measure online- and offline sexual harassment were not originally meant for this purpose and therefore not ideally designed. Even though some response alternatives might have appeared confusing, the questions were designed to include as many experiences as possible.

Disclosure statement

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

Additional information

Funding

The present research was financially supported by Västerbotten county council, award number RV-931721, and Swedish Psychiatry fund.