Abstract
Using data from the 2022 Minnesota Student Survey (N = 85,004), this study applies multivariate logistic analyses for rare events to understand better sexual violence committed by adolescents from a general strain theory perspective. Informed by this theory, measures of unmet goals, loss of positive stimuli, and exposure to negative stimuli were used to develop multivariate logistic models that were examined to improve our empirical understanding of adolescent perpetration of sexual violence. The study also investigates whether sex-based differences exist among perpetrators. Results indicate that older adolescents and males were more likely to perpetrate sexual violence, with various forms of strain increasing the risk of perpetration as well as demonstrating slightly similar effects across sexes.
Introduction
Although much of the research on sex offending focuses on adults, it is crucial to pay attention to juvenile sex offenders, as they account for a considerable proportion of sexual offenses (Ryan & Otonichar, 2016; Siria et al., 2020). For example, data from the Federal Bureau of Investigation’s 2019 Uniform Crime Report (FBI, 2019) demonstrates that juveniles represent approximately 16% of all arrests related to sex offenses, barring rape and prostitution, and roughly 17% of all rape-related arrests. A report from the National Incident Based Reporting System (NIBRS) indicates that juvenile offenders accounted for various sex crimes including rape (7.3%), sodomy (18.5%), and fondling (16%; FBI, 2015). In a study of 4,665 adolescent respondents from Michigan public schools, Ngo et al. (2018) similarly found that approximately 18% of the sample self-reported that they had engaged in some form of sexual perpetration at least once over the course of their 4-year study. Given that UCR and NIBRS data reflects only those juveniles coming into contact with law enforcement and that biases within self-reported data (e.g., social desirability and survey conditions) are likely, the actual number of adolescents perpetrating sexual offenses may be substantially higher than what is shown in these reports (Becker & Hicks, 2003). Therefore, identification of key risks that may lead youth to engage in sexual offending is required in order to effectively respond to this problem.
Prior studies have assessed numerous risk factors associated with the perpetration of sexual violence among youth (Fox, 2017; Ngo et al., 2018; Siria et al., 2020; Tharp et al., 2013; Ueda, 2017). In some of these investigations, researchers have applied diverse theoretical frameworks to explore various risk factors that contribute to adolescents who perpetrate sexual violence. When driven by theory, study findings may offer greater insight into the factors that motivate juveniles to sexually offend. However, the number of studies that have adopted the core propositions of Agnew’s (1992) General Strain Theory (GST) to explain why youth engage in this violent conduct is quite limited. Assessment of the connection between the experience of strain and sexual offending behaviors may also guide direction in the design of prevention and intervention strategies.
The primary research objective of the present study is to examine the associations between adolescents’ demographic characteristics, their failure to achieve goals, loss of positive stimuli, and exposure to negative stimuli (e.g., types of victimization) in perpetrating of sexual violence. A secondary research objective is to explore potential sex-based differences among adolescent sexual violence perpetrators. While considerable research attention has been attributed to the study of adult and juvenile sexual offenders who are male, there continues to be a lack of recognition aimed at juvenile sexual offending that occurs among females. GST has been previously applied as a framework to explain the gender gap in rates of delinquency participation, and to explain gender differences in criminal coping as a result of strain (e.g., Broidy & Agnew, 1997; Piquero & Sealock, 2004). However, these propositions have not been applied to examine potentially gendered risks that may impact sexual offending among youth. As a result, the abovementioned risk factors are examined through a GST lens to assess potential gender differences in juvenile sexual offending.
Literature Review
Adolescent Sexual Violence
The perpetration of sexual offenses is characterized as unwanted sexual contact exerted by an individual(s) against another person without consent (Heerde et al., 2015; Tharp et al., 2013). More importantly, sexual assault is generally accepted as a global health issue and a violation of human rights (Basile et al., 2018; Kuo et al., 2018). The consequences of sexual violence victimization are well-documented, encompassing physical, mental, and sexual health issues, affecting victims regardless of age, sex, or race/ethnicity (Centers for Disease Control and Prevention, 2022; Gewirtz-Meydan & Finkelhor, 2020; Krahé & Berger, 2017; Ybarra & Thompson, 2018). Furthermore, sexual violence perpetration in adolescence has also been found to be a predictor of negative mental health outcomes (i.e., depression and posttraumatic stress symptoms) and future sexual offending (see Meadows et al., 2022).
Existing scholarship highlights that adolescent aged 12 to 17 years are particularly susceptible to suffering from sexual violence (Banvard-Fox et al., 2020; Kuo et al., 2018). For instance, Felson et al. (2021) note that adolescents are at a greater risk of victimization by virtue of their limited capacity to ward off perpetrators and their regular contact with conceivable offenders such as parents or caregivers, family members, teachers or school staff, peers, acquaintances, and strangers (Lundgren & Amin, 2015). Furthermore, studies on sexual violence have demonstrated similarities between adult and adolescent victims, with both groups usually reporting that their experiences of sexual assault involved perpetrators with whom they were acquainted with (see Banvard-Fox et al., 2020; Gewirtz-Meydan & Finkelhor, 2020).
Finally, Ybarra and Thompson (2018) argue that sexual violence results from the presence of several factors that necessitate further examination to enhance our comprehension of its occurrence as well as enhancing existing preventative approaches. In response, several theoretical frameworks have been applied to better understand sexual violence victimization as well as its perpetration by adolescents. Yet, only one study thus far has applied general strain theory as a theorical guide to assess our understanding of the risks associated with the perpetration of adolescent sexual violence. Given the fact that a significant amount of research aimed at understanding juvenile delinquency through the lens of general strain theory continues to expand, the application of this framework to enhance our understanding of the perpetration of sexual violence among adolescents is lacking.
General Strain Theory
Departing from traditional explanations of criminal conduct that involve social structural mechanisms, Agnew’s (1992) revival of strain theory emphasizes various sources of individual strain. Whereas the earlier versions of strain theory centered primarily on class-based impediments toward achievement of monetary success (Cloward & Ohlin, 2013; Cohen, 1955; Merton, 1968), GST recognizes the social-psychological bases of strain and the numerous adaptations that may or may not involve criminal coping. In describing his core theoretical concept, Agnew (1992, p. 48) proposes that when individuals experience poor treatment within negative social relationships, this fuels negative emotions such as anger and frustration which can lead to engagement in delinquency/crime.
Other core theoretical concepts included within the GST framework are objective and subjective strains. Objective strains are described as being the sort of events or conditions that may be assumed to be disliked by most individuals (Agnew, 2001, pp. 320–321). Objective strains can include, for example, physical assault or the lack of adequate food or shelter. Subjective strains involve events and conditions that are disliked by the individual who is experiencing them or has previously experienced them (Agnew, 2001). A key finding that arose from studies that examined stress demonstrates that the subjective evaluation of objective strains is likely to be different across individuals (e.g., Wheaton, 1990). How an individual may cognitively appraise or respond to certain strains such as divorce or a death in the family is going to differ based on factors related to the individual (e.g., irritability and self-esteem), the social (e.g., social support), and/or circumstantial events of one’s life (Agnew, 2001).
According to Agnew (1992), there are three major categories of strains: (1) the failure, or the anticipated failure, toward achievement of positively valued goals (e.g., failure to achieve status or money-related goals), (2) the removal, or the anticipated removal, of positively valued stimuli (e.g., break-up with a romantic partner), and (3) the presence of, or the anticipated presence of, a noxious stimuli (e.g., verbal and physical abuse). Agnew (2001) argues that some strains are more likely to trigger a criminal response. These types of strains are identified as high in magnitude, viewed as unjust, associated with low levels of social control, and create pressure or incentive to engage in crime. Strains that are more likely to result in crime include, for example, parental rejection, harsh and erratic discipline, child abuse and neglect, homelessness, criminal victimization, and gender/racial discrimination.
Prior empirical studies have examined the theoretical propositions of GST across diverse populations which has generated wide support for the theory. Application of the GST perspective has also been applied to explain participation in various forms of delinquency and crime. However, to the authors’ knowledge, there has only been a single investigation of the strains experienced by adolescents and their connection to risks which may increase sexual offending. Colbert (2004) assessed several domains of risk including personality, psychological disorders, substance abuse, family factors (e.g., income level and witnessing violence), and peer interactions across delinquent groups. The 2001 Oklahoma Juvenile Data (n = 274) was used to examine these risks among two adolescent groups comprised of either violent offenders or those who perpetrated sexual offenses. Findings demonstrated that adolescent strains such as qualifying for reduced cost or free lunches, child maltreatment, negative interactions with peers (being hassled by peers) increased the odds of sexual offending. These results suggest that perpetration of sexual offenses may arise as a form of delinquent coping due to the vulnerabilities that arise with the experience of external stressors. However, it is evident that investigation of risks and their influence on juvenile sexual offending requires clarification. The present study attempts to address this limitation with an exploration of the risks associated with the juvenile sexual offending, which includes demographic predictors in addition to risk factors formulated from the basic tenets of GST.
Predictors of Sexual Violence Perpetration Among Adolescents
Demographic Predictors
Empirical studies on adolescent sexual violence perpetration have identified several demographic characteristics including age, biological sex, and race/ethnicity, that can impact its likelihood. The role of age as a predictor for sexual violence perpetration remains vague (Ngo et al., 2018; Ybarra & Thompson, 2018). For instance, Ybarra and Thompson (2018) used longitudinal data from a cohort of 1,586 youths aged 10 to 21 years to assess risk factors for the emergence of sexual violence in adolescence and discovered that age was positively associated with sexual assault perpetration. Conversely, Ngo et al. (2018) analyzed cross-sectional survey data over 4 years from a sample of 4,665 U.S. middle and high students and discovered that older students (i.e., 11th and 12th graders) exhibited a lower likelihood of engagement in sexual violence.
Research has also investigated the influence of biological sex on the likelihood of sexual violence perpetration, which has consistently shown that most adolescent sexual perpetrators are male (Fox, 2017; Gewirtz-Meydan & Finkelhor, 2020; Ngo et al., 2018; Ueda, 2017). For example, Ngo et al.’s (2018) investigation of adolescent sexual violence evidenced that males were more likely than females to report committing sexual violence. Furthermore, the impact of race/ethnicity on the perpetration of sexual violence remains unclear (Casey & Masters, 2017; Ngo et al., 2018; Ybarra & Thompson, 2018). Ybarra and Thompson’s (2018) examination of a youth cohort found no statistically significant relationship between race/ethnicity and committing sexual violence.
General Strain Factors
Failure to Achieve Goals
General strain theory has been utilized to understand the impact that certain strains experienced by adolescents have on delinquency and violence. Previous research has highlighted that academic achievement acts as a protective factor against adolescent sexual assault perpetration (Casey & Masters, 2017; Tharp et al., 2013; Vizard, 2013). For instance, Casey and Masters (2017) systematic review of sexual violence risk and protective factors demonstrated that male youth who had higher levels of academic achievement were less prone to engage in sexual violence. Research by Siria et al. (2020) on the characteristics of juvenile sexual offenders utilizing a sample of 73 incarcerated Spanish adolescents for committing a sexual offense demonstrated that roughly one third repeated a school year while and about two thirds repeated a school year more than once.
Although research on the impact of out-of-school suspension on adolescent sexual violence perpetration is scant, Tharp et al.’s (2013) systematic qualitative review of 191 studies assessing risk factors for sexual violence perpetration showed a mixed influence of school/conduct problems on its occurrence.
Loss of Positive Stimuli
Research on the impact of experience with the foster care system, homelessness, or having incarcerated parents and the propensity to engage in sexual violence is scant. However, meta-analyses or systematic reviews of the associations between youth homelessness, sexual offenses, sexual victimization, and sexual risk behavior have not demonstrated a clear association between homelessness and perpetrating sexual violence among youth (Heerde et al., 2014, 2015; Heerde & Hemphill, 2016).
Presentation of Negative Stimuli
The impact of exposure to household violence among adolescents as a risk factor for sexual violence perpetration remains unclear (Casey & Masters, 2017; Tharp et al., 2013; Vizard, 2013; Ybarra & Thompson, 2018). For instance, Tharp et al.’s (2013) systematic qualitative review of studies assessing risk factors for sexual violence perpetration revealed that prior exposure to household violence was strongly associated with its occurrence. These results align with Ybarra and Thompson’s (2018) assessment of sexual violence perpetration among youth (N = 1,586), affirming that adolescents exposed to household violence were more prone to perpetrate sexual violence. However, Fox’s (2017) assessment of identifying sexual and non-sexual juvenile offenders among 64,329 youth in Florida evidenced that witnessing violence at home was a significant predictor of non-sexual offending among the youth. It is also important to note that effects of exposure to household violence on sexual violence perpetration may be the result of the varying compositions of the samples studied.
Another important theme in the literature is the impact that parental abuse (i.e., physical and psychological) has on subsequent sexual violence in adolescence (Casey & Masters, 2017; Siria et al., 2020; Tharp et al., 2013; Vizard, 2013). For instance, Casey and Masters (2017) systematic review of previous studies that assessed risk and protective factors of perpetrating sexual violence evidenced a consistent relationship between physical or psychological abuse and sexual aggression in adolescence. These outcomes align with Siria et al.’s (2020) recent analysis of incarcerated juvenile sexual offenders (N = 73) in Spain that showed a significant proportion of these youth reported emotional abuse and neglect. Conversely, Fox’s (2017) study of juvenile offenders found that emotional abuse and physical abuse were not correlated with sexual or non-sexual offending. This outcome is interesting since Agnew (2013) has noted that from a GST perspective, child abuse is a severe form of strain likely to result in delinquency.
Research regarding the impact of dating violence on committing sexual violence in adolescence is limited. For example, Tharp et al. (2013) systematic review of risk and protective factors for sexual violence perpetration noted that specific characteristics of a perpetrator’s intimate relationships generally act as indictors of committing sexual violence. However, Tharp et al. (2013) note that many studies that explored this relationship have done so using data from adult males.
Furthermore, the cycle of violence model contends that forms of abuse and trauma during childhood can increase the odds of violent behaviors (see DeLisi et al., 2014). This assumption is germane to sexual abuse victimization and its impact on subsequent sexual violence perpetration has been noted in the literature (Casey & Masters, 2017; DeLisi et al., 2014; Fox, 2017, Siria et al., 2020). For example, DeLisi et al. (2014) utilized data from 2,520 incarcerated male delinquents from a large southern state to assess whether childhood sexual abuse victimization impacted subsequent juvenile sex offending and found that they were significantly associated. Similar research by Casey and Masters (2017), Fox (2017), and Siria et al. (2022) all demonstrated that past sex abuse was correlated with subsequent sexual violence. Finally, previous studies have highlighted that adolescent who bully their peers also engage in sexual violence (Espelage et al., 2015, 2018). However, research that assesses the impact of various forms of bullying victimization on committing sexual violence is lacking.
The Current Study
In the context of this exploratory study, understanding the dynamics of demographics as well as the impact of strains experienced by youth and their impact on sexual violence perpetration is important to explore from a general strain theory perspective. Applying this framework, the current study is guided by the following two questions: (1) Do factors grounded in general strain theory predict sexual violence perpetration among adolescents? (2) Do sex-based differences exist in perpetrating sexual violence among adolescents from a general strain theory perspective?
Methods
Data
Since 1989, the Minnesota Student Survey (MSS) has collected cross-sectional data on student health and safety with participants ranging in ages from 10 to 18 years. As a collaborative project with the Departments of Education, Human Services, Health, and Public Safety, the MSS provides comprehensive information about students’ health, well-being, as well their appraisal of school and home environments (Minnesota Department of Education, 2018). Over the years the MSS has expanded its collection of data on academic achievement, housing status, family relationships, and forms of adolescent victimization. The current study explores data derived from the 2022 MSS (N = 135,447). However, 49,160 cases from adolescents aged 10 to 13 years were excluded due to a lack of variation in sexual violence perpetration. Additionally, 1,283 cases with missing data were removed, resulting in a final sample of 85,004 responses.
Measures
The primary variables of interest in the current study include demographic factors, and various forms of strain and victimization. The measures applied in this study were specifically developed from survey items on measures grounded in the core propositions of general strain theory as well as the conditions around sexual assault perpetration revealed by adolescents’ responses to the MSS questions.
Dependent Variables
A measure of adolescent sexual assault perpetration was constructed from the following survey measure: “Have YOU ever pressured, tricked, or forced someone to do something sexual, or have you done something sexual to someone against their wishes?” Responses to this question were grouped to produce sexual violence (0 = No, 1 = Yes). This variable is measured at the nominal level to indicate whether adolescents have engaged in sexual violence This measure is also congruent with previous studies that examined sexual violence perpetration among adolescents (Basile et al., 2018; Krahé & Berger, 2017; Ngo et al., 2018; Tharp et al., 2013). Moreover, conducting multivariate logistic regression analyses to assess the impact that demographic factors as well as various forms of strain have on sexual violence perpetration necessitated the use of a dichotomous outcome measure.
Independent Variables
Demographic Measures
Several demographic variables were produced from students’ responses to questions asked during the 2022 MSS. Age indicates the age of the respondent during the distribution of the survey, ranging from 10 to 18 years of age and was measured at the continuous level. Two demographic characteristics were also developed at the nominal-level. These include measures of biological sex (0 = Female, 1 = Male), where female is the reference category, and responses to a question concerning race and ethnicity were aggregated to produce race/ethnicity (0 = White, 1 = Black, 2 = Asian, 3 = Hispanic, or 4 = Other), where White is the reference category.
Measures of Strain
First, two measures related to an individual’s failure to achieve their goals were created. Similar to prior assessments of school performance and strain (Sigfusdottir et al., 2012), a measure to identify students’ academic standing was produced from the following question: “How would you describe your grades this school year?” This measure captured responses on a 5-point scale (1 = Mostly A’s, 2 = Mostly B’s, 3 = Mostly C’s, 4 = Mostly D’s, and 5 = Mostly F’s). The responses were rearranged to generate an ordinal measure of academic achievement (1 = Mostly F’s, 2 = Mostly D’s, 3 = Mostly C’s, 4 = Mostly B’s, and 5 = Mostly A’s). Next, distinguishing students who were suspended from school was developed from the following survey item: “What are the reasons you missed a full or part of a day of school in the last 30 days? (Suspended from school).” The responses to this question were used to generate suspended (0 = No, 1 = Yes).
Second, three measures related to the loss of positive stimuli were created. Due to poor outcomes associated with youth in state care (Barn & Tan, 2012), the impact dislocation from family was assessed. Identifying students who lived in foster care stemmed from the survey question: Have you ever been in foster care? (During the past 12 months). The responses to this question were used to create foster care (0 = No, 1 = Yes). Discerning whether students experienced homelessness was established from two questions: “During the past 12 months, have you stayed in a shelter, somewhere not intended as a place to live, or someone else’s home because you had no other place to stay? (w/ parents or adult family member).” The same question was also presented but solicited whether the respondent has a similar experience without parents or adult family members. The responses to these questions were aggregated to generate homelessness (0 = No, 1 = Yes). An indicator of students who had a parent or guardian incarcerated was constructed from the following two survey items: Have any of your parents or guardians ever been in jail or prison? (I have a parent or guardian in jail or prison right now). The same question was also presented but solicited whether the respondent had an incarcerated parent or guardian in the past. The responses to these questions were grouped to generate incarcerated parents (0 = No, 1 = Yes).
Next, eight measures related to the presentation of negative stimuli were created.
Based on previous studies to assess abuse and conflict in the family, (see e.g., Sigfusdottir et al., 2012) a measure of exposure to violence in the student’s household was developed from the following question: “Have your parents or other adults in your home ever slapped, hit, kicked, punched or beat each other up?” The responses to this question were utilized to create exposure to domestic violence (0 = No, 1 = Yes). Students’ experiences of abuse (i.e., verbal and physical) at home is composed of two questions: “Does a parent or other adult in your home regularly swear at you, insult you or put you down?” and “Has a parent or other adult in your home ever hit, beat, kicked or physically hurt you in any way?” The responses to these questions were aggregated to develop parental abuse (0 = No, 1 = Yes). Adolescent sexual abuse was produced from two survey items: “Has anyone who was not a relative/family member ever pressured, tricked, or forced you to do something sexual or done something sexual to you against your wishes?” and “Has any relative/family member ever pressured, tricked, or forced you to do something sexual or done something sexual to you?” The responses to these survey items were grouped to produce sexual abuse (0 = No, 1 = Yes).
Lastly, measures of bullying victimization were also used to represent the introduction of negative stimuli. Based on previous studies that assess various forms of bullying victimization (Hay & Meldurm, 2010), a measure to identify students who experienced physical bullying was produced from the following question: “During the last 30 days, how often have other students at school pushed, shoved, slapped, hit, or kicked you when they weren’t kidding around?” This measure captured responses on a 5-point scale (1 = Never, 2 = Once or twice, 3 = About once a week, 4 = Several times a week, and 5 = Every day) to generate an ordinal measure of physical bullying. An indicator to identify students who experienced verbal bullying was produced from following questions: “During the last 30 days, how often have other students threatened to beat you up?” and “During the last 30 days, how often have others students at school made sexual jokes, comments, or gestures toward you?” This measure captured responses on a 5-point scale (1 = Never, 2 = Once or twice, 3 = About once a week, 4 = Several times a week, and 5 = Every day) and were aggregated to generate an ordinal measure of verbal bullying.
A measure to identify students who experienced social bullying was produced from following two questions: “During the last 30 days, how often have other students at school spread mean rumors or lies about you? and “During the last 30 days, how often have other students as school excluded you from friends, other students, or activities?” This measure captured responses on a 5-point scale (1 = Never, 2 = Once or twice, 3 = About once a week, 4 = Several times a week, and 5 = Every day) and were aggregated to generate an ordinal measure of social bullying. Finally, a measure to identify students who experienced cyber bullying was produced from the following question: “During the last 30 days, how often have you been cyberbullied? (Count being bullied through texting, Instagram, Snapchat, TikTok, or other social media).” This measure captured responses on a 5-point scale (1 = Never, 2 = Once or twice, 3 = About once a week, 4 = Several times a week, and 5 = Every day) to generate an ordinal measure of cyber bullying.
Analytical Strategy
Multivariate logistic regression models were used to explore the effect that demographic factors and various forms of strain experienced by adolescents have on sexual violence perpetration as well as whether sex differences exist among adolescents who perpetuate sexual violence. Given that adolescent sexual violence perpetration is a rare event in the 2022 MSS data (see Table 1), Firth’s (1993) logistic regression model for rare events was applied rather than traditional logistic regression models to minimize bias in the outcomes of these analyses. Therefore, Firth’s Bias-Reduced Logistic Regression R package (version 1.26.0), which includes data transformation functions, was utilized. All analyses were conducted using R (version 4.3.3) with RStudio (version 2024.04.1-748). Logistic regression models for rare events were applied to three distinct samples from the 2022 MSS dataset. Model 1 employed data from the entire sample of students from the 2022 MSS (N = 85,004) that included both female and male adolescents. Model 2 used data from a subsample of female adolescents (n = 41,993). Model 3 utilized data from a subsample of male adolescents (n = 43,011).
Descriptive Statistics for Study Measures (N = 85,004).
Results
Sample Characteristics
Table 1 illustrates demographic characteristics and forms of strain experienced by adolescents in the sample used for this study (N = 85,004). The mean age among the 85,004 students was approximately 15 years old (SD = 1). Females and males were proportionately represented in the sample (roughly 50%), approximately 67% of the respondents identified as being White. Regarding forms of strain, specifically failed to achieve goals, about 76% of students reported above average grades (i.e., Mostly A’s and B’s) and most students (roughly 99%) reported not being suspended from school. Table 1 also shows that most students did not report being in foster care, experiencing homelessness, or having incarcerated parents (approximately 99%, 97%, and 87%, respectively). Regarding the presentation of negative stimuli, roughly 95% of students did not report exposure to domestic violence and well as not experiencing parental abuse, dating violence, and sex abuse (approximately 84%, 86%, and 93%, respectively). Finally, Table 1 demonstrates that most students reported never experiencing physical, verbal, social, and cyber bullying victimization (roughly 93%, 85%, 87%, and 86%, respectively).
Multivariate Logistic Regression Outcomes
Multivariate logistic regression models were applied to a sample of adolescents from the 2022 MSS to explore the impact of demographic characteristics and forms of strain have on the perpetration of adolescent sexual violence. Model 1 used data from the entire sample (N = 85,004) to answer the first research question: Do factors grounded in general strain theory predict sexual violence perpetration among adolescents? Table 2 demonstrates that Model 1 was statistically significant and explained roughly one fifth of the variance in adolescent sexual violence (R 2 = .183).
Multivariate Logistic Regression on LET Variables to Predict Sexual Violence Perpetration Among Adolescents.
p < .05. **p < .01. ***p < .001.
Regarding demographic factors, age emerged as a statistically significant predictor of sexual violence (Exp(B) = 1.403, p < .001), indicating that older respondents were more prone to perpetrate sexual violence compared to their younger counterparts. Male adolescents were two times more likely (Exp(B) = 2.084, p < .001) to commit sexual violence compared to females. Students who reported living in the foster care system were three times more likely (Exp(B) = 3.00, p < .001) to commit sexual violence compared their peers who had not contact with the foster care system. Additionally, students who reported being homeless were more prone to perpetrate sexual violence (Exp(B) = 1.765, p < .001) in comparison to their peers who had not experienced homelessness.
Table 2 also demonstrated that students who reported parental abuse (Exp(B) = 1.598, p < .001) were more likely to engage in sexual violence compared to those who did not experience this type of abuse. With respect to dating violence, students who reported this type of victimization were five times more likely engage in sexual violence (Exp(B) = 5.038, p < .001) compared to their peers that did not experience this form violence. Students who reported experiencing sexual abuse were two times more likely (Exp(B) = 2.137, p < .001) to commit sexual violence in contrast to students who did not experience this form of violence. Regarding the measures of bullying victimization, students who reported verbal, social bullying, and cyber bullying were all more likely to perpetrate sexual violence compared to adolescents who did not report similar types of victimization (see Table 2).
Models 2 and 3 were established to answer the second research question: Do sex-based differences exist in perpetrating sexual violence among adolescents from a general strain theory perspective? Table 2 demonstrates that Model 2 was statistically significant and explained roughly one sixth of the variance in adolescent sexual violence (R 2 = .159), while Model 3 was statistically significant and explained one fourth of the variance in adolescent sexual violence (R 2 = .214). Although there were similarities regarding the significant associations of some of the factors and their impact on sexual violence used in this study, some differences were observed. For example, Table 2 demonstrates that age was a significant predictor of perpetrating sexual violence among both the subsamples of female (Model 2) and male adolescents (Model 3). Moreover, living in the foster care system and experiencing homelessness were significant predictors of perpetrating sexual violence among both the subsamples of female (Model 2) male adolescents (Model 3). Interestingly, among the subsample of males (Model 3), exposure to domestic violence was a significant predictor of sexual violence (Exp(B) = 1.466, p < .05) which was not observed in the full sample (Model 1) or subsample of female (Model 2) adolescents.
Table 2 illustrates that experiencing parental abuse was a significant predictor of committing sexual violence among both the subsamples of female (Model 2) and male adolescents (Model 3). Moreover, experiencing dating violence was a significant predictor of perpetrating sexual violence among the subsamples of female (Model 2) and male (Model 3) adolescents. However, the odds were higher for the subsample of females in Model 2 (Exp(B) = 6.503, p < .001) compared to the subsample of male adolescents in Model 3 (Exp(B) = 4.113, p < .001). Sexual abuse was also found to be significant predictor of perpetrating sexual violence among both subsamples of female (Model 2) and male (Model 3) adolescents. Interestingly, the odds were higher for the subsample of males in Model 3 (Exp(B) = 2.964, p < .001) compared to the subsample of female adolescents in Model 2 (Exp(B) = 1.701, p < .001). Finally, verbal, social, and cyber bullying were all found to be significant predictors of committing sexual violence among the subsample of male adolescents (see Model 3). However, these outcomes were not observed among the subsample of female adolescents (see Model 2).
Discussion
Drawing from GST and employing multivariate logistic regression models, this study aimed to explore the impact of demographic characteristics of adolescents, their failure to achieve goals, loss of positive stimuli, and exposure to negative stimuli in perpetrating of sexual violence. All three models utilized in the current study were significant in sexual violence preparation among adolescents, suggesting that some measures of GST are important contributors of sexual violence perpetration.
The multivariate logistic regression models in Table 2 revealed that age was a significant risk factor across all three models, indicating that older adolescents were at significantly higher risk of committing sexual violence compared to their younger counterparts. Although Ybarra and Thompson’s (2018) study showed a strong positive correlation between age and the perpetration of sexual violence, Ngo et al.’s (2018) analysis found no statistically significant relationship between age and perpetrating sexual violence. Thus, the current finding significantly contributes to the literature on the association between adolescents’ age and sexual violence perpetration. However, more research is needed to clarify this association.
Sex was a significant risk factor for sexual violence perpetration in Model 1, indicating that adolescent males had a greater likelihood to perpetrate sexual violence compared to their female peers. This outcome is congruent with previous investigations into juvenile sexual violence (Fox, 2017; Gewirtz-Meydan & Finkelhor, 2020; Ngo et al., 2018; Ueda, 2017). Results from the current study also show that across all three models, race/ethnicity was not a significant predictor of perpetrating sexual violence among adolescents. Although race was not a significant factor in any of the three models, these findings echo the mixed results evidenced in previous studies (Casey & Masters, 2017; Ngo et al., 2018; Ybarra & Thompson, 2018).
In assessing the impact of the core propositions of general strain theory, some of the measures of strain emerged as significant predictors of sexual violence among adolescents. For example, all three models in Table 2 demonstrated that adolescents’ experiences with the foster care system and homelessness increased the risk of sexual violence perpetration. Therefore, this finding significantly contributes to the literature on the association between adolescent experience with the foster care system and homelessness on sexual violence perpetration. However, more research is needed to clarify this association since previous studies did not demonstrate whether these factors directly impact the perpetration of sexual violence among adolescents (Heerde et al., 2014, 2015; Heerde & Hemphill, 2016).
Table 2 illustrates the association between parental abuse and the heightened risk of sexual violence perpetration was identified in all three models and aligns with previous studies (Casey & Masters, 2017; Siria et al., 2020; Tharp et al., 2013; Vizard, 2013). Interestingly, Table 2 demonstrates that dating violence victimization increased the risk of sexual violence perpetration and provides a unique contribution to the literature since research on this association is scant. The association between sex abuse victimization and committing sexual violence was demonstrated in all three models (see Table 2) and is consistent with previous studies on the subject (Casey & Masters, 2017; DeLisi et al., 2014; Fox, 2017, Siria et al., 2020).
Finally, although research on the relationship between engagement in bullying and perpetrating sexual violence exists (Espelage et al., 2015, 2018), this study contributes uniquely to the literature on adolescent sexual violence in that some forms of bullying (i.e., verbal, social, and cyber) victimization were found to heighten the risk of committing sexual assault in the full sample (Model 1) and among the subsample of male adolescents (Model 3) in this study. Thus, this finding demonstrates the need for further exploration of this relationship. More importantly, the outcomes of this study showed a heightened risk of sexual offending in response to some forms of strain which aligns with Agnew’s (2013) stance on general strain theory and its relationship to violence. These forms of prior victimization (i.e., parental abuse, dating violence, sexual abuse, and bullying) also aligns with the cycle of violence model (see DeLisi et al., 2014) that asserts prior experiences with abuse and trauma during childhood increases the risk of behaviors like sexual violence.
Limitations
Despite this study’s contributions to the literature on predictors of sexual violence perpetration among adolescents, several limitations should be discussed. First, it is important to acknowledge that the Minnesota Student Survey’s sample is drawn exclusively from adolescents attending middle or high schools in Minnesota, which limits the generalizability of the findings to other states. Moreover, the cross-sectional data utilized in this study limits the ability to infer causality, which would be more evident with longitudinal data. Nevertheless, the current study provides unique and robust models that contribute to the existing literature on adolescent sexual violence.
In addition, the data assessed in the current study consists of self-reported responses. Due to the sensitive nature of certain questionnaire items, such as questions about contact with the child welfare system, experiences of homelessness, or types of victimization, some adolescents may have been hesitant to provide truthful responses. In the case of sexual assault, there is a rather substantial limitation in that some individuals who have committed sexual assault may not accept that their behavior constitutes sexual assault. For example, engaging in sexual activity with another person in the absence of affirmative consent may constitute sexual assault, yet the offender may genuinely believe that it was consensual because there was not a verbal and/or physical rejection. Thus, the question used to form the dependent variable in the current study does not simply measure which youth in the sample committed sexual assault, but rather the subset of the group who have cognitively accepted that they have committed sexual assault.
Conclusion
Findings from this study underscore critical areas for policy implementation aimed at minimizing adolescents’ risk of committing sexual violence. For example, DeGue et al.’s (2014) systematic review of 140 evaluations of sexual violence preventative programs found only three programs, which utilized thorough controlled evaluation models, effectively prevented sexual violence perpetration. It is worth noting that these studies varied in their demographic composition, ranging from youth to young adults. Yet, Schneider and Hirsch’s (2020) recent assessment of effective sexual violence prevention strategies highlight the potential for K-12 comprehensive sexuality education, influenced by the National Sexuality Education Standards.
Schneider and Hirsch (2020) argue that while sex education has traditionally focused on reducing unplanned pregnancies and sexually transmissible infections, as well as informing youth about health risk behaviors, incorporating discussions on sexual violence perpetration can also be beneficial for students. The researchers acknowledge that some school administrators or parents may resist incorporating sex education to current school curriculum for young students, such as kindergarteners, but emphasize the need for comprehensive approaches. Schneider and Hirsch (2020) reference data from RAINN (Rape, Abuse & Incent National Network) that highlights the vulnerability of youth to sexual violence by peers or adults.
Informing students about this form of brutality could expand awareness and prevent both victimization and perpetration. Additionally, continued evaluations of existing sexual violence prevention programs, which address recognized risk factors identified in the extant literature and supported by this study, are necessary to minimize victimization and perpetration among adolescents.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
