Impacts of Low Self-control and Opportunity Structure on Cyberbullying Developmental Trajectories: Using a Latent Class Growth Analysis

Abstract

This study tested self-control and opportunities theories to examine cyberbullying developmental trajectories through the estimation of a latent class growth analysis. Data from a 6-year longitudinal study of middle- and high-school students from South Korea were analyzed to examine if there are unique growth trajectories for cyberbullying perpetration when accounting for low self-control and opportunity factors. Results suggest that there are three distinct subgroups: (1) a normative trajectory group, (2) an increasing and late-peak group, and (3) an early onset and decreasing group. Low self-control was found to be significantly associated with early onset/decreasing cyberbullying. Opportunity to utilize cyberspace was significantly related with increasing/late peak cyberbullying but did not significantly mediate the effect of low self-control on class membership.

Keywords

cyberbullying developmental trajectories cyber risky lifestyles low self-control latent growth mixture modeling

Introduction

Cyberbullying has become a significant social problem primarily affecting youth over the past decade (Kowalski et al., 2014; Zych et al., 2018). Prior research examining, the impact of cyberbullying has indicated that cyberbullying is linked to various detrimental outcomes such as: (a) feelings of sadness, anger, and depression (Mishna et al., 2010); (b) lower self-esteem (Patchin & Hinduja, 2010); (c) behavioral problems (Kokkinos et al., 2014); and/or (d) suicide (Hinduja & Patchin, 2010).

While it is clear that there are numerous undesirable outcomes associated with cyberbullying victimization, theoretical and scientific explanations addressing the underlying mechanisms responsible for cyberbullying perpetration still leave much room for question (Kowalski et al., 2014; Slonje et al., 2012). Nevertheless, there is a developing body of literature that addresses the commonalities of individuals linked to cyberbullying activities. For example, Guo’s (2016) meta-analysis of 77 studies examining the correlates of cyberbullying perpetration determined that cyberbully perpetrators tend to be, amongst other things: (1) older males, (2) involved in other forms of bullying offline, (3) associated with behavioral problems, (4) frequently online, (5) morally and ethically inept, (6) from unstable households with high conflict and low supervision, and (7) susceptible to deviant peer groups (pp. 441–442).

The classification for cyberbullies developed by Guo (2016) offers a comprehensive and descriptive review of the contextual factors associated with those who are likely to cyberbully. Others have examined motivational factors associated with cyberbullying perpetration, such as Mishna et al. (2010) who found that cyberbullying activities appears to make individuals feel: (a) powerful, (b) popular, and (c) have feelings as though they are funny. Conversely, Rice et al. (2015) found that certain opportunity factors are significantly correlated with cyberbullying, such as frequent use of social media and texting. Finally, Fletcher et al. (2014) found that cyberbullying perpetration is highly associated with: (a) aggressive behaviors in school, (b) a reduced quality of life, (c) psychological issues including diminished mental well-being, and (d) social and peer problems.

Although the factors previously assessed shed some light as to the variables/traits that are likely associated with cyberbullying perpetration, one of the larger issues with previous research in this domain is the lack of theoretical grounding to explain the behavior. Hagan (2012) stated that “without the generation of useful theoretical explanations, a field is intellectually bankrupt; it becomes merely a collection of ‘war stories’ and carefully documented encyclopedic accounts” (p. 21). Thus, it is imperative to integrate useful theoretical propositions when attempting to explain the underlying factors associated with cyberbullying behaviors. By integrating theory into this line of research, future studies will be better equipped to address the issue and facilitate practitioners who attempt to institute effective mechanisms to mitigate cyberbullying activities (see Andrews et al., 1990).

The current study advances the cyberbullying literature by integrating propositions from self-control and opportunity theories to assess if these factors are relevant to cyberbullying perpetration. Ascertaining such knowledge can be beneficial toward integrating theoretically grounded methods to reduce cyberbullying within the population (Akers & Jensen, 2008; Andrews et al., 1990). While there is a wealth of research pertaining to cyberbullying within the U.S., the current study relies on a longitudinal dataset of Korean adolescents across the middle- and high-school period to estimate the developmental trajectories of individuals involved in cyberbullying behaviors. Previous research pertaining to bullying behaviors in predominantly Asian countries has indicated that bullying statistics tend to be similar between Eastern and Western societies (Moon et al., 2011). Additionally, theoretical models developed in Western societies (i.e., differential association, low self-control, general strain, etc.) have proved useful in explaining bullying behaviors within a unique South Korean context (Cho, 2018; Cho et al., 2019; Cho & Lee, 2018; Chui & Chan, 2015; Moon & Alarid, 2015; Moon et al., 2011, 2012). Further, this study highlights the need to consider the heterogeneity of the sample by offering important developmental insights into a finite mixture of unobserved, homogeneous subgroups, each having distinct patterns of cyberbullying developmental trajectories. Thus, the study employs a new approach, the latent group-based trajectory analysis technique (latent class growth analysis) that incorporates a categorical latent trajectory variable that represents latent trajectory classes. The latent trajectory classes are derived from maximum likelihood estimation, meaning that individuals were classified into mutually exclusive subgroups in which they have the greatest probability of membership (Nagin, 2005). Thus, the current study draws from self-control theory and opportunities theories to examine the effect of self-control and opportunity factors that vary on latent trajectory classes. The results and limitations of the analyses conducted here are discussed.

Literature Review

Although there is a wealth of descriptive research explaining cyberbullying perpetration (Guo, 2016; Rice et al., 2015; Zych et al., 2018), this line of inquiry would greatly benefit from theoretical development within the context cyberbullying in order to guide societal responses which have the ability to inhibit the issue of cyberbullying within society. The current study draws from self-control (Gottfredson & Hirschi, 1990) and Lifestyles and Routine Activity theories (Cohen & Felson, 1979) to explain why and how cyberbullying perpetration transpires.

Self-control Theory

Gottfredson and Hirschi (1990) wrote the seminal work General Theory of Crime which offers the tenants associated with self-control theory. According to Gottfredson and Hirschi (1990), self-control is the ability of people to forego acts that provide immediate pleasure for an individual because the individual has the ability to understand the harm that may be linked to long-term interests when acting to fulfill immediate pleasures. Contrary to self-control, low self-control is associated with a lack of restraint, and it is conducive to impulsive behaviors, self-centeredness, short-tempers, risky behaviors, and other socially undesirable traits (Gottfredson & Hirschi, 1990). A prominent factor associated with low self-control is low self-control’s association with antisocial behavior and the tendency for individuals manifesting this trait to engage in delinquent or deviant behaviors (Gottfredson & Hirschi, 1990). Although control theories are widely tested and accepted when explaining delinquent and/or criminal behavior, only limited scholarship involving cyberbullying activities and self-control have come to fruition (Holt et al., 2012; Kim et al., 2017; Li et al., 2016; Vazsonyi et al., 2012).

Low self-control is likely a precursor to cyberbullying perpetration because individuals possessing low self-control seek immediate pleasure without considering the long-term effects of their actions (Gottfredson & Hirschi, 1990; Pratt & Cullen, 2000; Tittle et al., 2003). That being said, previous research has found that traditional bullying behaviors are associated with low self-control (Chui & Chan, 2013; Moon & Alarid, 2015; Unnever & Cornell, 2003). Additionally, traditional bullying activities tend to be significantly associated with cyberbullying perpetration (Guo, 2016; Kim et al., 2017). Thus, it is logical to integrate this theory in tests of cyberbullying perpetration as done by relatively few scholars.

To assess whether or not cyberbullying activities are associated with low self-control, Vazsonyi et al. (2012) examined the effects of low self-control on cyberbullying perpetration using a sample of youth from 25 European countries. The authors incorporated path models to examine both the direct and indirect effects of low self-control on cyberbullying perpetration and victimization. Results suggested that low self-control both directly and indirectly effects cyberbullying perpetration and victimization when accounting for the mediating effects of externalizing behaviors, offline victimization, and offline perpetration (Vazsonyi et al., 2012). The authors found differential effects between males and females and noted that there are contextual—or opportunity factors associated with access to cyberspace—which differed between genders in the sample used. As a result, Vazsonyi et al. (2012) concluded that it is important to account for the opportunity structure when analyzing cyberbullying activities because not all individuals have the same capacity to access cyberspace.

In a more recent test of low self-control on cyberbullying perpetration, Li et al. (2016) examined the effects of low self-control and social learning on cyberbullying activities using a sample of middle- and high-school students from Kentucky. The authors found that both theories significantly explained cyberbullying perpetration, and low self-control tends to be associated with delinquent peer associations which, in turn, increased the likelihood of cyberbullying activities. In contrast to Vazsonyi et al. (2012), the authors in this study found conflicting support for the notion that access to technology increases the likelihood for cyberbullying behaviors with blogging being the only technological measure that significantly affected cyberbullying perpetration across models (Li et al., 2016).

Overall, Gottfredson and Hirschi’s (1990) General Theory of Crime offers a practical explanation for the motivation of why socially undesirable activities occur (i.e., low self-control). Nevertheless, Gottfredson and Hirschi (1990) also acknowledged that the opportunity structure is a relevant mechanism which needs to be present for deviant behavior to occur. The authors noted that persons with low self-control are more likely to become involved with deviant behaviors when the opportunity is present; thus, overlooking the effects of opportunity structure may result in an improperly defined analysis (Cho & Lee, 2018; Gottfredson & Hirschi, 1990; Li et al., 2016; Vazsonyi et al., 2012).

Opportunity Perspective

Lifestyles and routine activities theories (LRAT) addresses the opportunity structure as it relates to crime and deviance in order to examine how involvement in risky lifestyles or behaviors, along with (1) motivated offenders, (2) suitable targets, and (3) capable guardianship, factor into the victim and offender dynamic (Cohen & Felson, 1979; Hindelang et al., 1978; Miethe & Meier, 1990). Cohen and Felson (1979) theorized that when the aforementioned factors converge in time and space, a victimizing event is likely to occur because all of the components necessary are present.

When discussing the notion of cyberbullying perpetration, an individual must have access to a platform which allows for them to bully in that domain; otherwise the opportunity to do so is nonexistent. This nuance becomes more conflictual because youth tend to be responsible for the majority of bullying incidents (U.S. Department of Education, 2011, 2018), and parents may be more likely to regulate the opportunity for youth to access cyberspace. Additionally, the contextual factor of potential victim may be limited due to socioeconomic, household, and relational factors associated with the ability to be victimized in cyberspace (Arntfield, 2015).

Nevertheless, various risk factors associated with cyberbullying perpetration have been observed through the amount of time youth spend using a computer each day (Ang, 2015; Mishna et al., 2012; Navarro & Jasinski, 2012). For example, Mishna et al. (2012) analyzed middle and high school students and found that students who were engaged in cyberbullying behaviors utilized computers for more hours in a given day than individuals who did not, and the authors noted that time spent accessing the Internet needs to be further assessed and considered as a risk factor for cyberbullying behaviors. Consistent with Mishna et al. (2012), Navarro and Jasinski (2012) found that suitable targets (i.e., the types of activities a teen engaged in while online, such as using social networking sites) significantly increased the potential for cyberbullying activities, while increased guardianship in cyberspace does not necessarily have a significant effect on cyberbullying perpetration. Finally, Ang (2015) found that guardianship factors associated with parental styles (e.g., a lack of knowledge about a child’s online activities) was found to be associated with cyberbullying activities. Ang (2015) provided a list of preventive strategies which may be useful in reducing cyberbullying activities.

While the aforementioned contextual factors are relevant in cross-sectional analyses of cyberbullying activities, further development of the contextual factors underlying cyberbullying perpetration can be realized through longitudinal assessment. Moreover, cyberbullying perpetration can perhaps be better analyzed by estimating how these acts evolve throughout adolescence, and by examining if there are independent categories of cyberbullying perpetrators.

Cyberbullying and Developmental Trajectories

A unique aspect within deviance research involves the notion of onset and desistance of offending behavior. For example, Moffitt (1993) suggested that there were distinct groups of delinquent developmental trajectories which have unique trends that can be observed in longitudinal analyses. Moffitt (1993) identified two distinct classes of offenders: (1) adolescent-limited offenders, or age specific offenders, who start and end deviant/delinquent involvement during adolescence and (2) life-course persistent offenders, also known as repeat offenders, who begin offending at a young age and continue well into adulthood.

When assessing the onset and desistance of behaviors, previous research pertaining to traditional bullying, as well as limited research on cyberbullying, has focused on the developmental trajectories of youth. In regard to traditional bullying behaviors, Pepler et al. (2008) analyzed the developmental trajectories of traditional bullying of youth between the ages 10 and 14 years. The authors accounted for contextual factors associated with a youth (i.e., moral disengagement, physical aggression, and relational aggression), family relations (i.e., parental involvement and conflict), and peer relations (i.e., peers who bully, conflict with peers, and susceptibility to peer pressure) in their analyses (Pepler et al., 2008). Four independent trajectories were determined for traditional bullying methods within the sample assessed, and Pepler et al. (2008) found that roughly 10% of the sample reported consistently high levels of bullying, approximately 13% indicated moderate levels of bullying which desisted to no bullying by the end of high school, roughly 35% of the sample manifested moderate levels of bullying, and approximately 41% of the sample indicated no bullying behaviors. Overall, the authors determined that those with an elevated risk of peer, parental, and individual factors were more likely to bully (Pepler et al., 2008).

Outside of traditional bullying, Modecki et al. (2013) integrated the developmental perspective to analyze how changes in certain risk factors (e.g., self-esteem and depressed mood) between 8th and 10th grade was associated with cyber-aggression in 11th grade. The authors determined that (1) early increases in problem behaviors were associated with greater levels of cyber perpetration and victimization, (2) greater decreases in self-esteem were associated with both cyber acts, and (3) greater levels of depression early in life were linked to cyber perpetration and victimization beyond the increase in depression which occurred between 8th and 11th grades for some youth (Modecki et al., 2013). The authors concluded that there are multiple developmental pathways which can explain both cyber perpetration and victimization, and they called for interventions addressing the emotional well-being and problem behaviors early in adolescence to reduce cyber-aggression within the population (Modecki et al., 2013).

Kim et al. (2017) examined the developmental pathways and motives for cyberbullying while accounting for measures associated with self-control theory in a sample of Korean youth. The authors determined that cyberbullying and traditional bullying behaviors covaried, and that there were similar developmental trajectories between the two groups of individuals (Kim et al., 2017). Additionally, parental involvement significantly affected the opportunity structure for cyberbullying activities, and low self-control was significantly associated with a unique grouping of cyberbullies (Kim et al., 2017). The results determined by Kim et al. (2017) slightly deviate from the findings of Lee and Shin (2017) who conducted a similar study and determined that parental attachment did not have a substantial effect on cyberbullying activities. Nevertheless, both authors contend that more research needs to be directed to this line of inquiry (Kim et al., 2017; Lee & Shin, 2017).

Finally, Pabian and Vandebosch (2016) examined the developmental trajectories of both traditional and cyberbullying perpetration using a four-wave panel study, and the authors conducted a latent class analysis to examine differential bullying trajectories over a 2-year period. Pabian and Vandebosch (2016) determined that there were four distinct groupings within the analysis which included: (1) nonstop traditional bullies, (2) traditional and cyberbullies with decreasing perpetration, (3) traditional and cyberbullies with increasing perpetration, and (4) noninvolved (Pabian & Vandebosch, 2016).

Overall, prior research pertaining to cyberbullying and developmental trajectories is rapidly developing given the social changes in relation to the collective shift toward cyberspace. The studies that have come to fruition over the past few years have demonstrated that there appears to be: (a) distinct subgroups of cyberbullies (Pabian & Vandebosch, 2016); (b) variation within the rate of which cyberbullying increases and/or decreases within distinct subgroups of cyberbullies (Modecki et al., 2013; Pepler et al., 2008); and (c) certain theoretical/contextual factors (i.e., parental attachment, parental involvement, self-esteem, self-control, and peer relations) tend to influence the varying categories associated with cyberbullying developmental trajectories (Kim et al., 2017; Pepler et al., 2008).

Present Study

According to the theoretical model, a key mechanism of self-control theory is that low self-control in an interaction with opportunity predicts anti-social or deviant behavior. In other words, individuals with low self-control are more likely to engage in deviant behavior when opportunity is present. However, Gottfredson and Hirschi (1990) did not explain any intervening mechanisms of opportunity on the relationship between low self-control and anti-social or deviant behavior (see Seipel & Eifler, 2010). In fact, many researchers have indicated that individuals with low self-control are more likely to be engage in deviant behavior when opportunity is present without the explanation of the concept of opportunity. For this reason, this study employs opportunity perspectives for an integration of individual traits (low self-control) and situational factors that create criminal opportunity to explain offending behavior. For instance, individuals with low self-control would be more likely to put themselves at certain situations that create/facilitate criminal opportunity, leading to an increased the likelihood of engaging in offending behavior. Because low self-control that is formed early in life preceded lifestyles, it is assumed to that the effect of low self-control on offending behavior might be fully or partially mediated by certain types of risky lifestyles. For this reason, this study hypothesizes that low self-control would influence cyberbullying perpetration directly and indirectly throughout risky lifestyles.

Regarding the methodological model, multiple pathways need to be accounted for because some children begin deviant/delinquent behavior while young, others begin when they are older, and some groups stop offending during adolescence or young adulthood while others do not (Moffitt, 1993). The group-based method is used in developmental criminology by modeling individual-level variability in developmental trends through a small number of subgroups or classes (Nagin, 1999, 2005) so each group is defined by unique sizes and shapes as well as unique growth curves (similarities in their onset and/or desistance from deviant behavior).

The current study examines whether or not youth involved in cyberbullying perpetration fit into distinct developmental trajectories and assesses the degree to which developmental trajectories increase, decline, and/or remain stable over time. The analyses conducted here further examine whether the trajectories are existent while accounting for indicators of low self-control and risky lifestyles which may create, facilitate, and/or mitigate cyberbullying activities. The data utilized in this study involves a 6-year longitudinal dataset of Korean adolescents, and the analyses integrate latent class growth models to assess the following research questions: (1) are there distinct patterns of cyberbullying developmental trajectories that can be attributed to the existence of unobserved subgroups within the population; (2) is low self-control related to the evolution of cyberbullying developmental trajectories; (3) are there direct effects of low self-control and opportunity factors on cyberbullying developmental trajectories; and (4) is the relationship between low self-control and cyberbullying developmental trajectories mediated by the opportunity structure?

Method

Data and Sample

The data for this study came from the National Youth Policy Institute’s (NYPI) Korean Children and Youth Panel Survey (KCYPS). The sample for the KCYPS was selected using a multi-stage stratified cluster sampling technique to achieve a nationally representative sample of youth while accounting for socioeconomic status, sex, and locational characteristics of the population. A total of 78 middle schools were selected from a list of national schools in 16 administrative districts. The schools were sampled proportionately to size, based on the average number of first-year middle school students per class. Middle school students and their parents were sampled in proportion to the number of students enrolled in selected middle schools in each administrative district. A total of 2,351 adolescents who were 14 years old were selected for the first analytic sample in Time 1 of the analysis (Survey Year [SY] 2010).

Among the sampled subjects, 2,280, 2,259, 2,108, 2,091, 2,058, and 1,881 students participated in the survey between SY2011 and SY2016, respectively. Subjects were 15 years old in the SY2011 assessment, and they were followed until they were 20 years old in SY2016. The current study incorporates six times to establish temporal-order inference; Time 1 for control variables, Time 2 for covariates, and Times 3 to 6 for developmental trajectories of cyberbullying perpetration. However, the last wave was excluded because Time 7 subjects were no longer adolescents.

One of the common methodological issues in longitudinal panel design research is missing data/attrition of subjects over time. Approximately 20% of the follow-up rate was addressed by using the robust maximum likelihood (MLR) estimator in Mplus 7.4 (Muthen & Muthen, 2015). All variables examined in the current analyses are provided in Table A in the Supplemental Appendix.

Measures

Dependent variable

Cyberbullying was assessed from SY2012 (Time 3) to SY2015 (Time 6). Consistent with prior research, this study assesses cyberbullying by including two measures which asks respondents whether they engaged in the following types of offending in the last year: (1) “Have you ever intentionally circulated false information on the Internet message boards about others,” and (2) “Have you ever cursed/insulted other people through chats/message boards” (Kim et al., 2017). Each item was coded 1 if the event occurred at least once, and 0 if there was no indication. The values were summed into ratings between 0 and 2. The variable was an ordinal scale of variety in types of cyberbullying from 0 (no cyberbullying) to 1 (only one type of cyberbullying) to 2 (both types of cyberbullying). Variety scores hold advantages over dichotomies or frequency scores (p = .03 at Time 3; p = .01 at Time 4; p = .00 at Time 5; p = .00 at Time 6 for the first item; p = .18 at Time 3; p = .06 at Time 4; p = .05 at Time 5; p = .03 at Time 6 for the first item) (Bendixen et al., 2003).

Independent variables

Low self-control was assessed in SY2011 (Time 2), using six items that were measured through the presence of impulsivity and temper out of the following six domains of self-control described by Gottfredson and Hirschi (1990): temper, simple tasks, risk seeking, physical activities, self-centeredness, and impulsivity (Arneklev et al., 1998; see Table A in the Supplemental Appendix). Response options to these items were administered on a 5-point Likert scale ranging from 1 = very untrue to 5 = very true. Confirmatory factor analysis was used to create a single construct (α = .81). Higher values on this latent variable reflect less self-control.

Computer use hours is a single-item measure in SY2011 (Time 2), asking respondents the average number of hours of computer usage in a day. Online lifestyles by computer were measured using six items in SY2011 (Time 2) (Navarro & Jasinski, 2012; Ngo & Paternoster, 2011; see Table A in the Supplemental Appendix). Online lifestyles by a smartphone was assessed in SY2011 (Time 2) using eight items asking respondents about online activities using a smartphone (see Table A in the Supplemental Appendix). Response options for all of the aforementioned items were assessed on a 5-point Likert scale, ranging from 1 (never) to 4 (often). As a result of the internal consistency estimate for online lifestyles by a computer (α = .72) and a smartphone (α = .70), confirmatory factor analysis was conducted to create a single variable for each. Higher scores reflect more frequent engagement in online lifestyles on either a computer or smartphone.

Delinquent peer association was at Time 2 using 11 items administered in SY2011 (see Table A in the Supplemental Appendix). Each item was recoded where 1 = yes and 0 = no, creating a summated scale ranging from 0 to 11. Higher values on this variable indicate more frequent delinquent peer associations.

Control variables

Sex (males = 1 and females = 0) and family structure (single parent = 1 and otherwise = 0) were measured dichotomously. Also, two types of parenting style were controlled for at Time 1: parent supervision and parental attachment (see Table A in the Supplemental Appendix). Response options were assessed on a 4-point scale ranging from 1 (very untrue) to 4 (very true). As a result of the estimated internal consistency values (α = .76 for parent supervision and α = .82 for parent attachment), confirmatory factor analysis was conducted to create a single construct for each parenting style. Higher values for both constructs represent greater levels of each parenting style. Further, measures for traditional bullying perpetration and victimization in SY2011 (Time 2) were included as controls (see Table A in the Supplemental Appendix). All items for each variable were coded as 1 = yes and 0 = no. The values were summed ranging from 0 to 4 to be considered a count variable. Higher values on each latent variable reflect more frequent traditional bullying perpetration and victimization.

Analysis

Latent class growth analysis (LCGA) was conducted to identify the heterogeneity in cyberbullying developmental trajectories, grouping cases into a certain number of classes with distinct class-specific trajectories, utilizing Mplus 7.4 (Muthen & Muthen, 2015). LCGA is a group-based semi-parametric approach based on finite growth mixture modeling (GMM) that uses a multinomial logistic regression modeling strategy to identify relatively homogeneous classes of developmental trajectories related to covariates (Muthen & Muthen, 2000).

The data analysis conducted here proceeded in a series of steps. The first step of the first stage estimated both the means and variances of latent growth factors: an intercept, representing an initial level/average starting point of cyberbullying at Time 1, and a slope, representing a rate of change across four time points. The second step was to select a final optimal classification model. Since there is no clear-cut decision rule on how to determine the final optimal class model, there are multiple model fit indices—absolute model fit (e.g., maximum log likelihood [LL] value) and relative model fit (e.g., AIC [the Akaike Information Criterion], BIC [the Bayesian Information Criterion], SABIC [the sample-size adjusted BIC], CAIC [consistent AIC], and AWE [Approximate Weight of Evidence Criterion]). Additionally, for likelihood ratio tests (LRT) statistics, both the Lo-Mendell-Rubin likelihood ratio test (LMR-LRT) and the bootstrapped LRT (BLRT) were used to compare two adjacent/nested class models in assessing whether a k-class model (H₀) demonstrates a better fit compared to a k+1 class model (H₁).

The second stage involved estimating the impact of covariates on 1-latent trajectory class membership. For this stage, a 3-step approach was implemented, providing an improvement to the “classify-then-analyze” procedure in which the samples were first paced into a certain number of classes, and then regressed on covariates (Nylund-Gibson & Masyn, 2016). This approach assumes that class membership is measured without measurement error (Roederet al., 1999). Not considering the error variance in the class enumeration process may result in statistical bias. In other words, simultaneously identifying class membership and estimating individual differences across classes may cause the uncertainty in the rates of class membership when estimating standard errors used in testing for class differences. In this study, the means of each class were fixed using the logit values resulting from the Mplus output of the identification of class membership. Thus, class formation and interpretation was not affected by the subsequent addition of covariates in the model. After the final optimal model was selected, covariates were included in the longitudinal mediation models. This study used the difference in the coefficients approach, depicting both the direct effect of low self-control on cyberbullying and the mediating effect¹ of risky lifestyles on the link between low self-control and cyberbullying. The study tested whether the finding satisfied the fourth assumption, identifying whether the coefficient relating low self-control to delinquency in the full model is smaller than the coefficient in the first model before controlling for the mediators (opportunity variables). The overall results showed that each of the covariates correlated with cyberbullying in the predicted direction (see Table 1).

Table 1.

Correlations Among the Study Variables (N = 2,351).

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
1.	Cyberbullying perpetration T3	1	00.249**	00.219**	00.148**	00.104**	00.129**	00.031	−0.036	00.078**	00.128**	00.033	−0.074**	−0.049*	00.229**	00.043
2.	Cyberbullying perpetration T4		1	00.283**	00.235**	00.105**	00.070**	−0.003	−0.070**	00.015	00.156**	00.058**	−0.043*	−0.015	00.192**	00.036
3.	Cyberbullying perpetration T5			1	00.215**	00.064**	00.035	−0.002	−0.0030	00.041	00.109**	00.058**	−0.032	00.017	00.137**	00.024
4.	Cyberbullying perpetration T6				1	00.012	00.091**	00.023	−0.017	00.032	00.089**	00.062**	−0.022	00.038	00.136**	00.054*
5.	Low self-control T2					1	00.108**	00.164**	00.088**	00.080**	00.128**	00.103**	−0.126**	−0.143**	−0.040	00.013
6.	Computer use hours T2						1	00.175**	00.027	00.042*	00.039	00.045*	−0.160**	−0.075**	00.091**	00.159**
7.	Online lifestyles-computer T2							1	00.331**	00.051*	00.034	00.020	00.008	−0.014	−0.197**	00.023
8.	Online lifestyles-smartphone T2								1	90.037	00.012	−0.014	00.050*	00.103**	−0.188**	−0.040
9.	Delinquent peer association T2									1	0.185**	0.060**	0.008	−0.002	00.082**	00.012
10.	Parent supervision T1										1	00.183**	−0.014	−0.009	0.139**	0.043*
11.	Parent attachment T1											1	−0.033	−0.044*	0.061**	0.019
12.	Traditional bullying perpetration T2												1	00.424**	00.047*	−0.102**
13.	Traditional bullying victimization T2													1	00.041	−0.080**
14.	Sex (Males = 1)														1	00.044*
15.	Single parents															1

Note. ^*p ≤ .05. ^**p ≤ .01. ^***p ≤ .001.

Results

Stage I: Model Selection

Cyberbullying developmental trajectories were assessed from SY2012 (Time 3) to SY2015 (Time 6) using the LGCM. Adolescents averaged 0.137 cyberbullying at Time 3 (age 16), whereas the average increase in cyberbullying over the 4-year period was −0.037 per year (p < .001). Also, the variances of both growth factors were significant (b = 0.040 at p < .01 and b = 0.002 at p < .001), indicating that inter-individual differences in the trajectories vary over time.

The second step of this stage was to determine the best fitting model. Although the 4-class model with the smallest value of AIC, BIC, the sample-size adjusted BIC, CAIC, and AWE provided the best fit to the data, the significant p-values of the LMR-LRT and BLRT provided evidence supporting a 3-class solution.² Entropy³ of the four models ranged from 0.988 to 0.989, representing better classification (see Table 2). Class 1 observed the highest most-likely membership (91.3%). The intercept (the initial level at Time 1) for the class 1 trajectory was lower than average (b = 0.024; p < .001), and the slope (the rate of change across times) was non-significant. That led us to label class 1 as the normative trajectory group. Class 2 was the smallest in terms of most-likely membership (3.2%). The intercept was non-significant, but the class 2 trajectory sharply increased thereafter, as evidenced by its significant positive linear slope (b = 0.310; p < .001) (see Figure 1). This led us to interpret class 2 as the increasing and late-peak group. The last trajectory accounted for over 5.5% of the variation. The intercept was significantly higher than average (b = 1.451; p < .001); however, it dropped steeply thereafter, as evidenced by its significant negative linear slope (b = –0.484; p < .001). We interpreted class 3 as the early onset and decreasing group.

Table 2.

Model Fit Indexes for Latent Class Growth Analysis and Grow Mixture Model (N = 2,351).

Model (K)	LL	N-par	AIC	BIC	SABIC	CAIC	AWE	H0: k classes; H1: k+1 classes			Entropy
Model (K)	LL	N-par	AIC	BIC	SABIC	CAIC	AWE	LRTS	Adj LMR p-value	Bootstrapped p-value	Entropy
LCGA
1-class	–43525.051	7	87064.10	87104.23	87081.99	8266.30	83304.43	2181.01	<.001	<.001	0.988
2-class	–42387.532	10	84795.06	84852.38	84820.61	84862.38	84959.70	1238.05	<.001	<.001	0.988
3-class	–41741.821	13	83509.64	83584.16	83542.85	83597.16	83723.67	964.28	.244	.049	0.989
4-class	–41238.897	16	82509.80	82601.50	82550.67	82617.50	82773.22	11.74	.6955	.6667	0.988
GMM
1-class	–42876.795	10	85773.59	85830.91	85799.14	85840.91	85938.23	na	na	na	na

Note. LL = Model maximum log likelihood value (the value shown in bold indicates a model with the smallest LL value that perfectly fits the data); npar = number of free parameters estimated in the model; AIC = the Akaike Information Criterion (the value shown in bold indicates the model with the smallest value); BIC = the Bayesian Information Criterion (the value shown in bold indicates the model with the smallest value); SABIC = the sample-size adjusted BIC (the value shown in bold indicates the model with the smallest value); CAIC = consistent AIC (the value shown in bold indicates the model with the smallest value); AWE = Approximate Weight of Evidence Criterion (the value shown in bold indicates the model with the smallest value); LRTS = likelihood ratio test statistics comparing a current model (k class) to a model with one more latent class (k+1 class); Adj LMR p-value = the adjusted Lo–Mendell–Rubin likelihood ratio test p-value (the value shown in bold represents the non-significant p-value, indicating the current model with the smaller number of classes is not rejected); Bootstrapped p-value = parametric bootstrapped p-value for the LRTS; BF = Bayes factor comparing the current model (k class) to a model with one more latent class (k+1 class) (the value shown in bold indicates a model with the smallest number of classes that is favored over a model with one more latent class); cmP(K) = the approximate correct model probability compared to all models (the value shown in bold indicates any models with cmP(K) > 0.10, showing strong evidence for the correct model. LCGA = Latent Class Growth Analysis; GMM = Growth Mixture Model, where variances and covariances are freed to be estimated for all classes (class-varying variances and covariances).

Figure 1.

Sample and estimated means of cyberbullying developmental trajectories from latent class growth modeling.

Stage II: Adding Between-Class Effects of Covariates to a LCGA

Class-specific impacts of low self-control and opportunity structure were estimated by using multinomial logistic regression (see Table 3). Low self-control at Time 2 was significantly associated with the odds of membership in the early onset/decreasing group in comparison to the normative trajectory group as a chosen reference group (b = 1.036, p < .001) as well as the increasing/late peak group (b = 0.778, p < .05) (see Model 1 in Table 3). The probabilities of membership in the increasing/late peak group were associated with parent supervision and parent attachment at Time 1 as well as traditional bullying perpetration at Time 2 (b = 0.527, p < .05; b = 0.096, p < .05; b = 0.527, p < .01), compared to the normative trajectory group. Parent attachment at Time 1 was significantly related to the early onset/decreasing group in comparison with the increasing/late peak group (b = −0.786, p < .05). Additionally, males were more strongly associated with both the early onset/decreasing (b = 2.023, p < .001) and increasing/late peak groups (b = 1.96, p < .001).

Table 3.

Multinomial Regression Model Estimating Direct and Mediated Effects on Class Membership in Cyberbullying.

	Model 1						Model 2
	Estimation of direct effects of low self-control						Estimation of intervening effects of opportunity factors
	Model A		Model B		Model C		Model A		Model B		Model C
	Early onset/decreasing (5.5%) versus Normative trajectory (91.3%)		Increasing/late peak (3.2%) versus Normative trajectory (91.3%)		Early onset/decreasing (5.5%) versus Increasing/late peak (3.2%)		Early onset/decreasing (5.5%) versus Normative trajectory (91.3%)		Increasing/late peak (3.2%) versus Normative trajectory (91.3%)		Early onset/decreasing (5.5%) versus Increasing/late peak (3.2%)
	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.
Control variables
Sex (Males = 1)	2.023***	0.276	1.96***	0.378	0.063	0.465	1.88***	0.282	1.866***	0.398	0.014	0.484
Single parents	0.183	0.297	0.564	0.316	−0.38	0.411	−0.015	0.302	0.386	0.327	−0.402	0.426
Parent supervision T1	−0.211	0.23	0.527*	0.17	0.237	0.317	−0.134	0.238	−0.459	0.257	0.325	0.34
Parent attachment T1	−0.079	0.24	0.096*	0.193	−0.786*	0.404	−0.057	0.253	0.845*	0.382	−0.902*	0.45
Bullying perpetration T2	0.22	0.143	0.527**	0.17	−0.308	0.195	0.313*	0.158	0.558**	0.191	−0.245	0.219
Bullying victimization T2	−0.02	0.184	0.096	0.193	−0.115	0.235	−0.01	0.187	0.188	0.189	−0.198	0.238
Self-control variable
Low self-control T2	1.036***	0.225	0.258	0.281	0.778*	0.344	1.141***	0.244	0.094	0.307	1.046**	0.376
Opportunity variables
Computer use hours T2							0.057	0.043	0.147**	0.052	−0.09	0.065
Online lifestyles-computer T2							0.274	1.385	3.405*	1.714	−3.13	2.128
Online lifestyles-smartphone T2							−1.455	1.032	−2.034	1.229	0.579	1.526
Delinquent peer association T2							−0.005	0.006	−0.004	0.006	−0.002	0.007
Model Fit Information
Akaike (AIC)	−5150.693						−5016.906
Bayesian (BIC)	−5003.165						−4825.110
Sample-Size Adjusted BIC	−5085.770						−4933.131

Note. In Models A and B, Reference group = the normative trajectory group. In Model C, Reference group = the increasing/late peak group.

p < .05; **p < .01; ***p < .001.

Opportunity factors were added to Model 2. Computer use hours and online lifestyles by a computer at Time 2 were significantly related to membership in the increasing/late peak group compared to the normative trajectory group. Specifically, the odds of being a member of the increasing/late peak group increased by 0.147 for every unit increase of computer use hours at Time 2 (p < .01), while those increased by 3.405 for every unit of online lifestyles by computer (p < .05). For the early onset/decreasing group, none of opportunity factors significantly predicted membership in comparison with the normative trajectory group. However, low self-control at Time 2 remained significant after controlling for opportunity factors compared to both the normative trajectory group (b = 1.141, p < .001) and the increasing/late peak group (b = 1.046, p < .01). Traditional bullying perpetration at Time 2 was significantly associated with both the early onset/decreasing and increasing/late peak groups in comparison to the normative trajectory group (b = 0.313, p < .05; b = 0.558, p < .01, respectively). Importantly, traditional bullying perpetration at Time 2 became significant for the early onset/decreasing group membership after controlling for opportunity factors in the full model. Parental attachment at Time 2 was significant for the early onset/decreasing group membership in comparison to the increasing/late peak group (b = -0.902, p < .05). Males demonstrated a significantly increased odds of membership in both the early onset/decreasing and increasing/late peak groups.

Discussion

Few longitudinal studies exist which identify the heterogeneity of the samples within an unique Korean context. The current study examined cyberbullying developmental trajectories to determine unique patterns of cyberbullying in a heterogeneous sample of Korean adolescents between the ages of 14 and 19. Additionally, this study examined the direct effect of low self-control and opportunity factors on cyberbullying trajectory membership as well as the mediating effect of opportunity factors on the relationship between low self-control and trajectory membership. The analyses conducted here occurred in two specific stages, and a discussion pertaining to the results identified here is provided below.

Phase I Discussion

The first phase of the current study involved assessing the patterns of cyberbullying developmental trajectories for youth between the ages 14 and 18. This study found much heterogeneity in the patterns, identifying three groups: (1) a normative trajectory group (accounting for 91.3% of the sample) which demonstrated the intercept for the class 1 trajectory was lower than the average, and the slope was not significant; (2) a increasing and late-peak group (accounting for 3.2% of the sample) which observed that the intercept was non-significant, but the trajectory sharply increased thereafter; and (3) an early onset and decreasing group (accounting for 5.5% of the sample) which demonstrated that the intercept was significantly greater than the average, but dropped steeply thereafter. Thus, the results from this portion of the analyses suggest that there are three distinct classes of youth involved in cyberbullying perpetration behaviors.

Prior research, however, has found between two and five offending trajectories. The higher number of subgroups classified is more in line with studies that use high-risk offenders (Sampson & Laub, 2003) rather than population-based samples (Barker et al., 2007). This study used low-risk adolescents with a population-based sample, and serious offending remained relatively underexplored. For this reason, future research would replicate serious offenders and their behaviors during a longer developmental period to determine generalizability across populations.

The majority of youth in this study fall into the category of normative trajectory, suggesting that the majority of adolescents in the sample have little or no experiences with cyberbullying, or consistent levels of cyberbullying over time. A small portion of youth fall into the early onset and decreasing group, indicating that trajectories peaked in early adolescence followed by a decrease thereafter in adolescence. This group observed a similar pattern to the age-crime curve (Barker et al., 2007; Sampson & Laub, 2003) and the adolescent-limited group of Moffitt’s (1993) dual pathway developmental theory. It means that adolescent-limited (age specific) offenders started and ended deviant/delinquent involvement during adolescence and reflect their temporary involvement in antisocial behavior due to the “maturity gap.” This group recognized the potential consequences of future criminal behavior and discontinued their criminal behavior. Last, the increasing and late-peak group was characterized by very low levels of cyberbullying in early adolescence followed by a substantial increase in later adolescence. This group is similar to the life-course persistent (repeat) offenders who begin offending at a young age and continue well into adulthood, characterized by cross situational behavioral consistency (Moffitt, 1993). This group exhibits “contemporary continuity,” meaning that repeat offenders bring the same criminal propensities that they exhibited as an adolescent to adulthood and act.

The information retained from the analyses conducted in this phase is useful for the construction and/or directing of efforts to mitigate cyberbullying within the population. Adequate efforts should be devised to specifically target the normative trajectory group because addressing this group of youth could substantially decrease cyberbullying in the aggregate considering the fact that so many youths fit into this category. By fully analyzing the risk associated with the normative trajectory group going forward, a needs assessment at the aggregate level would prove beneficial for social organizations that have the capacity to respond in a meaningful manner (Andrews et al., 1990).

Phase II Discussion

The second portion of the analysis conducted here examined whether adolescents’ distinct trajectories differed based on theoretically relevant covariates (i.e., low self-control and opportunity structure) and was assessed when youth in the sample were 14 years old. Findings indicate that low self-control is associated with an increased odds of membership in the early onset/decreasing group in comparison to both the normative trajectory group and the increasing/late peak group after controlling for opportunity factors. Adolescents with less self-control subsequently demonstrated a decreased rate of cyberbullying which seemed to peak at age 14 when compared to the other two groups. Consequently, low self-control has a significant effect on cyberbully perpetration, and this factor seems to have a substantial effect on the early onset/decreasing group of youth. Low self-control has previously been linked to increased levels of cyberbullying perpetration (Li et al., 2016; Vazsonyi et al., 2012); nevertheless, this finding is unique because the longitudinal nature of the analysis conducted here observed a significant decline away from cyberbullying perpetration for youth transitioning into adulthood. It may be the case that youth with greater levels of low self-control stop cyberbullying behaviors and/or socially undesirable behaviors later in the teenage years, or it could be a function of these youth moving on to other forms of deviant behaviors. Future research should investigate this class of individuals further to estimate differential outcome measures to determine if that is the case.

Computer use hours and online lifestyles by a computer of adolescents at the age of 14 was found to be significantly related to membership in the increasing/late peak group when compared to the normative trajectory group. Specifically, adolescents who used computers longer and engaged in online lifestyles were more likely to demonstrate an increased rate of committing cyberbullying and these activities peaked at age 18. This finding is consistent with prior research indicating that more time spent using a computer each day increases cyberbullying perpetration (Ang, 2015; Mishna et al., 2012; Navarro & Jasinski, 2012, 2013). Based on the results determined here, measures targeting the online opportunity structure can be manipulated to reduce cyberbullying perpetration, to some degree. Thus, it is ideal for caregivers of youth (i.e., parents, grandparents, siblings, etc.) to take note of not only what a youth is doing while in cyberspace, but also note how much time a youth is spending in cyberspace in order to reduce the opportunity for social harm. Limiting the amount of time and/or accessibility to cyberspace by caregivers can reduce the opportunity structure for youth to cyberbully in that domain.

Finally, low self-control at Time 2 remained significant after controlling for opportunity factors compared to both the normative trajectory or increasing/late peak groups. This indicated low self-control in and of itself appears to supersede accessing and time spent using the internet for some youth. According to Gottfredson and Hirschi (1990), self-control develops early in life and it is largely a function of parenting method/style. Establishing desirable levels of self-control is theorized to have a sustained impact throughout the life-course (Gottfredson & Hirschi, 1990). Thus, individual parenting methods could be responsible for the undesirable behaviors of adolescents and young adults, as demonstrated within the analyses conducted here, and it appears to be the case that opportunity structure does not significantly condition cyberbullying behaviors overtime when low self-control is present. Consequently, parenting styles are likely to have a greater impact on the cyberbullying activities of youth, and social mechanisms which help foster idealized parenting styles are subsequently likely to reduce the total amount of adolescent cyberbullying behaviors within society. Overall, the results here suggest that addressing undesirable parenting styles is perhaps a more desirable avenue to pursue to reduce cyberbullying activities than simply limiting the opportunity for you to engage in online activities.

Limitations and Recommendation for Future Research

Despite the strengths of this study, there are a number of limitations. First, the current analysis did not explain the aggregate-level effects of opportunities for cyberbullying. At the individual level, exposure to risk influences opportunities in a similar manner as opportunity perspectives suggest. At the aggregate level (e.g., within schools), this concept alters or moderates the individual-level effects on deviant opportunities because there is likely variation within nested groupings conducive to school climates. Spatial characteristics should therefore be integrated in future analyses to determine if there are nested effects associated with the opportunity structure which influence micro-level cyberbullying activities.

Additionally, some survey items necessary for a thorough assessment of the indicators utilized in this study were not available in the dataset analyzed. For example, the dataset used in the analyses did not include comprehensive items for cyberbullying activities. Cyberbullying perpetration is a broad construct and can involve harassing, insulting, threatening, socially excluding, and/or humiliating others on the Internet or by using mobile phones (Chisholm, 2014). The current study, however, included only two items to examine cyberbullying behaviors: (1) intentionally circulating false information and (2) insulting others on the Internet. Future research should attempt to access a more comprehensive list of items to measure cyberbullying activities as well as the other domains assessed in this study. Also, according to Olweus’s (1978) definition, bullying incorporates the following three characteristics: (1) repeated incident amongst the same perpetrators and victims over time; (2) perpetrators’ intention to harm victims; and (3) an imbalance power between perpetrators and victims. Both items of cyberbullying were not able to reflect the repeated nature of bullying.

Further, the analyses conducted here rely on a sample of youth from South Korea and the results may be limited to regionally specific contextual factors. With the idea of diversity in mind, future analyses need to be more inclusive when assessing the underlying causes of cyberbullying perpetration by integrating samples that are more diverse based on factors such as race, ethnicity, nationality, culture, and geographic location. Additionally, it would be productive for future studies to thoroughly isolate the causal process underlying bullying perpetration by accounting for gender-specific factors because doing so may highlight unique results that were not observed in the analyses conducted here.

Finally, self-control theory highlights that self-control is established in early childhood (by the age of 8) and remains stable throughout the life course. Thus, the theory emphasizes parental attachment and monitoring in order to establish direct control and subsequently self-control through proper parenting practices. Further, self-control theory states that attachment is no longer important after self-control has been established early in life (around age 8). For this reason, the analyses included parenting factors (parent supervision and parental attachment) that have been shown to influence self-control. Therefore, consistent with other studies (Botchkovar et al., 2015; Cullen et al., 2008; Jackson & Vaughn, 2018), these items were controlled in the analyses. However, these items were assessed at Time 1 when subjects were at the age of 14. Future research would assess parenting practices in early childhood (around age 8).

Conclusion

The current study examined indicators from self-control and opportunities theories to determine if there were distinct trajectories of cyberbullying perpetration within a sample of Korean youth. Results from the analyses identified heterogeneity within cyberbullying developmental trajectories over a 4-year period through adolescence. The analyses yielded an unobserved 3-class subgroup, each having a unique pattern of cyberbullying perpetration trajectories. Low self-control was a significant factor among members in the early onset/decreasing group, while opportunity factors were significant amongst members in the increasing/late peak group, compared to the normative trajectory group of cyberbullies. Consequently, there appears to be distinct classes of cyberbullies within the population and future research should further investigate classifications of cyberbullying subgroups to better inform social intervention methods. Additionally, it is important to continue this line of research while integrating theoretically relevant variables so practitioners will be better equipped to integrate evidence-based practices when addressing the issue of cyberbullying perpetration within the population.

Supplemental Material

Appendix – Supplemental material for Impacts of Low Self-control and Opportunity Structure on Cyberbullying Developmental Trajectories: Using a Latent Class Growth Analysis

Supplemental material, Appendix for Impacts of Low Self-control and Opportunity Structure on Cyberbullying Developmental Trajectories: Using a Latent Class Growth Analysis by Sujung Cho and Steven Glassner in Crime & Delinquency

Footnotes

Authors’ Note

“Informed consent was not needed because this study used the National Youth Policy Institute’s (NYPI) Korean Children and Youth Panel Survey (KCYPS) data set”.

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.

ORCID iDs

Sujung Cho

Steven Glassner

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Sujung Cho is an assistant professor in the Department of Criminology and Criminal Justice at Southern Illinois University Carbondale. Her primary research interest includes developmental criminology and victimology, juvenile delinquency, and quantitative research in cross cultural settings. She is currently working on developmental research, which presupposes developmental determinism where childhood delinquent behavior leads to patterns of peer delinquency trajectories or, alternatively, patterns of peer delinquency trajectories influence late adolescent delinquent behavior.

Steven Glassner is an assistant professor at Tarleton State University. His research focuses on early victimization and the negative outcomes associated with victimized youth. Dr. Glassner is currently working on analyzing the various health consequences associated with bullying victimization to see how these factors influence delinquency/criminality over the life course.

References

Akers

R. L.

Jensen

G. F.

(2008). The empirical status of social learning theory of crime and deviance: The past, present, and future. In Cullen

F. T.

Wright

J. P.

Blevins

K. R.

(Eds.), Taking stock: The status of criminological theory (pp. 37–76). Transaction Publishers.

Andrews

D. A.

Zinger

Hoge

R. D.

Bonta

Gendreau

Cullen

F. T.

(1990). Does correctional treatment work? A clinically relevant and psychologically informed meta-analysis. Criminology, 28(3), 369–404.

Ang

R. P.

(2015). Adolescent cyberbullying: A review of characteristics, prevention and intervention strategies. Aggression and Violent Behavior, 25(A), 35–42.

Arneklev

B. J.

Cochran

J. K.

Gainey

R. R.

(1998). Testing Gottfredson and Hirschi’s “low self-control” stability hypothesis: An exploratory study. American Journal of Criminal Justice, 23(1), 107–127.

Arntfield

(2015). Toward a cybervictimology: Cyberbullying, routine activities theory, and the anti-sociality of social media. Canadian Journal of Communication, 40, 371–388.

Barker

E. D.

Séguin

J. R.

White

H. R.

Bates

M. E.

Lacourse

Carbonneau

Tremblay

R. E.

(2007). Developmental trajectories of male physical violence and theft: relations to neurocognitive performance. Archives of General Psychiatry, 64(5), 592–599.

Baron

R. M.

Kenny

D. A.

(1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of personality and social psychology, 51(6), 1173–1182.

Bendixen

Endresen

I. M.

Olweus

(2003). Variety and frequency scales of antisocial involvement: Which one is better? Legal and Criminological Psychology, 8(2), 135–150.

Botchkovar

Marshall

I. H.

Rocque

Posick

(2015). The importance of parenting in the development of self-control in boys and girls: Results from a multinational study of youth. Journal of Criminal Justice, 43(2), 133–141.

10.

Chisholm

J. F.

(2014). Review of the status of cyberbullying and cyberbullying prevention. Journal of Information Systems Education, 25(1), 77–87.

11.

Cho

(2018). The impact of low self-control and delinquent peer associations on bullying perpetration and victimization among South Korean adolescents: Time-concurrent, time-lagged, and latent growth curve modeling. Journal of School Violence, 17(4), 500–520.

12.

Cho

Lee

J. R.

(2018). Impacts of low self-control and delinquent peer associations on bullying growth trajectories among Korean youth: a latent growth mixture modeling approach. Journal of Interpersonal Violence, 1–31. https://doi.org/10.1177/0886260518786495

13.

Cho

Glassner

Lee

J. M.

(2019). Impact of low self-control, parental involvement, and peer relationships on changes of bullying perpetration over time: A latent growth curve model of a sample of South Korean adolescents. Children and Youth Services Review. https://linkinghub.elsevier.com/retrieve/pii/S0190740919303433

14.

Cohen

L. E.

Felson

(1979). Social Change and crime rate trends: A routine activities approach. American Sociological Review, 44(4), 588–608.

15.

Chui

W. H.

Chan

H. C. O.

(2013). Association between self-control and school bullying behaviors among Macanese adolescents. Child Abuse and Neglect, 37(4), 237–242.

16.

Chui

W. H.

Chan

H. C. O.

(2015). Self-control, school bullying perpetration, and victimization among Macanese adolescents. Journal of Child and Family Studies, 24(6), 1751–1761.

17.

Cullen

F. T.

Unnever

J. D.

Wright

J. P.

Beaver

K. M.

(2008). Parenting and self-control. In Goode

(Ed.), Out of control: Assessing the general theory of crime (pp. 61–74). Stanford University Press.

18.

Fletcher

Fitzgerald-Yau

Jones

Allen

Viner

R. M.

Bonell

(2014). Brief report: Cyberbullying perpetration and its associations with socio-demographics, aggressive behavior at school, and mental health outcomes. Journal of Adolescence, 37(8), 1393–1398.

19.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford University Press.

20.

Guo

(2016). A meta-analysis of the predictors of cyberbullying perpetration and victimization. Psychology in the Schools, 53(4), 432–453.

21.

Hagan

(2012). Essentials of research methods in criminal justice and criminology (3rd ed.). Allyn and Bacon.

22.

Hindelang

Gottfredson

Garofalo

(1978). Victims of personal crime: An empirical foundation for a theory of personal victimization. Ballinger.

23.

Hinduja

Patchin

J. W.

(2010). Bullying, cyberbullying, and suicide. Archives of Suicide Research, 14(3), 206–221.

24.

Holt

T. J.

Bossler

A. M.

May

D. C.

(2012). Low self-control deviant peer associations and juvenile cyberdeviance. American Journal of Criminal Justice, 37(3), 378–395.

25.

Jackson

D. B.

Vaughn

M. G.

(2018). A multi-informant study of the role of household disorder in low self-control. Criminal Justice and Behavior, 45(12), 1977–1996.

26.

Kim

Song

Jennings

W. G.

(2017). A distinct form of deviance or a variation of bullying? Examining the developmental pathways and motives of cyberbullying compared with traditional bullying in South Korea. Crime and Delinquency, 63(12), 1600–1625.

27.

Kokkinos

C. M.

Antoniadou

Asdre

Voulgaridou

(2014). Parenting and internet behavior predictors of cyber-bullying and cyber-victimization among preadolescents. Deviant Behavior, 37(4), 439–455.

28.

Kowalski

R. M.

Giumetti

G. W.

Schroeder

A. N.

Lattanner

M. R.

(2014). Bullying in the digital age: A critical review and meta-analysis of cyberbullying research among youth. Psychological Bulletin, 140(4), 1073–1137.

29.

Lee

Shin

(2017). Prevalence of cyberbullying and predictors of cyberbullying perpetration among Korean adolescents. Computers in Human Behavior, 68(C), 352–358.

30.

C. K. W.

Holt

T. J.

Bossler

A. M.

May

D. C.

(2016). Examining the mediating effects of social learning on the low self-control-cyberbullying relationship in a youth sample. Deviant Behavior, 37(2), 126–138.

31.

Miethe

T. D.

Meier

R. F.

(1990). Opportunity, choice, and criminal victimization: A test of a theoretical model. Journal of Research in Crime and Delinquency, 27(3), 243–266.

32.

Mishna

Cook

Gadalla

Daciuk

Solomon

(2010). Cyber bullying behaviors among middle and high school students. American Journal of Orthopsychiatry, 80(3), 362–374.

33.

Mishna

Khoury-Kassabri

Gadalla

Daciuk

(2012). Risk factors for involvement in cyber bullying: Victims, bullies and bully-victims. Children and Youth Services Review, 34(1), 63–70.

34.

Modecki

K. L.

Barber

B. L.

Vernon

(2013). Mapping developmental precursors of cyber-aggression: Trajectories of risk predict perpetration and victimization. Journal of Youth and Adolescence, 42(5), 651–661.

35.

Moffitt

T. E.

(1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100(4), 674–701.

36.

Moon

Alarid

L. F.

(2015). School bullying, low self-control, and opportunity. Journal of Interpersonal Violence, 30(5), 839–856.

37.

Moon

Hwang

H. W.

McCluskey

J. D.

(2011). Causes of school bullying: Empirical test of a general theory of crime, differential association, and general strain theory. Crime and Delinquency, 57(6), 849–877.

38.

Moon

Morash

McCluskey

J. D.

(2012). General strain theory and school bullying: An empirical test in South Korea. Crime and Delinquency, 58(6), 827–855.

39.

Muthén

L. K.

(2000). Integrating person-centered and variable-centered analyses: Growth mixture modeling with latent trajectory classes. Alcoholism: Clinical and experimental research, 24(6), 882–891.

40.

Muthen

(2015). Mplus user’s guide. 6th edn. Los Angeles, CA: Muthen & Muthen.

41.

Nagin

D. S.

(1999). Analyzing developmental trajectories: a semiparametric, group-based approach. Psychological Methods, 4(2), 139–157.

42.

Nagin

D. S.

(2005). Group-based modeling of development. Harvard University Press.

43.

Navarro

J. N.

Jasinski

J. L.

(2012). Going cyber: Using routine activities theory to predict cyberbullying experiences. Sociological Spectrum, 32(1), 81–94.

44.

Navarro

J. N.

Jasinski

J. L.

(2013). Why girls? Using routine activities theory to predict cyberbullying experiences between girls and boys. Women & Criminal Justice, 23(4), 286–303.

45.

Ngo

F. T.

Paternoster

(2011). Cybercrime Victimization: An examination of Individual and Situational level factors. International Journal of Cyber Criminology, 5(1), 773–793.

46.

Nylund-Gibson

Masyn

K. E.

(2016). Covariates and mixture modeling: Results of a simulation study exploring the impact of misspecified effects on class enumeration. Structural Equation Modeling: A Multidisciplinary Journal, 23(6), 782–797.

47.

Olweus

(1978). Aggression in the schools: Bullies and whipping boys. Hemisphere.

48.

Pabian

Vandebosch

(2016). Developmental trajectories of (cyber) bullying perpetration and social intelligence during early adolescence. Journal of Early Adolescence, 36(2), 145–170.

49.

Patchin

J. W.

Hinduja

(2010). Cyberbullying and self-esteem. Journal of School Health, 80(12), 614–621.

50.

Pepler

Jiang

Craig

Connolly

(2008). Developmental trajectories of bullying and associated factors. Child Development, 79(2), 325–338.

51.

Pratt

T. C.

Cullen

F. T.

(2000). The empirical status of Gottfredson and Hirschi’s general theory of crime: A meta-analysis. Criminology, 38(3), 931–964.

52.

Rice

Petering

Rhoades

Winetrobe

Goldbach

Plant

Montoya

Kordic

(2015). Cyberbullying perpetration and victimization among middle-school students. American Journal of Public Health, 105(3), 66–72.

53.

Roeder

Lynch

K. G.

Nagin

D. S.

(1999). Modeling uncertainty in latent class membership: A case study in criminology. Journal of the American Statistical Association, 94(447), 766–776.

54.

Sampson

R. J.

Laub

J. H.

(2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41(3), 555–592.

55.

Seipel

Eifler

(2010). Opportunities, rational choice, and self-control: On the interaction of person and situation in a general theory of crime. Crime & Delinquency, 56(2), 167–197.

56.

Slonje

Smith

P. K.

Frisen

(2012). Processes of cyberbullying, and feelings of remorse by bullies: A pilot study. European Journal of Developmental Psychology, 9(2), 244–259.

57.

Tittle

C. R.

Ward

D. A.

Grasmick

H. G.

(2003). Gender, age and crime deviance: A challenge to self-control theory. Journal of Research in Crime and Delinquency, 40(4), 426–453.

58.

Unnever

J. D.

Cornell

D. G.

(2003). Bullying, self-control, and ADHD. Journal of Interpersonal Violence, 18(2), 129–147.

59.

U.S. Department of Education (2011, November). Student victimization in U.S. schools: Results from the 2009 crime supplement to the National Crime Victimization Survey. Retrieved August 5, 2020. https://nces.ed.gov/pubs2012/2012314.pdf

60.

U.S. Department of Education (2018). Student victimization in U.S. schools: Results from the 2015 school crime supplement to the National Crime Victimization Survey. https://nces.ed.gov/pubs2018/2018106.pdf

61.

Vazsonyi

A. T.

Machackova

Sevcikova

Smahel

Cerna

(2012). Cyberbullying in context: Direct and indirect effects by low self-control across 25 European countries. European Journal of Developmental Psychology, 9(2), 210–227.

62.

Vermunt

J. K.

Magidson

(2003). Latent class models for classification. Computational Statistics & Data Analysis, 41(3–4), 531–537.

63.

Zych

Baldry

A. C.

Farrington

D. P.

Llorent

V. J.

(2019). Are children involved in cyberbullying low on empathy? A systematic review and meta-analysis of research on empathy versus different cyberbullying roles. Aggression and Violent Behavior, 45, 83–97. https://doi.org/10.1016/j.avb.2018.03.004

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.16 MB