The Timing Effect of Bullying in Childhood and Adolescence on Developmental Trajectories of Externalizing Behaviors

Abstract

This longitudinal study investigated the timing effect of bullying on developmental trajectories of externalizing behaviors from middle childhood to adolescence. We focused on the relation of (a) only an early experience of bullying (i.e., desisters) to subsequent externalizing behaviors in adolescence and (b) only a late experience of bullying (i.e., late-onsetters) to the concurrent externalizing behaviors in adolescence. Their trajectories of externalizing behaviors were compared with the persisters and to the non-experience group. Individual growth curve modeling was conducted using 440 child–mother dyads from the Springfield Child Development Project, a community-representative, longitudinal study over a 6-year period that included four time interviews. We modeled the changes in child aggression and delinquency from 7 to 19 years of age as a function of bully status group. Results indicated that the levels of aggression and delinquency for the desisters decreased over time (with the cessation of bullying in adolescence) and were significantly lower than those of the persisters and similar to those of the non-involved group at the end of the trajectory (cessation effect). For the late-onsetters, the level of delinquency increased over time (with the onset of bullying behaviors in adolescence) and were significantly higher than those of the non-involved group and similar to those of the persisters at the end of the trajectory (onset effect). The aggression for the late-onsetters, however, did not support the onset effect. This study implies that we need to pay more attention to intervening for late-onset.

Keywords

bullying developmental trajectories externalizing behaviors adolescence childhood

Not only does bullying harm victims but can also have a negative effect on children and adolescents who themselves engage in bullying. When a child engages in chronic bullying behaviors, the long-term consequences for that child are likely to include continued externalizing behavior problems and maladjusted relationships (Bender & Lösel, 2011; Broidy et al., 2003; Gibb, Horwood, & Fergusson, 2011; Pepler, Jiang, Craig, & Connolly, 2008, Ryan & Smith, 2009; Ttofi, Farrington, & Lösel, 2012). What are the antecedent factors of bullying that might lead to problems following bullying experiences? Two important antecedent factors may be the timing of the onset and the duration of the bullying in which a child has engaged. For example, engaging in bullying behaviors in primary school seems to provide an increased risk for engaging in bullying in secondary school (Schäfer, Korn, Brodbeck, Wolke, & Schulz, 2005). However, not all adolescents who bully begin bullying behaviors in childhood, and not all children who bully continue to do so in adolescence (Sourander, Helstelä, Helenius, & Piha, 2000).

Some studies have confirmed different trajectories of bullying behaviors (Kim, Boyce, Koh, & Leventhal, 2009; Pepler et al., 2008) or aggression (Broidy et al., 2003; Martino, Ellickson, Klein, McCaffrey, & Edelen, 2008; Underwood, Beron, & Rosen, 2009) based on the timing of the behaviors (e.g., the trajectory of desisters—that is, early starters who desist later, of late-onsetters, and of persisters—that is, early starters who continue bullying). Regarding the trajectories of bullying behaviors, researchers have identified four bullying status groups and investigated potential predictors of which group a child will fall in (e.g., gender, physical characteristics, socio-economic status [SES], parents’ education, and family environment; in Korean students, Kim et al., 2009) and risk factors from child, peer, and family domains (Pepler et al., 2008). For example, to investigate the possible predictors that influence diverse developmental pathways in bullying behaviors, Pepler and her colleagues (2008) identified the type of bullying trajectories from ages 10 to 17 years (i.e., the persistent group with high levels of bullying, the group that desisted by the end of high school, the group that slightly increased over time, and the non-experience group) and the risk factors related to the different bullying trajectories. Transition probabilities linking the early risk classes to the trajectories of bullying over time showed that the majority of the no-risk group fell into the low-bullying trajectory group and the majority of the high-risk group fell into the high or moderately increasing bullying trajectory group. However, we still do not know about the specific timing effects of bullying in childhood and adolescence, and bullies’ subsequent and concurrent maladjustments in adolescence; there is a lack of studies that investigate the timing of bullying behaviors during specific periods and its impact on externalizing trajectories from childhood to adolescence (e.g., Ttofi et al., 2012). To our knowledge, no research investigated bullying experiences during specific time periods of childhood or adolescence as a predictor of the developmental trajectories of externalizing behaviors. Thus, the main objective of this study was to examine the developmental trajectories of externalizing behaviors (i.e., aggression and delinquency) and, in particular, whether the bullying status from childhood to adolescence was related to the trajectories.

By using individual growth modeling, we could interpret the rate of change from childhood to adolescence and the end point of externalizing trajectories in adolescence who desisted in bullying from middle childhood to adolescence or persisted in bullying or first started bullying in adolescence or showed no bullying in childhood or adolescence. Specifically, we wanted to determine whether a child who had engaged in bullying behaviors at an early age but desisted later would show a carry-over effect of the previous bullying experiences on externalizing behaviors or a cessation effect. Also, we questioned whether a child who has engaged in bullying behaviors only at a later age would show a corresponding onset effect of bullying on externalizing behaviors or a resilience effect (because that child had not been involved in bullying behaviors during middle childhood).

Not surprisingly, a child who started bullying others earlier in life and did so over a long period of time might be expected to develop a more aggression and maladaptive relationship patterns than a child who started bullying later or engaged in bullying for only a short time (e.g., continuity model by Caspi, Elder, & Bem, 1987; Wicks-Nelson & Israel, 2006). Stouthamer-Loeber, Loeber, Wei, Farrington, and Wikström (2002) suggested that the reason for the association between the early onset of offending and more persistent, violent, and serious delinquency in the older sample (age between 13 and 19 years) as compared with the younger sample (age between 7 and 12) may be that the impact of risk effects for the older group accumulated over a longer period. Supporting the cumulative effect, one study by Huesmann, Dubow, and Boxer (2009) found that adolescence-limited and childhood-limited aggressiveness did not significantly differ from a life-course-persistent low aggressiveness (i.e., low aggressiveness group) on negative outcomes at age 48. This finding offers hope regarding transient aggressiveness in childhood and adolescence—hope that the harmful impact of the preceding experience of aggression in childhood and adolescence could be reduced in adulthood. However, it is not known whether a transient groups’ recovery might also be the case for childhood-limited bullies and their outcomes in adolescence. Moreover, many studies focused on antisocial and aggressive behaviors that are persistent over the life span and those limited to adolescence (e.g., Huesmann et al., 2009; Moffitt & Caspi, 2001; Patterson, DeBaryshe, & Ramsey, 1989; Underwood et al., 2009). Thus, the current study was meant to fill the gap in this research area.

Definitions

Researchers have defined bullying as a specific form of aggression in which the aggressor acts from a position of power, intimidation, and dominance and intentionally and repeatedly aggresses toward others (e.g., Griffin & Gross, 2004; Olweus, 1993; Solberg & Olweus, 2003). Various studies have distinguished between bullying and more general aggression (e.g., Griffin & Gross, 2004; Hay, Payne, & Chadwick, 2004). Although bullying and aggression are closely interrelated (Salmivalli & Nieminen, 2002), not all highly aggressive children and adolescents are involved in bullying behaviors (Hay et al., 2004), and not all types of bullying accompany aggressive acts (Olweus, 1993). Recently, emphasizing bullying behaviors which are intentional, repeated, and usually involve imbalance of power between the bully and the victim (Olweus, 1993), Wang, Iannotti, and Luk (2012) reviewed four types of bullying in the literature: physical bullying (e.g., hitting or pushing), verbal bullying (e.g., name-calling), social exclusion (e.g., ignoring or leaving out others), and spreading rumors (e.g., telling lies about others). In the current study, bullying behaviors consisted of two types of bullying, physical and verbal forms of bullying. Aggression included more general violent tendencies (i.e., gets into many fights) and oppositional characteristics (i.e., temper tantrums) as defined by Child Behavior Checklist (CBCL; Achenbach, 1991), whereas bullying behaviors included more intentional and relational components between the perpetrator and victim, such as pushing the victim, calling the victim names, or saying mean things, as conceptualized by Olweus on the Bully/Victim Questionnaire (e.g., Solberg & Olweus, 2003; Wang et al., 2012). Moreover, along with the physical and verbal offenses in the interpersonal relationships, identifying oneself as a “bully” was included in the bullying scale as opposed to questions identifying more general aggressive tendencies. We also confirmed two related, but different, constructs by testing construct validity (see the analysis of convergent and discriminant validities).

Research Hypothesis

Previous literature has consistently shown a close relation between bullying and both more general aggression and delinquency in childhood and adolescence (e.g., Baldry & Farrington, 2000; Bender & Lösel, 2011; Farrington, Loeber, Stallings, & Ttofi, 2011; Pepler et al., 2008; Scholte, Engels, Overbeek, de Kemp, & Haselager, 2007; Sourander et al., 2000; White & Loeber, 2008). Based on the previous link between bullying and externalizing behaviors, we first questioned whether the externalizing behaviors for the early-onset desisters (i.e., desisters) would carry over into adolescence or cease to exist. If the desisters would have the carry-over effect in adolescence, we hypothesized that their rates of change in the externalizing behaviors (measured by a slope term in this study) would be maintained similar to the rate of change for the early-onset persistent bullies (i.e., persisters). Consequently, at the end of the trajectory (measured by an intercept term), the desisters would show a high level of externalizing behaviors similar to the persisters but different from the non-involved group. As a competing hypothesis, if the desisters would show the cessation effect in adolescence, we hypothesized that the level of externalizing behaviors for the desisters would decrease over time as compared with the persisters, concurrent with the cessation of bullying in adolescence. Consequently, at the end of trajectory, they would show a low level of externalizing behaviors similar to the non-involved group but different from the persisters.

Second, regarding the late-onset bullies (i.e., late-onsetters), we questioned whether they would show an increase of externalizing behaviors in adolescence (onset effect) or no increase (resilience effect). If the late-onsetters would show the onset effect in adolescence, we hypothesized that their rate of change would increase over time, concurrent with the onset of bullying in adolescence, as compared with the children who were not engaged in bullying early on while maintaining their normal status later as well. Consequently, at the end of the trajectory, they would show a high level of externalizing behaviors similar to the desisters but different from the non-involved group. As a competing hypothesis, if the late-onsetters would show some resilience effect from the normal experience in their early life, we hypothesized that the rate of change in the externalizing behaviors would not increase as compared with that in the non-involved group. Finally, at the end of adolescent period, the late-onsetters would show a low level of externalizing behaviors similar to the non-involved and different from the persisters.

For a more accurate interpretation of the cessation and onset effects, it would be important to investigate the rate of changes in externalizing behaviors throughout the developmental trajectories. In addition, the magnitude of the subsequent outcomes for the transient groups has not been compared with those for the persisters or the non-involved group. For example, if the desisters were to show low levels of externalizing behaviors, similar to the levels of the non-involved group, and also show no differences in levels of externalizing behaviors from the persisters, this result would not meet the criteria of a cessation effect. This study is the first study to our knowledge that compared the magnitudes of the rate of change and the end point of trajectories driven by the transient groups (i.e., the desisters and the late-onsetters) with those showing consistent levels of bullying (i.e., the persisters and the non-involved).

Method

Participants

The data used in this study were drawn from the Springfield Child Development Project (SCDP), a longitudinal study of externalizing behaviors and other risk behaviors in middle childhood and adolescence. The sample consisted of a community-based, representative sample of 440 mother–child dyads from the city of Springfield, MA (Sheehan & Watson, 2008; Watson, Fischer, Burdzovic Andreas, & Smith, 2004). Sampling procedures for the original project were designed to produce a probability sample of all children living in the city of Springfield, MA, between 7 and 13 years of age, which provided a microcosm of ethnic groups, families, and neighborhood troubles found in many U.S. cities. Springfield exhibits a wide range of SES levels in its residents. To disengage from potential bias that could come from using a convenience sample of selected schools, the experimenters drew a random, computer-generated sample from a state-mandated census list in Massachusetts. After obtaining a street list of all women between the ages of 25 and 44 years (N = 4,518)—the age range in which mothers are most likely to have children between 7 and 13 years of age—a final sample of 440 eligible mother–child dyads (86.3% of those contacted) living in the city of Springfield, MA, was determined using recruitment letters and phone calls. Baseline data were collected when children were between 7 and 13 years old (Time 1, hereafter called T1; M = 9.92, SD = 1.99, N = 440). Follow-up interviews were conducted when the children ranged in age from 8 to 15 years (T2; M = 10.98, SD = 1.99, n = 391), 11 to 18 years (T3; M = 13.91, SD = 2.00, n = 357), and 12 to 19 years (T4; M = 14.95, SD = 2.01, n = 333). In the SCDP study, several age cohorts were followed over time (e.g., children between 7 and 13 years of age at T1 and children between 12 and 19 years of age at T4). Thus, instead of studying a developmental trajectory from meaningless assessment times, the hypotheses were tested by modeling changes in childhood and adolescent externalizing behaviors from age 7 to 19 as a function of bully status as measured at T1 through T4 (cf. Burdzovic Andreas & Watson, 2009; Raudenbush & Chan, 1993).

Descriptive statistics are presented in Table 1. Level 1 represents within-individual-level analysis throughout four time points (total 1,760 surveys), and Level 2 represents between-individual-level analysis (total 440 individuals). Some zip codes were over-sampled to obtain a more equally balanced representation of African American (38%), Hispanic American (29%), and European American (32%) children. Mothers’ years of education was used as an index of SES in the original data survey, and the means and standard deviations of the mother’s years of education for each ethnicity were 12.83 (2.11) for African Americans, 11.60 (2.55) for Hispanic Americans, and 13.53 (2.31) for European Americans. In the sample, 63% of African American, 48% of Hispanic American, and 67% of European American children had working mothers, and 27% of African American, 37% of Hispanic American, 72% of European American children had an intact, two-parent family.

Table 1.

Descriptive Statistics.

Within-Subject Variables	n	M	SD	Level of Model
Age-18	1,760	−5.52	2.89	1
(Age-18)²	1,760	38.85	33.07	1
Aggression	1,760	7.92	6.36	1
Delinquency	1,760	2.42	2.67	1
In childhood (age ≤ 12)
Childhood aggression	894	8.30	6.07
Childhood delinquency	894	2.13	2.11
Childhood bullying	894	2.04	2.52
In adolescence (age ≥ 13)
Adolescence aggression	866	7.54	6.63
Adolescence delinquency	866	2.71	3.12
Adolescence bullying	866	1.77	2.54
Between-Subject Variables	Total N	Subject n	%	Level of Model
Bully status groups	440			2
Desisters		72	16
Late-onsetters		63	14
Persisters		38	9
Non-involved		267	61
Gender	440			2
Male		242	55
Ethnicity	440			2
Majority		299	75
Maternal education	440	12.7 (M)	2.4 (SD)	2

Note. Ethnicity was dummy-coded so that African American and Hispanic children (minority) were contrasted with European American children (majority).

Measures

Fighting behavior

The Fighting Behavior Scale (Guerra & Slaby, 1990; Slaby & Guerra, 1988) measured bullying and victimization behaviors. In this study, we used a child reported bullying subscale for the perpetrators of bullying. Each group of bullies was assessed only bully-related items in the scale: a six-item bullying scale related to physical (e.g., “hit” or “push”) and verbal (e.g., “call names” or “say mean things”) bullying behaviors and self-identification of oneself as a bully (i.e., “Are you a bully?”). The frequency for each question was indicated using a 4-point scale (0 = never, 1 = sometimes, 2 = often, 3 = almost always). The Cronbach’s alpha values were .76, .62, .70, and .74 at T1 through T4.

To construct the accurately designated bully status groups (i.e., the desisters, the late-onsetters, the persisters, and the non-involved), we used an upper 25% cutoff score of bullying behaviors based on previous literature (e.g., Demaray & Malecki, 2003; Rueger, Malecki, & Demaray, 2011). For the score of bullying behaviors, we used the two mean scores of the childhood and adolescent bullying behaviors produced from all observed scores per person from the childhood period (age ≤ 12) and the adolescence period (age ≥ 13; for example, considering age 13 as a starting age of adolescence as was done in Copeland et al., 2013; Wong et al., 2013). Then, we applied the upper 25% cutoff to the total sample mean of childhood and adolescent bullying scores. Thus, in this study, desisters consist of children who experienced a higher average level of bullying in their childhood (above the upper quartile) but not in adolescence (below the lower quartile), and late-onsetters consist of the opposite cases. If the bullying scores were below the lower quartile in both childhood and adolescence, they were placed in the non-involved group, and if the scores were above the upper quartile in both periods, they were assigned to the persisters. For these classifications, means and standard deviations of bully scores during childhood and adolescence are presented in Table 2. When we analyzed the distribution of scores on the bullying scale, most participants (around 90%) in each group of desisters, late-onsetters, and persisters answered sometimes (1 point), often (2 points), or always (3 points) on three or more items of the total six items in the bullying scale. This finding implies that categorization was based on a wide-range of the bullying scale items rather than on one or two items only.

Table 2.

Classification of Bully Status Groups: Means and SDs of Bully Score During Childhood and Adolescence.

Time Periods	Cutoff	n	M (SD)	Four Bully Status Groups
Time Periods	Cutoff	n	M (SD)	Desisters	Late-Onsetters	Persisters	Non-Involved
Childhood (age ≤ 12)
Bully group	Upper quartile	110	4.64 (2.07)	Yes	No	Yes	No
Non-bully group	Lower quartile	275	1.10 (2.07)	No	Yes	No	Yes
	Total	385
Adolescence (age ≥ 13)
Bully group	Upper quartile	101	4.64 (2.38)	No	Yes	Yes	No
Non-bully group	Lower quartile	279	1.86 (1.20)	Yes	No	No	Yes
	Total	380

Note. In all, 385 participants out of 440 participants had at least one time point observed during childhood, and 380 participants out of 440 participants were observed at least once during adolescence.

CBCL

For externalizing behaviors, children’s delinquency (property and status violations) and aggression (aggression and opposition) were assessed by the mother’s reports of the CBCL-parent report that consisted of an 83-item questionnaire (with a 3-point scale; 0 = not true, 1 = somewhat or sometimes true, 2 = very true or often true; Achenbach, 1991). The questionnaire drew only 13 of these items from the Delinquent Behaviors subscale and 20 items from the Aggressive Behavior subscale. The CBCL has been shown to have high reliability and validity (Achenbach, 1991; Achenbach & Edelbrock, 1981). The aggression subscale had Cronbach’s alpha values of .87, .86, .90, and .86 at T1 through T4, and the delinquency subscale had Cronbach’s alpha values of .64, .73, .79, and .73.

Theoretically, aggression and bullying are different constructs but are interrelated. Thus, to test discriminant validity between the two factors of aggression and bullying, we performed the variance extracted test (Fornell & Larcker, 1981) based on Pepler’s et al. (2008) approach by conducting a confirmatory factor analysis (CFA) using AMOS (IBM SPSS AMOS 22). According to Fornell and Larcker (1981), discriminant validity is demonstrated if the square root of the average variance extracted per factor is larger than the correlation between the two factors. This study showed evidence of discriminant validity; the square root of the average variance of bullying and aggression from T1 through T4 (i.e., .74, .64, .71, and .72 for bullying and .73, .71, .75, and .73 for aggression) was larger than the correlation from T1 to T4 (i.e., .35, .22, .20, and .27). Regarding the convergent validity suggested by Anderson and Gerbing (1988), we obtained convergent validity for bullying and aggression through the factor loadings of measurement items which were significant at p < .001. Also, convergent validity was supported through exceeding the suggested item-to-total correlation threshold of .40 to .60 except for talk much, attacks, and bragging in aggression—but they were just below the threshold at .39, .38, and .33.

Missing Data

Out of 440 total mother–child dyads, 281 dyads (63.9%) had complete data at all four collection points. European American participants (25.5% attrition rate) were more likely to continue with the study than were African American participants (40.5%), χ²(1, n = 309) = 7.67, p = .006, and Hispanic participants (42.2%), χ²(1, n = 269) = 8.36, p = .004. Also, compared with participants who had any missing data, participants who had complete data were more likely to be engaged in bullying behaviors at T1 (M = 3.00, SD = 2.90 vs. M = 2.12, SD = 2.49), t(286.65) = −3.19, p = .002. The sample was slightly truncated regarding bullying behaviors, but the dropped participants were the more normative participants who were engaged in bullying less frequently. The significant differences above implied that our data were not “missing completely at random (MCAR)” but “missing at random (MAR).” Multiple imputation (MI) is one of the most robust and effective methods in modeling missing data, at least with MAR, as compared with using listwise deletion or older imputation (Durand & Wang, 2011; Raudenbush & Bryk, 2002). Thus, we created five data sets using MIs at each time point by the Markov Chain Monte Carlo (MCMC) imputation method (Yuan, 2011) in SAS 9.2, and ran those five data sets through the MI feature of HLM 7 (hierarchical linear modeling; Raudenbush, Bryk, & Congdon, 2010). Through the use of multiple data sets, the average of estimated parameters for variables was used to provide a single estimate in this study (Raudenbush & Bryk, 2002).

Analytic Plan

We used individual growth modeling (Raudenbush & Bryk, 2002; Singer & Willett, 2003) to test the hypotheses. Each child’s individual trajectory of externalizing behaviors was modeled by child age at a Level 1 model, and the variation in the trajectory was modeled by child bully status and controlling variables (i.e., gender, ethnicity, and years of maternal education) at a Level 2 model. More specifically, the Level 1 model represents the “within individual differences” that describes the individual growth trajectory of externalizing behaviors for each child over time. Based on the hypotheses, the children’s chronological age variable was centered on 18 years of age—The intercept in the Level 1 model was an indicator of behaviors at age 18 (a later adolescent status). The parameters of the intercept at (Age-18) and the (Age-18) slope and the quadratic age slope, that is, (Age-18)² at Level 1, were then modeled at Level 2, which represents “between individual differences” as a function of bully status and controlling variables. The intercepts and slopes of the transient bully groups were compared with those of the persisters and the non-involved.

Results

Table 3 presents the final models of the trajectories of aggression and delinquency. At each step, all the deviance tests were significant, demonstrating improved fit for each outcome. From the five imputations, five deviance tests were generated at each step of model building. As suggested by Brennan and his colleagues in their studies (cf. Civian, Richman, Shannon, Shulkin, & Brennan, 2008), from each step of the imputed models in delinquency and aggression, the median among the five models was compared with the five models estimated in the next step, and then the final deviance statistic was confirmed.

Table 3.

Results of Final Models: Trajectories of Aggression and Delinquency as a Function of Bully Status From Childhood to Adolescence.

	Aggression (Compared with the Persisters)				Delinquency (Compared with the Persisters)
	Coefficient	SE	t	p	Coefficient	SE	t	p
Fixed effects
For intercept at Age-18
Intercept (γ₀₀)	10.62	1.76	6.03	<.001	4.70	0.77	6.13	<.001
Gender (γ₀₁)	1.03	0.87	1.18	.241	0.12	0.51	0.23	.822
Ethnicity (γ₀₂₎	0.47	1.14	0.42	.682	0.80	0.59	1.37	.188
Education (γ₀₃)	−0.13	0.16	−0.77	.443	−0.04	0.09	−0.51	.614
Non-involved (γ₀₄)	−7.17	1.84	−3.89	<.001	−3.14	0.81	−3.86	<.001
Desisters (γ₀₅)	−6.52	2.55	−2.56	.027	−2.56	0.96	−2.68	.009
Late-onset (γ₀₆)	−3.98	1.90	−2.10	.040	−0.96	0.94	−1.02	.313
For Age-18 slope
Intercept (γ₁₀)	−0.31	0.30	−1.04	.312	0.24	0.09	2.54	.011
Gender (γ₁₁)	0.19	0.11	1.77	.081	0.04	0.06	0.65	.521
Ethnicity (γ₁₂)	−0.06	0.15	−0.39	.704	0.03	0.07	0.41	.690
Education (γ₁₃)	−0.02	0.02	−1.05	.296	−0.01	0.01	−0.77	.443
Non-involved (γ₁₄)	−0.38	0.24	−1.63	.120	−0.23	0.10	−2.41	.018
Desisters (γ₁₅)	−0.73	0.33	−2.33	.051	−0.28	0.11	−2.43	.018
Late-onset (γ₁₆)	−0.18	0.25	−0.73	.469	−0.00	0.11	0.01	.990
For (Age-18)² slope
Intercept (γ₂₀)	−0.03	0.01	−2.04	.050
(Compared with the Non-involved) For intercept at Age-18
*Desisters (γ_05n)	0.65	1.35	0.48	.637	0.57	0.64	0.89	.375
*Late-onset (γ_06n)	3.18	1.22	2.61	.010	2.18	0.68	3.22	.002
For Age-18 slope
*Desisters (γ_15n)	−0.35	0.17	−1.99	.059	−0.05	0.08	−0.58	.566
*Late-onset (γ_16n)	0.20	0.16	1.29	.200	0.23	0.08	2.84	.006
Random Effects	SD	Variance Components	χ	p	SD	Variance Components	χ	p
Level 1 Residual (r)	3.49	12.18			1.65	2.72
Level 2 Intercept (u₀)	6.37	40.57	1,431.92	<.001	3.36	11.28	1,762.18	<.001
Level 2 Slope (u₁)	.59	.35	694.63	<.001	.33	.11	838.97	<.001

Note. The coefficients of *Desisters and *Late-onset refer to the coefficients as compared with the non-involved group, instead of the persisters.

Developmental Trajectories of Aggression Based on Bully Status Groups

The results for the null model showed that average aggression varied significantly from individual to individual, the variance of u₀ (Level 2 residual), ${}^{\land}{σ_{intercept}^{2}}$ = 25.91, p < .001. The interclass correlation showed that 44.03% of the variance in aggression resided between children. The main effect of age was significant and negative, γ_age = −.32, p < .001, which means aggression decreased with increasing age. Also, the quadratic term of age, that is, slope of (Age-18)², was marginally significant, γ_(Age-18)² = −.03, p = .066. The random effects of the intercept and the slope of Age-18 were significant, ${}^{\land}{σ_{intercept}^{2}}$ = 45.04, ${}^{\land}{σ_{slope}^{2}}$ = 0.41, p < .001, but not for the slope of (Age-18)². The results for the random effects suggested that the addition of substantive predictors (e.g., bully status) in the intercept and slope at Level 2 can help explain significant individual differences in child aggression.

Table 3 presents the final model of aggression trajectories as a function of age and bully group status after controlling for gender, ethnicity, and maternal education. First of all, as compared with the persisters, the non-involved had a significantly lower level of aggression at age 18 (γ₀₄), but the rate of change was not significantly different between the two groups (γ₁₄). Second, in support for the hypothesis of the cessation effect, the desisters showed a significantly more precipitous decreasing trajectory over time compared with the persisters (marginal significance, γ₁₅, p = .051), and consequently, at age 18, the desisters had a significantly lower intercept of aggression than the persisters (γ₀₅) and showed no significant differences from the non-involved (γ_05n). Third, supporting neither the onset nor resilience effect, the late-onsetters were in between the persisters and the non-involved at age 18; they not only had a significantly lower intercept of aggression than the persisters (γ₀₆) but also showed a significantly higher intercept of aggression than the non-involved (γ_06n). Moreover, the slope of the late-onsetters was not different from that of either the persisters (γ₁₆) or the non-involved (γ_16n). The plotted growth curves are shown in Figure 1.

Figure 1.

Trajectories of aggression and delinquency depending on the bully status groups after controlling children’s gender, ethnicity, and maternal education.

Developmental Trajectories of Delinquency Based on Bully Status Groups

Moving now from the externalizing outcome of aggression to that of delinquency, the results for the null model showed that average delinquency varied significantly from individual to individual, the variance of u₀ (Level 2 residual), ${}^{\land}{σ_{intercept}^{2}}$ = 3.90, p < .001. The interclass correlation showed that 44.66 % of the variance in delinquency occurred between children. When the parameter of age (i.e., Age-18) was entered at Level 1, the main effect of the age variable was significant and positive, γ_Age-18 = .09, p = .002. The quadratic term of age was not significant, γ_(Age-18)² = .003, p = .551. The random effects of the intercept and the slope of Age-18 were significant, ${}^{\land}{σ_{intercept}^{2}}$ = 12.74, ${}^{\land}{σ_{slope}^{2}}$ = 0.12, p < .001.

First of all, as revealed in the final model of the delinquency trajectories (shown in Table 3), compared with the persisters, the non-involved had a significantly lower level of delinquency at age 18 (γ₀₄) and also had a significantly less steep trajectory (γ₁₄). Second, consistent with the results on aggression, the cessation effect was supported as well. The slope of the desisters was significantly different from than that of the persisters (γ₁₅) by showing a decreasing slope, and accordingly, at age 18, they showed a significantly lower level of delinquency than the persisters (γ₀₅) and similar levels to the non-involved group (γ_05n). Third, unlike the aggression results, we found an onset effect of bullying on delinquency. The late-onsetters showed a significantly steeper increasing slope over time than the non-involved group (γ_16n). Consequently, at age 18, they showed a significantly higher level of delinquency from the non-involved (γ_06n) and did not significantly differ from the persisters (γ₀₆). Figure 1 illustrates the growth curves of different groups based on this model.

Discussion

We asked two questions in this study: (a) whether the desisters, who had early bullying experiences but escaped from the experiences later, would have a carry-over effect of bullying on externalizing behaviors (which we called the cessation effect) and (b) whether the late-onsetters, who had no early bullying experiences but later began engaging in bullying, would show an onset effect on externalizing behaviors or alternatively at resilience effect (i.e., showing resilience to becoming involved in externalizing behaviors). In general, children who had shown bullying experiences in both periods (the persisters) presented higher levels of delinquency and aggression, which is not surprising, and children without any bullying experiences (the non-involved) showed lower levels of externalizing problems over time, which is also not surprising. The desisters and the late-onsetters were in between, supporting different hypotheses regarding the effects. Our results supported the cessation effect of bullying on aggression and delinquency for the desisters, and the onset effect for delinquency for the late-onsetters. The results revealed decreasing quadratic trajectories of aggression and increasing linear trajectories of delinquency. These results were consistent with previous studies showing that general levels of aggression decrease with age (Barker, Tremblay, Nagin, Vitaro, & Lacourse, 2006). However, the levels of delinquency that increase with age (Broidy et al., 2003) were found only for late-onsetters and persisters in this study—the non-involved showed little increases, and the desisters showed little decreases.

In support of the cessation effect, this result implies a recovery effect on externalizing behaviors for the desisters; the preceding experience of early bullying when it stopped during adolescence was not highly related to subsequent externalizing behaviors, in this case, delinquency and aggression. In contrast to assumptions of the long-term risk of the early experience of bullying on externalizing behaviors (e.g., Bender & Lösel, 2011; Broidy et al., 2003; Gibb et al., 2011; Olweus, 1993; Sourander et al., 2000; Sourander et al., 2007), the results show that early onset of bullying in childhood, when it desisted in adolescence, did not lead to more subsequent externalizing behaviors in adolescence. This result has an important implication related to prior findings. Previously, some findings suggested that there is a strong link between childhood bullying experiences and adolescent delinquency behaviors (e.g., Baldry & Farrington, 2000; Kumpulainen & Räsänen, 2000; Sourander et al., 2000; Sourander et al., 2007). The results from the study by Baldry and Farrington (2000) showed that about 70% of delinquents during adolescence were involved in bullying behaviors in their childhood, and it explained the relation between bullying and delinquency as bullying being an earlier state of delinquency. Broidy et al.’s (2003) and Nagin and Tremblay’s (1999)’s results showed physical aggression and opposition in childhood to be related to later violent delinquency. Also, Sourander et al. (2007) found that childhood bullying was related to adolescent delinquency such as property offense and traffic offense. Along these lines, a recent meta-analysis of 28 longitudinal studies measuring school-aged bullying behaviors and later outcomes indicated that school-aged children and adolescents who were involved in bullying behaviors showed delinquency and offending behavior later in life (Ttofi, Farrington, Lösel, & Loeber, 2011). However, it was not known whether the association remains consistent for bullying behaviors that either cease to exist or continue into adolescence. Our results support the contention that early experience with bullying in childhood did not have a carry-over effect on delinquency when the experience stopped or desisted in adolescence. A study by Huesmann et al. (2009) supported the recovery effect of previous childhood-limited and adolescence-limited aggressiveness on negative outcomes at age 48, and, along these lines, our study also indicated the recovery effect of childhood-limited bullying on externalizing behaviors in adolescence.

Regarding delinquency, despite the limited experiences in adolescence only, the late-onsetters showed concurrent delinquent behaviors in adolescence (onset effect). This implies that there is no clear resilience effect of the childhood non-bullying status on later externalizing behaviors when children are newly engaged in bullying behaviors in their adolescence.

However, regarding aggression, results supported neither the onset nor resilience effect; the level of aggression of the late-onsetters was between the levels of the two standard groups in adolescence. No onset effect on aggression is somewhat consistent with the results by O’Brennan, Bradshaw, and Sawyer (2009). As compared with the non-involved, they found that high school bullies showed less risk for aggression in a high school period (odds ratio [OR] = 2.4-3.9) than middle school bullies in a middle school period (OR = .2.5-6.6). This implies that the impact of bullying on aggression in each period of childhood and adolescence would decrease with age. However, little was known about the continuous impact of childhood-limited bullying on adolescent aggression or about the concurrent impact of adolescence-onset bullying on adolescent aggression based on a developmental trajectory. Our study used longitudinal perspectives and also found that the late-onset bullies (adolescence-onset bullies) showed significantly less risk of aggression in adolescence as compared with the persisters at the end of trajectory. Although this reduced risk of aggression in adolescence may imply a resilience effect for the late-onset bullies, the other results did not support the resilience effect from the normal experience in bullies’ early lives. This more conservative conclusion is taken because we also found late-onsetters’ aggression levels to be higher than those of the non-involved group, whose rate of change in aggression did not increase compared with that of the non-involved group. According to results found by Pepler et al. (2008), early risk-experienced group predicted moderately increasing bullying trajectories. However, early normal experience did not guarantee later decreasing aggression for late-onsetters. These incongruent results for neither onset nor resilience effects also emphasize the importance of having a dual comparison group as well as testing the rate of changes to investigate the magnitudes of the transient groups for more accurate and meaningful interpretation.

These different results of aggression and delinquency also imply that the timing of congruency between bullying and externalizing behaviors would depend on types of externalizing behaviors. Timing of bullying seems to be important for subsequent levels of delinquency. Delinquency level decreased as bullying behaviors desisted in adolescence (similar to the non-involved but different from the persisters), and delinquency level increased as bullying behaviors increased in adolescence (similar to the persisters but different from the non-involved). In aggression, aggression level decreased as bullying behaviors desisted (similar to the non-involved but different from the persisters). However, as bullying behavior increased in adolescence, aggression did not increase (lower than the persisters but higher than the non-involved). These findings suggest that whether the child was engaged in bullying in both periods or in either one of the periods is a more decisive factor than timing for subsequent aggression in adolescence.

The reason that there was an increased risk for delinquency but not as much as for aggression could be explained by two reasons given in previous literature. First, later bullies may have lesser risk for aggressive behaviors but spawn an increased risk for delinquency. As mentioned by Nansel et al. (2001), it could also reflect the general transition in adolescence of their direct and overt aggression into more indirect and covert forms of delinquency due to the acquisition of the social norm that aggression is not an appropriate behavior in social settings. Consistent with this explanation, Lahey et al. (2000) studied the prevalent rates in childhood and adolescence (7 to 19 years old) and found that aggression generally peaked near the middle of the age range and then decreased gradually, whereas delinquent behaviors increased at older ages. Second, during early adolescence, teens rely more on peers for social support, acceptance, and popularity, while seeking autonomy from their parents; thus, the increasing pressure to obtain peer acknowledgment, social hierarchy, and superiority over other students may be related to the increase in bullying (Espelage & Holt, 2001). Similarly, delinquency in adolescence is more likely to be conducted with peers and influenced by peers, instead of family (e.g., Church, Wharton, & Taylor, 2009; Sullivan, 2006). Moreover, risk-taking behaviors rarely occurred when a person is alone but significantly increased with the presence of peers among teenagers (Steinberg, 2007). Unlike the relation between bullying and aggression, this study indicated the relation between bullying and delinquency in adolescence was strong even though adolescents who bully did not have a history of bullying in childhood.

Limitations and Conclusions

First, based on previous research (e.g., Scholte et al., 2007; Sourander et al., 2000), this study helped us better understand the effect of peer bullying on externalizing behaviors rather than the effect of externalizing behaviors on children’s bullying experiences. However, this study cannot provide causal influences or the exact direction of the effects. It is also conceivable that the externalizing tendency may leave a child vulnerable to peer bullying experiences (e.g., Shakoor et al., 2012). Moreover, Salmivalli, Lagerspetz, Björkqvist, Österman, and Kaukiainen (1996) contended that how individuals accept bullying situations is determined by personal characteristics (e.g., an aggressive gene pattern or impulsivity) and social contexts (e.g., peer groups), but little is known of how different weights of temperament and environment might be related to engaging in bullying at different ages. A future study that accounts for the initial levels of aggression and delinquency could investigate how environments as well as a dispositional tendency to externalizing behaviors can lead to bullying involvement in childhood or adolescence.

Second, as Huesmann et al. (2009) used the cutoff scores to examine specific timing of aggression (e.g., childhood-limited and adolescent-limited aggressiveness) as predictors of various outcomes, this study also used cutoff scores to categorize different bullying status groups and participants’ ages into childhood and adolescence. Based on the objective of this study, to compare the rate of changes in externalizing trajectories given the timing of bullying experiences in specific time periods (i.e., childhood-limited bullies and adolescence-onset bullies, separately), the use of cutoff scores in this study was inevitable. By dichotomizing bullying behaviors rather than using a graduated scale of bullying, this bullying classification driven by the cutoff scores may not include all the information of the data. However, the four classifications used in this study have been confirmed by other researchers who investigated the patterns of bullying or aggressive behaviors throughout childhood and adolescence (Martino et al., 2008; Pepler et al., 2008). Despite the drawbacks of using cutoff scores, the results of this study on how bully status at different ages is related to trajectories of externalizing behavior suggest the importance of distinguishable prevention and intervention for specific transient bully groups in childhood and adolescence. The cessation effect of the childhood-limited bullying on later aggression and delinquency in adolescence and onset effect of adolescence-limited bullying on current delinquency in adolescence would lead us to emphasize the positive effects of early cessation and a concern for late-onset bullying as well.

Third, data on bullying were based on self-report, and certainly, social desirability factors could have impacted participants’ responses. Peer reports might have been preferable, although peer reports were not available in the longitudinal data set. Similarly, maternal reports about child aggression and delinquency might be less accurate because mothers tend to have less access to their children’s behaviors as they get older, and mothers may have a tendency to not acknowledge actual levels of these aggressive and antisocial behaviors in their children. For further study, multiple raters would be preferable. Finally, in this study, bullying was defined by the specific behaviors of physical and verbal bullying behaviors, but we did not account for other forms of victimization possibly involved within bullying behaviors, nor did we account for the extent to which the bullying repeatedly targeted specific victims. We also did not account for children who were involved in bullying behaviors and victimization. Thus, there is a need for further research to provide more generalizability of these results.

In sum, the desisters showed the cessation effect on aggression and delinquency, but the late-onsetters supported the onset effect on delinquency. Previously, evidence of differential bullying program effectiveness for childhood or adolescence had been inconclusive (Smith & Ananiadou, 2003), and there was evidence of more effectiveness for the older than the younger children due to their increasing cognitive abilities, rational decision-making ability, and decreasing impulsiveness (Ttofi & Farrington, 2011). Nevertheless, most anti-bullying programs have focused on bullying behaviors in childhood periods more than bullying behaviors in adolescent periods (see Baldry & Farrington, 2007 for a review of detailed programs). Finally, these results do not undermine the importance of intervention for childhood bullying behaviors, but they imply we need to pay more attention to the late-onset bullies as well. Also, programs that target adolescents who bully should also intervene in their spilled-over delinquent behaviors.

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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by NICHD Grant HD032371 to Malcolm Watson (P.I.).

Author Biographies

Yoona Lee is a postdoctoral fellow and an adjunct faculty in the Department of Psychology at Brandeis University, Waltham, MA. She received her PhD in developmental psychology from Brandeis University and her master degrees in psychology from Yonsei University, Seoul, Korea, and from Boston University, Boston, MA. Her major research interests include child maltreatment, child discipline, bullying, and children/adolescent psychosocial behaviors in developmental and cross-cultural perspectives.

Xiaodong Liu is an assistant professor of psychology at Brandeis University. He received his EdD in human development and psychology from Harvard Graduate School of Education. Before joining Brandeis University, he worked as an assistant professor in quantitative methods, measurement, and evaluation in College of Education at Ohio State University. His main interests are in applied statistics (modeling in particular) and in the methods related to research on development (broadly defined) and child development in particular.

Malcolm W. Watson received his PhD in developmental psychology from the University of Denver in 1977 and has been at Brandeis University in Waltham, MA, ever since. He is the George and Frances Levin professor of psychology at Brandeis and has served as the dean of the Graduate School of Arts and Sciences. He has published numerous articles and edited books primarily in three areas of child development: the development of play and fantasy in children, children’s developing social role concepts, and antecedents of aggression and bullying in children and adolescents. His research has been funded by NIMH and NICHD.

References

Achenbach

T. M.

(1991). Manual for the Child Behavior Checklist/4-16 and 1991 profile. Burlington, VT: University Associates in Psychiatry.

Achenbach

T. M.

Edelbrock

C. S.

(1981). Behavioral problems and competencies reported by parents of normal and disturbed children aged four through sixteen. Monographs of the Society for Research in Child Development, 46(1), 1-82. doi:10.2307/1165983

Anderson

J. C.

Gerbing

D. W.

(1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103, 411-423. doi:10.1037/0033-2909.103.3.411

Baldry

A. C.

Farrington

D. P.

(2000). Bullies and delinquents: Personal characteristics and parental styles. Journal of Community & Applied Social Psychology, 10, 17-31. doi:10.1002/(SICI)1099-1298(200001/02)10:1<17::AID-CASP526>3.0.CO;2-M

Baldry

A. C.

Farrington

D. P.

(2007). Effectiveness of programs to prevent school bullying. Victims & Offenders, 2, 183-204. doi:10.1080/15564880701263155

Barker

E. D.

Tremblay

R. E.

Nagin

D. S.

Vitaro

Lacourse

(2006). Development of male proactive and reactive physical aggression during adolescence. Journal of Child Psychology and Psychiatry, 47, 783-790. doi:10.1111/j.1469-7610.2005.01585.x

Bender

Lösel

(2011). Bullying at school as a predictor of delinquency, violence and other anti-social behaviour in adulthood. Criminal Behaviour and Mental Health, 21, 99-106. doi:10.1002/cbm.799

Broidy

L. M.

Nagin

D. S.

Tremblay

R. E.

Bates

J. E.

Brame

Dodge

K. A.

Vitaro

(2003). Developmental trajectories of childhood disruptive behaviors and adolescent delinquency: A six-site, cross-national study. Developmental Psychology, 39, 222-245. doi:10.1037/0012-1649.39.2.222

Burdzovic Andreas

Watson

M. W.

(2009). Moderating effects of family environment on the association between children’s aggressive beliefs and their aggression trajectories from childhood to adolescence. Development and Psychopathology, 21, 189-205. doi:10.1017/S0954579409000121

10.

Caspi

Elder

G. H.

Bem

D. J.

(1987). Moving against the world: Life-course patterns of explosive children. Developmental Psychology, 23, 308-313. doi:10.1037/0012-1649.23.2.308

11.

Church

W. I.

Wharton

Taylor

J. K.

(2009). An examination of differential association and social control theory: Family systems and delinquency. Youth Violence and Juvenile Justice, 7, 3-15. doi:10.1177/1541204008324910

12.

Civian

J. T.

Richman

A. L.

Shannon

L. L.

Shulkin

Brennan

R. T.

(2008). Using a multi-organization database: Research methods, strengths, and limitations. Community, Work & Family, 11, 139-148. doi:10.1080/1366880080205179

13.

Copeland

W. E.

Adair

C. E.

Smetanin

Stiff

Briante

Colman

Angold

(2013). Diagnostic transitions from childhood to adolescence to early adulthood. Journal of Child Psychology and Psychiatry, 54, 791-799. doi:10.1111/jcpp.12062

14.

Demaray

M. K.

Malecki

C. K.

(2003). Perceptions of the frequency and importance of social support by students classified as victims, bullies, and bully/victims in an urban middle school. School Psychology Review, 32, 471-489.

15.

Durand

V. M.

Wang

(2011). Clinical trials. In Thomas

J. C.

Hersen

(Eds.), Understanding research in clinical and counseling psychology (2nd ed., pp. 201-227). New York, NY: Routledge.

16.

Espelage

D. L.

Holt

M. L.

(2001). Bullying and victimization during early adolescence: Peer influences and psychosocial correlates. Journal of Emotional Abuse, 2, 123-142. doi:10.1300/J135v02n02_08

17.

Farrington

D. P.

Loeber

Stallings

Ttofi

M. M.

(2011). Bullying perpetration and victimization as predictors of delinquency and depression in the Pittsburgh youth study. Journal of Aggression, Conflict and Peace Research, 3, 74-81.

18.

Fornell

Larcker

D. F.

(1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18, 39-50. doi:10.2307/3151312

19.

Gibb

S. J.

Horwood

L. J.

Fergusson

D. M.

(2011). Bullying victimization/perpetration in childhood and later adjustment: Findings from a 30 year longitudinal study. Journal of Aggression, Conflict and Peace Research, 3, 82-88.

20.

Griffin

R. S.

Gross

A. M.

(2004). Childhood bullying: Current empirical findings and future directions for research. Aggression and Violent Behavior, 9, 379-400. doi:10.1016/S1359-1789(03)00033-8

21.

Guerra

N. G.

Slaby

R. G.

(1990). Cognitive mediators of aggression in adolescent offenders: II. Intervention. Developmental Psychology, 26, 269-277. doi:10.1037/0012-1649.26.2.269

22.

Hay

D. F.

Payne

Chadwick

(2004). Peer relations in childhood. Journal of Child Psychology and Psychiatry, 45, 84-108. doi:10.1046/j.0021-9630.2003.00308

23.

Huesmann

L. R.

Dubow

E. F.

Boxer

(2009). Continuity of aggression from childhood to early adulthood as a predictor of life outcomes: Implications for the adolescent-limited and life-course-persistent models. Aggressive Behavior, 35, 136-149. doi:10.1002/ab.20300

24.

Kim

Y. S.

Boyce

W. T.

Koh

Leventhal

B. L.

(2009). Time trends, trajectories, and demographic predictors of bullying: A prospective study in Korean adolescents. Journal of Adolescent Health, 45, 360-367. doi:10.1016/j.jadohealth.2009.02.005

25.

Kumpulainen

Räsänen

(2000). Children involved in bullying at elementary school age: Their psychiatric symptoms and deviance in adolescence: An epidemiological sample. Child Abuse & Neglect, 24, 1567-1577. doi:10.1016/S0145-2134(00)00210-6

26.

Lahey

B. B.

Schwab-Stone

Goodman

S. H.

Waldman

I. D.

Canino

Rathouz

P. J.

Jensen

P. S.

(2000). Age and gender differences in oppositional behavior and conduct problems: A cross-sectional household study of middle childhood and adolescence. Journal of Abnormal Psychology, 109, 488-503. doi:10.1037/0021-843X.109.3.488

27.

Martino

S. C.

Ellickson

P. L.

Klein

D. J.

McCaffrey

Edelen

M. O.

(2008). Multiple trajectories of physical aggression among adolescent boys and girls. Aggressive Behavior, 34, 61-75. doi:10.1002/ab.20215

28.

Moffitt

T. E.

Caspi

(2001). Childhood predictors differentiate life-course persistent and adolescence-limited antisocial pathways among males and females. Development and Psychopathology, 13, 355-375. doi:10.1017/S0954579401002097

29.

Nagin

Tremblay

R. E.

(1999). Trajectories of boys’ physical aggression, opposition, and hyperactivity on the path to physically violent and nonviolent juvenile delinquency. Child Development, 70, 1181-1196. doi:10.1111/1467-8624.00086

30.

Nansel

T. R.

Overpeck

Pilla

R. S.

Ruan

W. J.

Simons-Morton

Scheidt

(2001). Bullying behaviors among US youth: Prevalence and association with psychosocial adjustment. Journal of the American Medical Association, 285, 2094-2100. doi:10.1001/jama.285.16.2094

31.

O’Brennan

L. M.

Bradshaw

C. P.

Sawyer

A. L.

(2009). Examining development differences in the social-emotional problems among frequent bullies, victims, and bully/victims. Psychology in the Schools, 46, 100-115. doi:10.1002/pits.20357

32.

Olweus

(1993). Bullying at school: What we know and what we can do. Oxford, UK: Blackwell.

33.

Patterson

G. R.

DeBaryshe

B. D.

Ramsey

(1989). A developmental perspective on antisocial behavior. American Psychologist, 44, 329-335. doi:10.1037/0003-066X.44.2.329

34.

Pepler

Jiang

Craig

Connolly

(2008). Developmental trajectories of bullying and associated factors. Child Development, 79, 325-338. doi:10.1111/j.1467-8624.2007.01128.x

35.

Raudenbush

S. W.

Bryk

A. S.

(2002). Hierarchical linear models: Applications and data analysis methods. Newbury Park, CA: Sage.

36.

Raudenbush

S. W.

Bryk

Congdon

(2010). HLM 7: Hierarchical linear and nonlinear modeling. Skokie, IL: Scientific Software International.

37.

Raudenbush

S. W.

Chan

(1993). Application of a hierarchical linear model to the study of adolescent deviance in an overlapping cohort design. Journal of Consulting and Clinical Psychology, 61, 941-951. doi:10.1037/0022-006X.61.6.941

38.

Rueger

S. Y.

Malecki

C. K.

Demaray

M. K.

(2011). Stability of peer victimization in early adolescence: Effects of timing and duration. Journal of School Psychology, 49, 443-464. doi:10.1016/j.jsp.2011.04.005

39.

Ryan

Smith

J. D.

(2009). Antibullying programs in schools: How effective are evaluation practices? Prevention Science, 10, 248-259. doi:10.1007/s11121-009-0128-y

40.

Salmivalli

Lagerspetz

Björkqvist

Österman

Kaukiainen

(1996). Bullying as a group process: Participant roles and their relations to social status within the group. Aggressive Behavior, 22, 1-15. doi:10.1002/(SICI)1098-2337(1996)22:1<1::AID-AB1>3.0.CO;2-T

41.

Salmivalli

Nieminen

(2002). Proactive and reactive aggression among school bullies, victims, and bully-victims. Aggressive Behavior, 28, 30-44. doi:10.1002/ab.90004

42.

Schäfer

Korn

Brodbeck

F. C.

Wolke

Schulz

(2005). Bullying roles in changing contexts: The stability of victim and bully roles from primary to secondary school. International Journal of Behavioral Development, 29, 323-335. doi:10.1177/01650250544000107

43.

Scholte

R. J.

Engels

R. E.

Overbeek

de Kemp

R. T.

Haselager

G. T.

(2007). Stability in bullying and victimization and its association with social adjustment in childhood and adolescence. Journal of Abnormal Child Psychology, 35, 217-228. doi:10.1007/s10802-006-9074-3

44.

Shakoor

Jaffee

S. R.

Bowes

Ouellet-Morin

Andreou

Happé

Arseneault

(2012). A prospective longitudinal study of children’s theory of mind and adolescent involvement in bullying. Journal of Child Psychology and Psychiatry, 53, 254-261. doi:10.1111/j.1469-7610.2011.02488.x

45.

Sheehan

M. J.

Watson

M. W.

(2008). Reciprocal influences between maternal discipline techniques and aggression in children and adolescents. Aggressive Behavior, 34, 245-255. doi:10.1002/ab.20241

46.

Singer

J. D.

Willett

J. B.

(2003). Applied longitudinal data analysis: Modeling change and event occurrence. New York, NY: Oxford University Press.

47.

Slaby

R. G.

Guerra

N. G.

(1988). Cognitive mediators of aggression in adolescent offenders: I. Assessment. Developmental Psychology, 24, 580-588. doi:10.1037/0012-1649.24.4.580

48.

Smith

P. K.

Ananiadou

(2003). The nature of school bullying and the effectiveness of school-based interventions. Journal of Applied Psychoanalytic Studies, 5, 189-209. doi:10.1023/A:1022991804210

49.

Solberg

M. E.

Olweus

(2003). Prevalence estimation of school bullying with the Olweus Bully/Victim Questionnaire. Aggressive Behavior, 29, 239-268. doi:10.1002/ab.10047

50.

Sourander

Helstelä

Helenius

Piha

(2000). Persistence of bullying from childhood to adolescence—A longitudinal 8-year follow-up study. Child Abuse & Neglect, 24, 873-881. doi:10.1016/S0145-2134(00)00146-0

51.

Sourander

Jensen

Rönning

J. A.

Niemelä

Helenius

Sillanmäki

Almqvist

(2007). What is the early adulthood outcome of boys who bully or are bullied in childhood? The Finnish “From a Boy to a Man” study. Pediatrics, 120, 397-404. doi:10.1542/peds.2006-2704

52.

Steinberg

(2007). Risk taking in adolescence: New perspectives from brain and behavioral science. Current Directions in Psychological Science, 16, 55-59. doi:10.1111/j.1467-8721.2007.00475.x

53.

Stouthamer-Loeber

Loeber

Wei

Farrington

D. P.

Wikström

P. H.

(2002). Risk and promotive effects in the explanation of persistent serious delinquency in boys. Journal of Consulting and Clinical Psychology, 70, 111-123. doi:10.1037/0022-006X.70.1.111

54.

Sullivan

C. J.

(2006). Early adolescent delinquency: Assessing the role of childhood problems, family environment, and peer pressure. Youth Violence and Juvenile Justice, 4, 291-313. doi:10.1177/1541204006292656

55.

Ttofi

M. M.

Farrington

D. P.

(2011). Effectiveness of school-based programs to reduce bullying: A systematic and meta-analytic review. Journal of Experimental Criminology, 7, 27-56. doi:10.1007/s11292-010-9109-1

56.

Ttofi

M. M.

Farrington

D. P.

Lösel

(2012). School bullying as a predictor of violence later in life: A systematic review and meta-analysis of prospective longitudinal studies. Aggression and Violent Behavior, 17, 405-418. doi:10.1016/j.avb.2012.05.002

57.

Ttofi

M. M.

Farrington

D. P.

Lösel

Loeber

(2011). The predictive efficiency of school bullying versus later offending: A systematic/meta-analytic review of longitudinal studies. Criminal Behaviour and Mental Health, 21, 80-89. doi:10.1002/cbm.808

58.

Underwood

M. K.

Beron

K. J.

Rosen

L. H.

(2009). Continuity and change in social and physical aggression from middle childhood through early adolescence. Aggressive Behavior, 35, 357-375. doi:10.1002/ab.20313

59.

Wang

Iannotti

R. J.

Luk

J. W.

(2012). Patterns of adolescent bullying behaviors: Physical, verbal, exclusion, rumor, and cyber. Journal of School Psychology, 50, 521-534. doi:10.1016/j.jsp.2012.03.004

60.

Watson

M. W.

Fischer

K. W.

Burdzovic Andreas

Smith

K. W.

(2004). Pathways to aggression in children and adolescents. Harvard Educational Review, 74, 404-430.

61.

White

N. A.

Loeber

(2008). Bullying and special education as predictors of serious delinquency. Journal of Research in Crime & Delinquency, 45, 380-397. doi:10.1177/0022427808322612

62.

Wicks-Nelson

Israel

A. C.

(2006). Behavior disorders of childhood (6th ed.). Englewood Cliffs, NJ: Prentice Hall.

63.

Wong

T. L.

Loeber

Slotboom

Bijleveld

C. H.

Hipwell

A. E.

Stepp

S. D.

Koot

H. M.

(2013). Sex and age differences in the risk threshold for delinquency. Journal of Abnormal Child Psychology, 41, 641-652. doi:10.1007/s10802-012-9695-7

64.

Yuan

(2011). Multiple imputation using SAS software. Journal of Statistical Software, 45, 1-25.