Journal of Affective Disorders 246 (2019) 603–610

Contents lists available at ScienceDirect

Journal of Affective Disorders

journal homepage: www.elsevier.com/locate/jad

Research paper

Does bullying victimization really influence adolescents’ psychosocial T

problems? A three-wave longitudinal study in China
⁎ ⁎
Xiao-Wei Chua,b, Cui-Ying Fana,b, , Shuai-Lei Liana,b, Zong-Kui Zhoua,b,
a
Key Laboratory of Adolescent Cyberpsychology and Behavior (CCNU), Ministry of Education, Wuhan 430079, China
b
School of Psychology, Central China Normal University, Wuhan 430079, China

A R T I C LE I N FO A B S T R A C T

Keywords: Background: Traditional bullying and cyberbullying victimization are associated with adolescents’ mental health
Traditional bullying problems, but the temporal sequence of these relations remains unclear. The present study analyzed the temporal
Cyberbullying and reciprocal relationships between both types of bullying victimization and six frequent psychosocial problems
Psychosocial problems during adolescence: depression, general anxiety, stress, low self-esteem, social anxiety, and loneliness.
Adolescents
Methods: A total of 661 Chinese adolescents in grades 7 and 8 at the initial wave of the study responded to the
Longitudinal
survey items assessing their bullying victimization experiences and self-reported psychosocial problems. This
research was conducted at three separate time points, over a year and a half period.
Results: Bullying victimization did not predict adolescents’ psychosocial problems. Instead, adolescents with
psychosocial problems were more likely to experience bullying victimization. Depression and general anxiety
were two common predictors of traditional bullying and cyberbullying victimization. Stress was a specific
predictor of traditional bullying victimization, whereas self-esteem, social anxiety, and loneliness were specific
predictors of cyberbullying victimization. Psychosocial problems did not differentially predict the two types of
bullying victimization. Boys with higher levels of loneliness were more likely than girls to experience cyber-
bullying victimization.
Limitations: The present study is limited in terms of its sample selection and self-reported instruments.
Conclusions: This study emphasizes adolescents’ psychosocial problems as the potential risk factors for their
future bullying victimization. Education professionals should pay special attention to adolescents’ mental health
problems when formulating bullying interventions and programs.

1. Introduction were associated with numerous psychological health problems, such as

depression, anxiety, and low self-esteem (Fisher et al., 2016; Reijntjes
Bullying victimization is a growing concern for adolescents. et al., 2010).
Researchers have differentiated two distinct forms: traditional bullying In China, bullying, especially cyberbullying, is a relatively new
and cyberbullying victimization. These two behaviors often involve phenomenon that attracts increased attention from researchers.
three primary features of bullying: intentional aggression, repetition, Prevalence rates of traditional bullying victimization have ranged from
and power imbalance between the perpetrator(s) and the victim. 2% to 66%, whereas rates of cyberbullying victimization have ranged
Cyberbullying has some unique features that distinguish it from tradi- from 12% to 75% in Chinese adolescents (Chan and Wong, 2015; Chu
tional bullying, including dependence on technical skills, perpetrators’ et al., 2018a). Research that sampled Chinese adolescents has revealed
anonymity, accessibility of the victims, greater potential audience, and cross-sectional links between bullying victimization and mental health
permanence of bullying incident (Kowalski et al., 2014). A meta-ana- problems, such as depression, anxiety, low self-esteem, and loneliness
lysis of 80 English-language studies in adolescents has revealed mean (Chu et al., 2018b; Huang et al., 2013). However, no research to date
prevalence rates of 36% for traditional bullying victimization and 15% has investigated the longitudinal relationships between bullying victi-
for cyberbullying victimization (Modecki et al., 2014). Meta-analytical mization and psychosocial problems in Chinese adolescents. The tem-
studies also demonstrated that both types of bullying victimization poral sequence of these relations also remains unexamined. To address

⁎
Corresponding author at. School of Psychology, Central China Normal University, Wuhan 430079, China.
E-mail addresses: [email protected] (X.-W. Chu), [email protected] (C.-Y. Fan), [email protected] (S.-L. Lian),
[email protected] (Z.-K. Zhou).

https://doi.org/10.1016/j.jad.2018.12.103
Received 24 September 2018; Received in revised form 20 December 2018; Accepted 24 December 2018
Available online 26 December 2018
0165-0327/ © 2018 Elsevier B.V. All rights reserved.
X.-W. Chu et al. Journal of Affective Disorders 246 (2019) 603–610

these questions can shed light on the etiology and causal relationship of the reduced chances for perpetrators’ empathy and remorse, the large
bullying victimization and mental health problems. It also can help to potential audience, and the permanence of bullying incident. In addi-
formulate targeted measures aimed at reducing adolescents’ psycholo- tion, research suggested that psychosocial problems (i.e., social anxiety
gical and behavioral health problems. For these reasons, this study and loneliness) were more predictive of online than of real-life bullying
examined the temporal and bidirectional relationships between bul- victimization (van den Eijnden et al., 2014). One possible explanation is
lying victimization and six frequent psychosocial problems during that adolescents with psychosocial problems tend to engage in online
adolescence: depression, general anxiety, stress, low self-esteem, social contact with strangers as a kind of compensation for their un-
anxiety, and loneliness. The first four problems may result from ado- satisfactory relationships in real life (Valkenburg and Peter, 2007). The
lescents’ subjective appraisals about intrapersonal state, whereas the risk of being bullied online will be particularly high in case of inter-
last two problems may result from their subjective appraisals about acting with strangers (Slovak and Singer, 2011). This study hypothe-
interpersonal relationships. Additional objectives were to analyze the sized that the bidirectional relationships between cyberbullying victi-
differential relations of traditional bullying and cyberbullying victimi- mization and psychosocial problems would be stronger than the
zation with psychosocial problems and the extent to which these rela- relations between traditional bullying victimization and psychosocial
tions would differ by gender. problems.
Two models were used to explain the relation between stressors and Higher susceptibility to interpersonal stress in female adolescents
psychosocial problems, namely stress exposure model and stress gen- compared to males raises the possibility of gender differences in the
eration model. The stress exposure model implies that exposure to ne- association between bullying victimization and psychosocial problems.
gative events precedes and increases the risk for various health-related Two longitudinal research has demonstrated that the association be-
outcomes (e.g., depression, anxiety, and low self-esteem). Psychological tween bullying victimization and psychosocial symptoms was stronger
problems are seen as the individual's passive reactions to stressors, in females than in males (Bannink et al., 2014; Kim et al., 2018). When
particularly interpersonal stressors. Bullying victimization causes ado- encountering interpersonal stressors, girls tend to focus inward and
lescents’ psychosocial problems because it might be interpreted as ne- ruminate on the potential causes and consequences of their problematic
gative peer evaluation or social exclusion, which may in turn reinforce relationships (e.g., bullying victimization) (Nolen-Hoeksema and
their negative self-evaluations (van den Eijnden et al., 2014). The Girgus, 1994). In this sense, girls may experience more intense and
model has been supported by previous longitudinal studies revealing prolonged tension as a result of interpersonal stress (Hankin et al.,
psychosocial problems as the outcomes of bullying victimization (Fahy 1998). In addition, female adolescents with psychosocial problems tend
et al., 2016; Stapinski et al., 2015). The stress generation model posits to be more vulnerable, self-abased, and socially withdrawn, which may
that health-related problems precede and increase the occurrence of elicit others’ bullying behaviors more frequently. However, male ado-
negative life events (e.g., bullying victimization). The person is viewed lescents tend to distract themselves from the negative emotions by
as an actor shaping, construing, and responding to the environment. engaging in dangerous activities (e.g., drinking alcohol) (Nolen-
Certain characteristics (e.g., maladaptive coping and problem-solving Hoeksema and Morrow, 1991). This study hypothesized that the bi-
styles) and circumstances (e.g., parental psychopathology) developed directional relations between bullying victimization and psychosocial
by the individuals with psychosocial problems may contribute to their problems would be stronger in girls than in boys.
experience of excessive stressors (Hammen, 2006). Regarding bullying,
research has demonstrated that psychological health problems were 2. Methods
precursors of bullying victimization (Gibb and Hanley, 2010; Yang
et al., 2013). 2.1. Procedure and participants
Based on above theoretical analyses, a reciprocal relationship be-
tween bullying victimization and psychosocial problems may be es- We conducted a three-wave panel study with 6-month intervals
tablished. In a meta-analysis, Reijntjes and colleagues (2010) concluded among Chinese adolescents between November 2016 and November
a reciprocal relationship between traditional bullying victimization and 2017. The data were collected from an ordinary junior high school in
internalizing problems (e.g., depression and anxiety). The conclusion Wuhan city in central China. The Ethics Committee of the authors’
was also supported in other studies (Gámez-Guadix et al., 2013; Rose university approved this study. Informed consent was obtained from the
and Tynes, 2015; Siegel et al., 2009). For example, Siegel and collea- headmaster and participants. Students were told that they could refuse
gues (2009) found that traditional bullying victimization was both a or discontinue participation at any time without penalty, and that any
predictor and consequence of social anxiety over time. Regarding cy- of their responses would not be disclosed to anyone. They finally
berbullying victimization, its bidirectional relations with depression completed a paper-pencil questionnaire written in Chinese during one
and anxiety have also been demonstrated (Gámez-Guadix et al., 2013; class hour.
Rose and Tynes, 2015). However, some other research only supports At T1, the sample comprised 661 adolescents (39.2% girls) in grades
the stress generation model. Depressive symptoms were found as 7 (49.6%) and 8 (50.4%). They ranged in age from 11 to 15 years old
antecedents of offline and online bullying victimization rather than as (M = 12.86, SD = 0.73). The majority of participants (92.4%) were in a
outcomes (Frison et al., 2016; Tran et al., 2012). Social anxiety and two-parent family. The distribution of family income was described as
loneliness predicted increases in latter offline and online bullying vic- 3.5% low, 93.4% medium, and 3.2% high. On average adolescents
timization rather than the reverse (Pabian and Vandebosch, 2016; van came from families where parents had less than a senior high school
den Eijnden et al., 2014). Taken together, this study predicted that the education. Of the adolescents, 598 (78.8%) participated in all three
relationship between bullying victimization and psychosocial problems measurement waves, 135 (17.8%) participated in two waves, and 26
would be bidirectional, but the predictive effect of psychosocial pro- (3.4%) participated in one wave. Regarding the sample attrition or
blems on bullying victimization may be stronger than the reverse. addition, 27 dropped-out and 90 dropped-in at T2 (n = 724), and 43
Given the substantial overlap between traditional bullying and cy- dropped-out and 24 dropped-in at T3 (n = 705). We further compared
berbullying victimization (Kowalski et al., 2014), a key issue that needs adolescents who participated at all three waves (n = 598) with those
to be considered is whether either form is associated with psychosocial who dropped-out or dropped-in (n = 161). Following Goodman and
problems after controlling for the other. Cyberbullying victimization Blum (1996), we regressed a dichotomous variable (0 = those who
usually results in stronger negative psychosocial effects than traditional participated in all waves; 1 = those who dropped-out or dropped-in) on
bullying victimization (Campbell et al., 2012). This may be due to some all research variables. The analyses only showed a small effect of social
core features of cyberbullying (Kim et al., 2018; van den Eijnden et al., anxiety on participants’ status (B = −0.39, p < .05), indicating that the
2014), including the difficulty for victims to escape from cyberbullying, data were missing at random generally.

604
X.-W. Chu et al. Journal of Affective Disorders 246 (2019) 603–610

Fig. 1. The proposed conceptual model.

Note. TBV = Traditional bullying victimization; CBV = Cyberbullying victimization; T1 = Time 1; T2 = Time 2; T3 = Time 3.

2.2. Measures discomforts in the presence of others. A sample item is “I feel anxious
when I speak in front of a group” . Response options included “0 = not
2.2.1. Traditional bullying victimization (TBV) at all like me”, “2 = a little like me”, “3 = somewhat like me”, and
We used the victimization subscale of Traditional Bullying Scale “4 = a lot like me” . Cronbach's alpha in this sample was 0.68, 0.74, and
(Li et al., 2012). This subscale contains six items regarding the fre- 0.79 at T1–T3, respectively.
quency with which adolescents had experienced three forms of tradi-
tional bullying behaviors (i.e., verbal, physical, and relational) during 2.2.6. Loneliness
the previous six months. A sample item is “People hit, kicked, punched An eight-item short form of the UCLA Loneliness Scale (ULS–8)
or threatened me” . Response options included “1 = never”, “2 = once (Hays and DiMatteo, 1987) was used to assess adolescents’ loneliness.
or twice”, “3 = 2–3 times a month”, “4 = once a week”, and This scale describes individuals’ feelings of being cut off or separated
“5 = several times a week” . Cronbach's alpha in this sample was 0.77, from others. A sample item is “There is no one I can turn to” . Parti-
0.79, and 0.80 at T1–T3, respectively. cipants rated each item on a 4-point scale ranging from 1 (never) to 4
(always). The reliability and validity of this scale have been supported
2.2.2. Cyberbullying victimization (CBV) among Chinese students (Ma, 2018; Wu and Yao, 2008). Cronbach's
We used the victimization subscale of Chinese version of the Revised alpha in this sample was 0.81, 0.76, and 0.79 at T1–T3, respectively.
Cyberbullying Inventory (Chu and Fan, 2017). This subscale describes
14 activities of being bullied electronically during the previous six 2.3. Data analysis
months. A sample item is “People sent threatening or hurtful text
messages to me” . Response options included “1 = never”, “2 = once”, To test hypotheses, six latent cross-lagged panel analyses were se-
“3 = twice or three times”, and “4 = more than three times” . Cron- parately conducted with Mplus 7.0. Due to the non-normality of our
bach's alpha in this sample was 0.85, 0.86, and 0.87 at T1–T3, re- data, the research employed the robust maximum likelihood estimation
spectively. method. Missing data were handled using the full information max-
imum likelihood estimator. We constructed three parcels as indicators
2.2.3. Depression, general anxiety, and stress of each latent variable. The item-to-construct balance parceling method
Participants completed the Chinese version of the 21-item was used to create parcels for all variables except for traditional bul-
Depression Anxiety Stress Scale (DASS–21) (Chan et al., 2012). Each of lying victimization, which was indicated by its three subtypes. This
the three subscales contains seven items. Sample items are “I felt that study tested the cross-lagged paths controlling for autoregressive and
life was meaningless”, “I felt scared without any good reason”, and “I within-time paths among all research variables. Participants’ gender,
found it difficult to relax” . Participants rated each item on a 4-point grade, family structure, family income, and parental education were
scale ranging from 0 (did not apply to me at all) to 3 (applied to me very analyzed as covariates. Fig. 1 presents the proposed conceptual model.
much, or most of the time). In this sample, Cronbach's alpha was 0.82, For clarity, the measurement details, cross-sectional relationships, and
0.85, and 0.84 for depression scale, 0.76, 0.79, and 0.81 for anxiety control variables are not shown. Model fit was evaluated through CFI,
scale, and 0.78, 0.82, and 0.83 for stress scale at T1–T3, respectively. TLI, RMSEA, and SRMR. Models fit the data well when CFI and TLI
scores are greater than 0.95, and RMSEA and SRMR scores are lower
2.2.4. Self-esteem than 0.06 and 0.08, respectively (Hu and Bentler, 1999). For model
Participants completed the Chinese version of the Rosenberg Self- comparison, no difference exists in fit indices if ΔCFI and ΔTLI are less
Esteem Scale (Wang et al., 1999). This scale includes 10 items that than 0.01, and ΔRMSEA is less than 0.015 (Chen, 2007).
evaluate individuals’ positive and negative feelings about the self. A
sample item is “On the whole I am satisfied with myself” . Participants 3. Results
rated each item on a 4-point scale ranging from 1 (strongly disagree) to
4 (strongly agree). Cronbach's alpha in this sample was 0.85, 0.86, and 3.1. Preliminary analyses
0.87 at T1–T3, respectively.
Eight separate three-way mixed ANOVAs were used to determine
2.2.5. Social anxiety whether the change in research variables from T1 to T3 varied as a
Adolescents’ social anxiety was assessed by the Chinese version of function of gender and grade. In each ANOVA, time was the repeated
Social Anxiety Subscale in the Self-Consciousness Scale (Wang et al., factor, and gender and grade were two between-subjects factors.
1999). This subscale includes six items describing individuals’ Results revealed a decreasing trend from T1 to T3 in TBV (F(2,

605
 Table 1
Means, standard deviations and bivariate correlations of all study variables.
X.-W. Chu et al.

Variables M (SD) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

1. TBV T1 1.36
(0.56)
2. CBV T1 1.22 .40
(0.35)
3. Depression T1 1.42 .32 .37
(0.52)
4. GA T1 1.53 .33 .35 .69
(0.52)
5. Stress T1 1.73 .30 .36 .72 .77
(0.58)
6. Self-esteem T1 2.98 −0.23 −0.19 −0.41 −0.29 −0.27
(0.57)
7. SA T1 1.13 .12 .06 .28 .37 .31 −0.26
(0.63)
8. Loneliness T1 1.86 .32 .17 .42 .42 .36 −0.48 .36
(0.62)
9. TBV T2 1.29 .49 .28 .17 .18 .16 −0.14 .05 .19
(0.52)
10. CBV T2 1.19 .32 .53 .24 .25 .22 −0.14 .04 .18 .44
(0.33)
11. Depression T2 1.45 .25 .24 .52 .42 .43 −0.35 .23 .43 .28 .29
(0.54)
12. GA T2 1.57 .24 .20 .44 .52 .46 −0.26 .25 .37 .28 .31 .72
(0.55)

606
13. Stress T2 1.74 .25 .22 .46 .48 .54 −0.24 .28 .34 .26 .28 .76 .79
(0.60)
14. Self-esteem T2 3.05 −0.20 −0.17 −0.35 −0.30 −0.27 .66 −0.19 −0.44 −0.21 −0.18 −0.48 −0.37 −0.34
(0.54)
15. SA T2 1.10 .11 .06 .25 .29 .26 −0.25 .59 .38 .12 .12 .35 .39 .42 −0.29
(0.64)
16. Loneliness T2 1.72 .21 .10 .28 .30 .24 −0.37 .28 .58 .29 .16 .45 .40 .39 −0.46 .40
(0.56)
17. TBV T3 1.23 .38 .25 .14 .13 .13 −0.15 .03 .18 .46 .35 .29 .27 .27 −0.12 .11 .23
(0.47)
18. CBV T3 1.17 .24 .44 .19 .20 .20 −0.16 .02 .17 .38 .62 .27 .27 .21 −0.16 .11 .16 .41
(0.33)
19. Depression T3 1.45 .22 .17 .42 .35 .36 −0.28 .18 .30 .18 .20 .53 .48 .46 −0.36 .24 .33 .35 .33
(0.52)
20. GA T3 1.51 .21 .11 .38 .46 .42 −0.27 .26 .34 .19 .23 .47 .62 .53 −0.31 .33 .35 .32 .36 .70
(0.53)
21. Stress T3 1.70 .20 .13 .36 .37 .47 −0.19 .23 .27 .22 .19 .45 .54 .54 −0.27 .33 .34 .31 .32 .72 .79
(0.59)
22. Self-esteem T3 3.01 −0.16 −0.14 −0.28 −0.20 −0.20 .49 −0.20 −0.26 −0.10 −0.13 −0.35 −0.28 −0.26 .58 −0.23 −0.25 −0.24 −0.21 −0.54 −0.40 −0.37
(0.59)
23. SA T3 1.12 .04 .02 .20 .26 .23 −0.25 .55 .30 .04 .08 .28 .32 .31 −0.27 .64 .29 .15 .14 .38 .44 .43 −0.35
(0.69)
24. Loneliness T3 1.76 .21 .12 .24 .30 .22 −0.27 .25 .47 .17 .15 .35 .36 .31 −0.33 .31 .55 .38 .23 .51 .47 .44 −0.48 .42
(0.59)

Note. TBV = Traditional bullying victimization; CBV = Cyberbullying victimization; GA = General anxiety; SA = Social anxiety; T1 = Time 1; T2 = Time 2; T3 = Time 3.
The underscored figures are significant at p < .05, and the bold figures are significant at p < .01 (2-tailed).
Journal of Affective Disorders 246 (2019) 603–610
X.-W. Chu et al. Journal of Affective Disorders 246 (2019) 603–610

594) = 18.43, p < .001), CBV (F(2, 594) = 4.91, p < .01), and loneliness and cross-lagged path coefficients in the final structural models (i.e.,
(F(2, 594) = 17.19, p < .001). Boys were more likely than girls to ex- Model 2). The autoregressive paths were all highly correlated and sig-
perience TBV (F(1, 594) = 8.47, p < .01) and CBV (F(1, 594) = 4.25, nificant for TBV, CBV, depression, general anxiety, stress, self-esteem,
p < .05). Students in grade 7 had greater symptoms of depression (F(1, social anxiety, and loneliness between T1 and T2 as well as between T2
594) = 4.13, p < .05), general anxiety (F(1, 594) = 3.87, p = .05), and and T3. As for the significant cross-lagged paths in the six structural
stress (F(1, 594) = 4.43, p < .05) than those in grade 8. The two-way and models, depression at T2 positively predicted TBV and CBV at T3.
three-way interaction effects among time, gender, and grade were not General anxiety at T2 positively predicted TBV and CBV at T3. Stress at
revealed in any variable. Table 1 presents the means, standard devia- T2 positively predicted TBV at T3. Self-esteem at T2 negatively pre-
tions, and bivariate correlations of all study variables at three waves. dicted CBV at T3. Social anxiety at T2 positively predicted CBV at T3.
Loneliness at T1 and T2 positively predicted CBV at T2 and T3. The
3.2. Measurement models results indicate that psychosocial problems are predictors, rather than
outcomes, of the two types of bullying victimization.
Before analyzing structural models, we examined whether the A series of tests on the differences in path coefficients was further
measurement model was equivalent over time. A freely estimated used to examine whether psychosocial problems were more predictive
measurement model (i.e., no specified parameter restrictions) was first of cyber than of traditional bullying victimization. The Wald test of
established for all latent variables (TBV, CBV, depression, general an- parameter constraints showed no differences between the path coeffi-
xiety, stress, self-esteem, social anxiety, and loneliness). It represented a cients from psychosocial problems to TBV and the path coefficients
close fit of the data (S-Bχ2(2160) = 3,539.68, p < .001, CFI = 0.95, from psychosocial problems to CBV (ps > 0.05). These results suggest
TLI = 0.94, RMSEA = 0.03, SRMR = 0.04). The standardized factor that psychosocial problems do not differentially predict the two types of
loadings were significant and varied between 0.59 and 0.92 (M = 0.78, bullying victimization. Besides, this study analyzed the six structural
SD = 0.07). Then, a restrictive model (i.e., factor loadings were speci- models above separately for girls and boys. As displayed in Table 3, the
fied as equal across time) was compared to the unrestrictive model, and paths from depression, general anxiety, and stress at T2 to TBV at T3
it did not become statistically worse (ΔCFI = 0.000, ΔTLI = 0.001, were significant for girls. The paths from depression and general an-
ΔRMSEA = 0.000). Finally, a more restrictive model (i.e., both factor xiety at T2 to TBV at T3, and the paths from depression, general an-
loadings and item intercepts were specified as equal across time) was xiety, social anxiety, and loneliness at T2 to CBV at T3 were significant
compared to the restrictive model, and it also did not become worse for boys. The Wald test only revealed a significant gender difference in
(ΔCFI = 0.002, ΔTLI = 0.002, ΔRMSEA = 0.000). These results in- the path from loneliness at T2 to CBV at T3 (W(1) = 5.93, p < .05). The
dicate that the measurement of latent variables with the indicators is predictive effect of loneliness on CBV was stronger in boys than in girls.
appropriate and invariant across time.
4. Discussion
3.3. Structural models
Bullying victimization is one of the most common interpersonal
Based on the conceptual model (see Fig. 1), six separate structural stressors during adolescence. Both traditional and cyber forms of bul-
models (i.e., bullying victimization and six psychosocial problems) lying victimization are associated with adolescents’ mental health
were performed to test the hypotheses. As shown in Table 2, the fit problems. However, little is known about the direction of these rela-
indexes were satisfactory for the unconstrained models including all tions, particularly under Chinese cultural background. This study fills in
cross-lagged paths (i.e., Model l). Further, the models were re-estimated the gap by analyzing the temporal and reciprocal relationships between
after deleting the nonsignificant cross-lagged paths (i.e., Model 2). They bullying victimization and psychosocial problems (i.e., depression,
fit the data well and were not statistically worse than the originally general anxiety, stress, low self-esteem, social anxiety, and loneliness).
unconstrained models. Table 3 displays all significant autoregressive Results indicate that psychosocial problems prospectively predict

Table 2
Model fit and comparison of all alternative structural models in this study.
Models Model fit Model comparison

S-Bχ 2
df CFI TLI RMSEA [90% CI] SRMR ΔCFI ΔTLI ΔRMSEA
BV & Depression
Model 1 540.28⁎⁎ 373 .974 .966 .024 [.020, 0.029] .042
Model 2 538.07⁎⁎ 379 .975 .968 .024 [.019, 0.028] .044 .001 .002 .000
BV & GA
⁎⁎
Model 1 560.92 373 .970 .961 .026 [.021, 0.030] .041
Model 2 562.17⁎⁎ 379 .971 .963 .025 [.021, 0.030] .041 .001 .002 .001
BV & Stress
Model 1 611.53⁎⁎ 373 .964 .953 .029 [.025, 0.033] .045
Model 2 615.43⁎⁎ 380 .965 .955 .029 [.024, 0.033] .047 .001 .002 .000
BV & Self-esteem
Model 1 562.40⁎⁎ 373 .976 .968 .026 [.021, 0.030] .044
Model 2 567.78⁎⁎ 380 .976 .969 .026 [.021, 0.030] .047 .000 .001 .000
BV & SA
⁎⁎
Model 1 613.55 373 .963 .952 .029 [.025, 0.033] .048
Model 2 617.82⁎⁎ 380 .964 .954 .029 [.025, 0.033] .048 .001 .002 .000
BV & Loneliness
Model 1 593.28⁎⁎ 373 .969 .959 .028 [.024, 0.032] .045
Model 2 598.49⁎⁎ 379 .969 .960 .028 [.023, 0.032] .046 .000 .001 .000

2
Note. BV = Bullying victimization; GA = General anxiety; SA = Social anxiety; S-Bχ = Satorra-Bentler scaled chi-square; df = degrees of freedom;
CFI = comparative fit index; TLI = Tucker-Lewis index; RMSEA = root mean square error of approximation; 90% CI = 90% confidence interval;
SRMR = standardized root mean squared residual; Δ = change in parameter. Model 1 refers to the unconstrained models including all cross-lagged paths, and Model
2 refers to the models after deleting non-significant cross-lagged paths.
⁎⁎
p < .01

607
X.-W. Chu et al. Journal of Affective Disorders 246 (2019) 603–610

Table 3
Unstandardized and standardized parameter estimates of the final structural models.
Models Paths Overall models Models for girls Models for boys

B (SE) β B (SE) β B (SE) β

BV & Depression Autoregressive paths
⁎⁎ ⁎⁎
TBV T1→TBV T2 .45 (0.08) .55 .27 (0.12) .48 .54 (0.11) .58⁎⁎
TBV T2→TBV T3 .45 (0.12) .47⁎⁎ .62 (0.14) .67⁎⁎ .40 (0.14) .42⁎⁎
CBV T1→CBV T2 .54 (0.08) .61⁎⁎ .56 (0.09) .70⁎⁎ .53 (0.11) .57⁎⁎
CBV T2→CBV T3 .66 (0.08) .63⁎⁎ .77 (0.16) .69⁎⁎ .63 (0.09) .61⁎⁎
Depression T1→Depression T2 .64 (0.07) .60⁎⁎ .59 (0.07) .67⁎⁎ .69 (0.12) .57⁎⁎
Depression T2→Depression T3 .55 (0.06) .58⁎⁎ .70 (0.12) .66⁎⁎ .47 (0.07) .53⁎⁎
Cross-lagged paths
Depression T2→TBV T3 .14 (0.06) .17* .12 (0.06) .18 ± .14 (0.08) .16#
Depression T2→CBV T3 .08 (0.03) .11* .07 (0.04) .10+
BV & GA Autoregressive paths
TBV T1→TBV T2 .45 (0.08) .56⁎⁎ .27 (0.12) .48⁎⁎ .56 (0.11) .59⁎⁎
TBV T2→TBV T3 .45 (0.13) .47⁎⁎ .65 (0.14) .70⁎⁎ .39 (0.15) .41⁎⁎
CBV T1→CBV T2 .55 (0.08) .61⁎⁎ .56 (0.09) .70⁎⁎ .54 (0.11) .58⁎⁎
CBV T2→CBV T3 .66 (0.08) .63⁎⁎ .77 (0.15) .69⁎⁎ .59 (0.09) .58⁎⁎
GA T1→GA T2 .64 (0.07) .63⁎⁎ .62 (0.10) .63⁎⁎ .64 (0.10) .62⁎⁎
GA T2→GA T3 .73 (0.06) .74⁎⁎ .80 (0.09) .78⁎⁎ .71 (0.08) .72⁎⁎
Cross-lagged paths
GA T2→TBV T3 .15 (0.05) .18⁎⁎ .09 (0.05) .14+ .16 (0.08) .18*
GA T2→CBV T3 .08 (0.03) .12* .11 (0.04) .15*
BV & Stress Autoregressive paths
TBV T1→TBV T2 .45 (0.08) .55⁎⁎ .27 (0.12) .48⁎⁎ .55 (0.11) .59⁎⁎
TBV T2→TBV T3 .45 (0.13) .47⁎⁎ .63 (0.14) .68⁎⁎ .45 (0.15) .47⁎⁎
CBV T1→CBV T2 .54 (0.08) .61⁎⁎ .55 (0.09) .70⁎⁎ .54 (0.11) .57⁎⁎
CBV T2→CBV T3 .70 (0.08) .66⁎⁎ .76 (0.16) .69⁎⁎ .66 (0.10) .64⁎⁎
Stress T1→Stress T2 .70 (0.06) .65⁎⁎ .63 (0.07) .73⁎⁎ .74 (0.09) .61⁎⁎
Stress T2→Stress T3 .59 (0.05) .62⁎⁎ .75 (0.10) .71⁎⁎ .52 (0.06) .57⁎⁎
Cross-lagged paths
Stress T2→TBV T3 .11 (0.05) .16* .09 (0.05) .15 ±
BV & Self-esteem Autoregressive paths
TBV T1→TBV T2 .46 (0.08) .56⁎⁎ .26 (0.12) .48⁎⁎ .56 (0.11) .59⁎⁎
TBV T2→TBV T3 .53 (0.12) .54⁎⁎ .68 (0.15) .72⁎⁎ .47 (0.15) .49⁎⁎
CBV T1→CBV T2 .54 (0.08) .61⁎⁎ .55 (0.09) .70⁎⁎ .53 (0.11) .57⁎⁎
CBV T2→CBV T3 .69 (0.08) .65⁎⁎ .75 (0.16) .68⁎⁎ .67 (0.10) .64⁎⁎
Self-esteem T1→Self-esteem T2 .61 (0.04) .72⁎⁎ .57 (0.06) .69⁎⁎ .64 (0.05) .75⁎⁎
Self-esteem T2→Self-esteem T3 .74 (0.05) .65⁎⁎ .86 (0.06) .78⁎⁎ .67 (0.07) .57⁎⁎
Cross-lagged paths
Self-esteem T2→CBV T3 −0.04 (0.02) −0.06*
BV & SA Autoregressive paths
TBV T1→TBV T2 .46 (0.09) .56⁎⁎ .27 (0.12) .48⁎⁎ .56 (0.11) .60⁎⁎
TBV T2→TBV T3 .53 (0.13) .54⁎⁎ .66 (0.15) .72⁎⁎ .48 (0.15) .50⁎⁎
CBV T1→CBV T2 .54 (0.08) .61⁎⁎ .55 (0.09) .70⁎⁎ .53 (0.11) .57⁎⁎
CBV T2→CBV T3 .69 (0.08) .65⁎⁎ .74 (0.15) .67⁎⁎ .64 (0.10) .62⁎⁎
SA T1→SA T2 .75 (0.06) .73⁎⁎ .71 (0.07) .78⁎⁎ .79 (0.10) .70⁎⁎
SA T2→SA T3 .79 (0.06) .73⁎⁎ .80 (0.09) .78⁎⁎ .80 (0.09) .70⁎⁎
Cross-lagged paths
SA T2→CBV T3 .05 (0.03) .07 ± .08 (0.04) .11*
BV & Loneliness Autoregressive paths
⁎⁎ ⁎⁎
TBV T1→TBV T2 .47 (0.08) .57 .25 (0.12) .47 .57 (0.11) .59⁎⁎
TBV T2→TBV T3 .54 (0.13) .55⁎⁎ .65 (0.15) .71⁎⁎ .51 (0.15) .51⁎⁎
CBV T1→CBV T2 .52 (0.08) .58⁎⁎ .55 (0.09) .70⁎⁎ .53 (0.11) .57⁎⁎
CBV T2→CBV T3 .68 (0.08) .65⁎⁎ .74 (0.15) .67⁎⁎ .63 (0.09) .61⁎⁎
Loneliness T1→Loneliness T2 .65 (0.05) .66⁎⁎ .62 (0.07) .64⁎⁎ .67 (0.06) .67⁎⁎
Loneliness T2→Loneliness T3 .65 (0.06) .61⁎⁎ .62 (0.07) .61⁎⁎ .67 (0.08) .60⁎⁎
Cross-lagged paths
Loneliness T1→CBV T2 .06 (0.03) .10*
Loneliness T2→CBV T3 .04 (0.02) .06# .09 (0.03) .14⁎⁎

Note. BV = Bullying victimization; TBV = Traditional bullying victimization; CBV = Cyberbullying victimization; GA = General anxiety; SA = Social anxiety;
T1 = Time 1; T2 = Time 2; T3 = Time 3.
⁎
p < .05;
⁎⁎
p < .01;
+
p = .05;
±
p = .07;
#
p = .08

bullying victimization, but the reverse is not supported. Specifically, not differentially predict the two types of bullying victimization. The
depression and general anxiety are two common predictors of tradi- predictive effect of loneliness at T2 on cyberbullying victimization at T3
tional bullying and cyberbullying victimization. Stress is a specific is stronger in boys than in girls. These findings can shed light on the
predictor of traditional bullying victimization, whereas self-esteem, temporal relationships between bullying victimization and adolescents’
social anxiety, and loneliness are specific predictors of cyberbullying mental health problems. It is also beneficial for education professionals
victimization. Further analyses suggest that psychosocial problems do to formulate universal and targeted interventions aimed at reducing

608
X.-W. Chu et al. Journal of Affective Disorders 246 (2019) 603–610

students’ bullying experiences. However, Chinese boys are often required to be strong and brave. They
Bullying victimization is not a risk factor, but a detrimental outcome may be considered weak and incompetent if they expose their problems
of adolescents’ psychosocial problems. The findings support stress in peer relations. Thus they are more likely to develop new friendships
generation model rather than stress exposure model. According to the with strangers online and are more easily to be bullied.
stress generation model (Hammen, 2006), individuals can actively The present study is not without limitation. First, this study used a
shape and construct their social environments. Adolescents with more self-report questionnaire to gather the data. Self-reports may be subject
psychological health problems tend to select and create negative social to increased biases (e.g., social desirability). Some measures were
environments. Given their personal characteristics (e.g., maladaptive adopted to handle this potential problem, including telling all partici-
coping styles) and the negative circumstances they create (e.g., nega- pants that there was no right or wrong answer to the items, and that we
tive peer relations), these individuals have a tendency to experience would not tell their answers to anyone. To provide further evidence for
increased interpersonal stressors (e.g., bullying victimization). The the findings, we recommend that future research considers the reports
findings are consistent with previous research revealing psychosocial from multiple informants (e.g., parents, peers, and teachers). Second,
symptoms like depression, social anxiety, and loneliness as the pre- this study did not distinguish the three subtypes of traditional bullying
dictors rather than outcomes of traditional bullying and cyberbullying victimization (i.e., physical, verbal, and relational). These three sub-
victimization (Frison et al., 2016; Pabian and Vandebosch, 2016; Tran types and cyberbullying victimization may show different associations
et al., 2012; van den Eijnden et al., 2014). However, the present study with adolescents’ psychosocial problems. Future research can examine
largely extends these research by concurrently investigating traditional this possibility. Finally, participants were recruited from an ordinary
and cyber forms of bullying victimization, with a focus on up to six junior high school in Wuhan, a representative city in central China. This
frequent psychosocial problems in Chinese adolescents. convenience sample, on average, can represent the whole Chinese early
Depression and general anxiety are two common predictors for adolescents. That is, the present findings may be generalized to other
traditional bullying and cyberbullying victimization. This may be at- Chinese students. However, due to the cultural differences (collectivism
tributed to the overlap in definitions of traditional bullying and cy- vs. individualism), our findings should be generalized with caution to
berbullying victimization, including intentional aggression, repetition, the students from other countries. Further research can recruit the
and power imbalance (Kowalski et al., 2014). Another possibility is due samples from more countries to test the findings in this study.
to the substantial overlap between offline and online bullying victimi- This study contributes to a better understanding of the bidirectional
zation trajectories (Sumter et al., 2012). Adolescents who experience relationships between two types of bullying victimization and psycho-
one type of bullying victimization may have an increased chance of logical health problems. The findings suggest that education profes-
experiencing other types of bullying victimization, because bullying sionals should include strategies to prevent psychosocial problems
victimization changes their social-cognitive processing and generates within the interventions in bullying behaviors during adolescence.
negative cognitive biases, which make them acting awkwardly in peer Specifically, prevention programs of bullying should be enriched by
interactions (Sumter et al., 2012). The findings that social anxiety and promoting adolescents’ positive appraisals regarding their intrapersonal
loneliness are specific predictors of cyberbullying victimization may state and peer relationships. The psychological health education course
imply that cyberbullying victimization is the particular result of ado- needs to foster students’ self-evaluation, emotion management, and
lescents’ negative appraisals regarding their functioning in peer and social skills. Adults can provide increased social support for those with
friend contexts. Adolescents who feel socially anxious and lonely are psychosocial problems since they may become the potential victims of
more likely to socialize with online strangers as a kind of compensation, bullying. Interventions on traditional bullying and cyberbullying should
which may increase their risk of being bullied online (Valkenburg and be equally integrated together. Bullying professionals may need to pay
Peter, 2007). special attention to boys in the Chinese context.
Inconsistent with hypotheses, psychosocial problems are not more
predictive of cyberbullying than of traditional bullying victimization. Conflict of interest
Adolescents with psychosocial problems may tend to communicate with
strangers online and experience increased cyberbullying victimization. The authors have declared that they have no competing or potential
However, in the face-to-face context, psychosocial maladjustments are conflicts of interest.
more noticeable, and have more direct and influential effects on stu-
dents’ aggressive behaviors (van den Eijnden et al., 2014). In this sense, Authors' Contributions
there may be similar possibility for adolescents with psychosocial
problems to be the targets of traditional bullying and cyberbullying. Xiao-Wei Chu conceived of the study, participated in its design and
The findings are inconsistent with one previous research revealing that coordination, performed the statistical analysis, interpreted the data,
loneliness and social anxiety are more predictive of online than of real- and drafted the manuscript; Cui-Ying Fan authored the funding appli-
life bullying victimization (van den Eijnden et al., 2014). One possibi- cation, participated in the design of the study, and collected the data;
lity may be due to cultural differences. In China, collective interests and Shuai-Lei Lian participated in the coordination of the study, performed
group harmony are highly emphasized. When students encounter psy- the measurement, and helped revising the manuscript for important
chosocial problems, they are more likely to become unpopular and be intellectual content; Zong-Kui Zhou conceived of the study, participated
excluded from the peer group. Hence, they may more sensitively and in the design of the study, and helped revising the manuscript for im-
easily report their bullying victimization experiences. portant intellectual content. All authors read and approved the final
Although the predictive effects of psychosocial problems on bullying manuscript.
victimization seem to be different across gender, further analyses do not
demonstrate the effects being stronger in girls than in boys. On the Funding sources
contrary, boys with higher levels of loneliness are more likely to ex-
perience cyberbullying victimization compared with girls. One possible This work was supported by the Major Program of National Social
explanation is that Chinese girls are usually treated as a vulnerable Science Foundation of China [grant number 11&ZD151], the Talent
group, and they always receive special care and protection from adults. Project of the National Cultural Celebrities and “Four Batches”, the
Therefore, when they encounter problems in peer relationships (e.g., Fundamental Research Funds of Central China Normal University
feeling lonely), they can gain parents’ or teachers’ extra assistance and [grant number CCNU14Z02004], the Major Cultivation Program of the
emotional support. This may reduce their likelihood of using the Collaborative Innovation Center of Assessment toward Basic Education
Internet to compensate for a lack of satisfying relationships in real life. Quality at Beijing Normal University [grant numbers 2019-04-009-

609
X.-W. Chu et al. Journal of Affective Disorders 246 (2019) 603–610

BZPK01, 2019-04-003-BZPK01], and the Excellent Doctoral 128–140.

Dissertation Cultivation Program of Central China Normal University Hays, R.D., DiMatteo, M.R., 1987. A short-form measure of loneliness. J. Pers. Assess. 51
(1), 69–81.
[grant number 2018YBZZ090]. Hu, L.t., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure ana-
lysis: conventional criteria versus new alternatives. Struct. Equat. Model. 6 (1), 1–55.
Acknowledgements Huang, H., Hong, J.S., Espelage, D.L., 2013. Understanding factors associated with bul-
lying and peer victimization in Chinese schools within ecological contexts. J. Child
Fam. Stud. 22 (7), 881–892.
None. Kim, S., Colwell, S.R., Kata, A., Boyle, M.H., Georgiades, K., 2018. Cyberbullying victi-
mization and adolescent mental health: evidence of differential effects by sex and
mental health problem type. J. Youth Adolesc. 47 (3), 661–672.
Supplementary materials 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
Supplementary material associated with this article can be found, in youth. Psychol. Bull. 140 (4), 1073–1137.
Li, H.-l., Zhang, W.-x., Yu, F.-j., 2012. The relationship between victimization and de-
the online version, at doi:10.1016/j.jad.2018.12.103.
pression of adolescents. Psychol. Dev. Educ. 1, 77–82.
Ma, C.M.S., 2018. A latent profile analysis of internet use and its association with psy-
References chological well-being outcomes among Hong Kong Chinese early adolescents. Appl.
Res. Qual. Life 13, 727–743.
Modecki, K.L., Minchin, J., Harbaugh, A.G., Guerra, N.G., Runions, K.C., 2014. Bullying
Bannink, R., Broeren, S., van de Looij-Jansen, P.M., de Waart, F.G., Raat, H., 2014. Cyber prevalence across contexts: a meta-analysis measuring cyber and traditional bullying.
and traditional bullying victimization as a risk factor for mental health problems and J. Adolesc. Health 55 (5), 602–611.
suicidal ideation in adolescents. PLoS One 9 (4), e94026. Nolen-Hoeksema, S., Girgus, J.S., 1994. The emergence of gender differences in depres-
Campbell, M., Spears, B., Slee, P., Butler, D., Kift, S., 2012. Victims’ perceptions of tra- sion during adolescence. Psychol. Bull. 115 (3), 424–443.
ditional and cyberbullying, and the psychosocial correlates of their victimisation. Nolen-Hoeksema, S., Morrow, J., 1991. A prospective study of depression and posttrau-
Emot. Behav. Diffic. 17 (3–4), 389–401. matic stress symptoms after a natural disaster: the 1989 Loma Prieta earthquake. J.
Chan, H.C., Wong, D.S.W., 2015. Traditional school bullying and cyberbullying in Pers. Soc. Psychol. 61 (1), 115–121.
Chinese societies: prevalence and a review of the whole-school intervention ap- Pabian, S., Vandebosch, H., 2016. An investigation of short-term longitudinal associations
proach. Aggress. Violent Behav. 23, 98–108. between social anxiety and victimization and perpetration of traditional bullying and
Chan, R.C.K., Xu, T., Huang, J., Wang, Y., Zhao, Q., Shum, D.H.K., . . ., Potangaroa, R., cyberbullying. J. Youth Adolesc. 45 (2), 328–339.
2012. Extending the utility of the Depression Anxiety Stress Scale by examining its Reijntjes, A., Kamphuis, J.H., Prinzie, P., Telch, M.J., 2010. Peer victimization and in-
psychometric properties in Chinese settings. Psychiatry Res. 200 (2), 879–883. ternalizing problems in children: a meta-analysis of longitudinal studies. Child Abuse
Chen, F.F., 2007. Sensitivity of goodness of fit indexes to lack of measurement invariance. Negl. 34 (4), 244–252.
Struct. Equat. Model. 14 (3), 464–504. Rose, C.A., Tynes, B.M., 2015. Longitudinal associations between cybervictimization and
Chu, X.-w., Fan, C.-y., 2017. Revision of the revised cyber bullying inventory among mental health among U.S. adolescents. J. Adolesc. Health 57 (3), 305–312.
junior high school students. Chin. J. Clin. Psychol. 25 (6), 1031–1034. Siegel, R.S., La Greca, A.M., Harrison, H.M., 2009. Peer victimization and social anxiety in
Chu, X.-W., Fan, C.-Y., Liu, Q.-Q., Zhou, Z.-K., 2018a. Stability and change of bullying adolescents: prospective and reciprocal relationships. J. Youth Adolesc. 38 (8),
roles in the traditional and virtual contexts: a three-wave longitudinal study in 1096–1109.
Chinese early adolescents. J. Youth Adolesc. 47 (11), 2384–2400. Slovak, K., Singer, J.B., 2011. School social workers’ perceptions of cyberbullying. Child.
Chu, X.-W., Fan, C.-Y., Liu, Q.-Q., Zhou, Z.-K., 2018b. Cyberbullying victimization and Sch. 33 (1), 5–16.
symptoms of depression and anxiety among Chinese adolescents: examining hope- Stapinski, L.A., Araya, R., Heron, J., Montgomery, A.A., Stallard, P., 2015. Peer victi-
lessness as a mediator and self-compassion as a moderator. Comput. Hum. Behav. 86, mization during adolescence: concurrent and prospective impact on symptoms of
377–386. depression and anxiety. Anxiety, Stress Coping 28 (1), 105–120.
Fahy, A.E., Stansfeld, S.A., Smuk, M., Smith, N.R., Cummins, S., Clark, C., 2016. Sumter, S.R., Baumgartner, S.E., Valkenburg, P.M., Peter, J., 2012. Developmental tra-
Longitudinal associations between cyberbullying involvement and adolescent mental jectories of peer victimization: off-line and online experiences during adolescence. J.
health. J. Adolesc. Health 59, 502–509. Adolesc. Health 50 (6), 607–613.
Fisher, B.W., Gardella, J.H., Teurbe-Tolon, A.R., 2016. Peer cybervictimization among Tran, C.V., Cole, D.A., Weiss, B., 2012. Testing reciprocal longitudinal relations between
adolescents and the associated internalizing and externalizing problems: a meta- peer victimization and depressive symptoms in young adolescents. J. Clin. Child
analysis. J. Youth Adolesc. 45 (9), 1727–1743. Adolesc. Psychol. 41 (3), 353–360.
Frison, E., Subrahmanyam, K., Eggermont, S., 2016. The short-term longitudinal and Valkenburg, P.M., Peter, J., 2007. Internet communication and its relation to well-being:
reciprocal relations between peer victimization on Facebook and adolescents’ well- identifying some underlying mechanisms. Media Psychol. 9 (1), 43–58.
being. J. Youth Adolesc. 45 (9), 1755–1771. van den Eijnden, R., Vermulst, A., van Rooij, A.J., Scholte, R., van de Mheen, D., 2014.
Gámez-Guadix, M., Orue, I., Smith, P.K., Calvete, E., 2013. Longitudinal and reciprocal The bidirectional relationships between online victimization and psychosocial pro-
relations of cyberbullying with depression, substance use, and problematic internet blems in adolescents: a comparison with real-life victimization. J. Youth Adolesc. 43
use among adolescents. J. Adolesc. Health 53 (4), 446–452. (5), 790–802.
Gibb, B.E., Hanley, A.J., 2010. Depression and interpersonal stress generation in children: Wang, X.-D., Wang, X.-L., Ma, H., 1999. Handbook of rating scales for mental health
prospective impact on relational versus overt victimization. Int. J. Cogn. Ther. 3 (4), (Revised Edition). Chinese Mental Health Journal Press, Beijing.
358–367. Wu, C.-h., Yao, G., 2008. Psychometric analysis of the short-form UCLA Loneliness Scale
Goodman, J.S., Blum, T.C., 1996. Assessing the non-random sampling effects of subject (ULS-8) in Taiwanese undergraduate students. Personal. Individ. Differ. 44 (8),
attrition in longitudinal research. J. Manag. 22 (4), 627–652. 1762–1771.
Hammen, C., 2006. Stress generation in depression: reflections on origins, research, and Yang, S.-J., Stewart, R., Kim, J.-M., Kim, S.-W., Shin, I.-S., Dewey, M.E., . . ., Yoon, J.-S.,
future directions. J. Clin. Psychol. 62 (9), 1065–1082. 2013. Differences in predictors of traditional and cyber-bullying: a 2-year long-
Hankin, B.L., Abramson, L.Y., Moffitt, T.E., Silva, P.A., McGee, R., Angell, K.E., 1998. itudinal study in Korean school children. Eur. Child Adolesc. Psychiatry 22 (5),
Development of depression from preadolescence to young adulthood: emerging 309–318.
gender differences in a 10-year longitudinal study. J. Abnorm. Psychol. 107 (1),

610