Temporal patterns and predictors of bullying roles among adolescents in Vietnam: a school-based cohort study

Abstract

Although many cross-sectional studies have examined bullying experiences and correlated factors among adolescents in schools, relatively little is known about the extent to which bullying roles are stable or fluid over time. This short-term quantitative longitudinal study in Vietnam examined temporal patterns and predictors of bullying roles over an academic year. A total of 1424 middle and high school students aged 12–17 years completed two anonymous, self-administered questionnaires six months apart in 2014 and 2015. Young people were classified into different bullying roles as follow: not-involved (38.9%), victims only (24%), bullies only (6.6%), and bully-victims (40.4%) across the two times. About 60% of all surveyed students experienced bullying either as victim, bully, or bully-victim during the year. Of these students, nearly three in four indicated unstable bullying roles over time. Multivariate multinomial logistic regressions indicated factors ranging from individual (age, gender, and mental health) to family (social support, parental supervision and monitoring, witnessing parental violence, and conflict with siblings), school (perceived social support, teachers’ attempt to stop bullying at school), and peers (social support, students’ attempt to stop bullying at school) have significant associations with levels of bullying involvement. Implications for bullying prevention programs nationally and internationally are discussed.

Keywords:

• Bullying
• cyberbullying
• adolescent
• longitudinal design
• mental health
• Vietnam

Introduction

Bullying is intentional and repeated aggression via physical, verbal, relational or cyber forms in which the targets cannot defend themselves (Olweus, 1994; Smith, Mahdavi, Carvalho, Fisher, Russell, & Tippett, 2008). This type of interpersonal aggression has been studied for over thirty years in Western countries (Smith, 2016) and more recently in North Asian countries, such as Japan (Toda, 2016; Yoneyama & Naito, 2003), Korea (Yang et al., 2006, 2013) and China (Chan & Wong, 2015).

The Southeast Asian region has a population of more than 600 million people, but to date there has been relatively limited standardised survey research into bullying, and no relevant cohort studies in schools in this region (Sittichai & Smith, 2015). In Vietnam, some research in schools has addressed various forms of violence among young people (Nguyen, 2012; Nguyen & Tran, 2013) and child maltreatment experiences (Le, Holton, Nguyen, Wolfe, & Fisher, 2015; Nguyen, Dunne, & Le, 2010). The studies that focus specifically on school bullying in Vietnam have used in-depth qualitative methods to explore the characteristics and contexts of bullying (Horton, 2011; Horton, Lindholm, & Nguyen, 2015; UNESCO, 2015).

Globally, scholars have identified four main bullying roles that to classify children’s involvement as bully, victim, bully-victim (Gumpel, 2008; Gumpel, Zioni‐Koren, & Bekerman, 2014; Salmivalli, Lagerspetz, Björkqvist, Österman, & Kaukiainen, 1996; Thornberg, 2007). International estimates suggest that from 10% to 30% of children and adolescents are victims only (Gumpel, 2008; Gumpel et al., 2014; Lereya et al., 2015; Solberg, Olweus, & Endresen, 2007; Vlachou, Andreou, Botsoglou, & Didaskalou, 2011). The proportions of those who bully others range from 3 to 15 % of adolescents (Gumpel et al., 2014; Lereya et al., 2015; Solberg et al., 2007). Further, bully-victims or ‘provocative victims’ (Gumpel, 2008) who are both victimised and bully others comprise between 15 and 20% of populations of young people (Lereya, Copeland, Zammit, & Wolke, 2015; Salmivalli, 2010).

Bullying is a very complex phenomenon that is related to multiple causal factors ranging from individual characteristics to school, family and social environments (Swearer Napolitano & Espelage, 2011). At the individual level, numerous factors have been linked with bullying roles, such as younger age (Chan & Wong, 2015; Hong & Espelage, 2012) and male gender. Risk of exposure may vary in different geographic regions. Boys are more likely to be engaged in bullying both as victims or perpetrators than are girls in some Asian countries (Chan & Wong, 2015; Yang et al., 2013); while females are more likely than males to report being bullied in some Western countries (Rigby & Johnson, 2016). Researchers have found significant associations between social media use and online or cyber victimisation or cyber perpetration (Mishna, Khoury-Kassabri, Gadalla, & Daciuk, 2012; Zhou et al., 2013). Notably, previous studies have reported the relationships between mental health problems and various types of victimisation in both online and offline spaces (Lereya et al., 2015; Yang et al., 2013).

Features of parent–child relationships can be either protective or risk factors for engagement in bullying. For instance, parental monitoring of Internet usage may prevent adolescents from bullying others online (Zhou et al., 2013). High parental social support can protect adolescents from being victimised by their peers (Holt & Espelage, 2007) or bullying others online and in traditional ways (Fanti, Demetriou, & Hawa, 2012; Wang, Iannotti, & Nansel, 2009). There are also many converse family influences. Adolescents who witness parental violence may be more likely to be involved in bullying others in school and online (Hemphill et al., 2012). Adolescents who are bullied by siblings are more likely to be victimised by their peers (Tucker, Finkelhor, Turner, & Shattuck, 2014) or to experience bully-victim roles in offline and online settings (Tanrikulu & Campbell, 2014).

School environment appears to strongly affect bullying involvement. Students who lack teacher’s restrictions on their mobile phone usage (Zhou et al., 2013) have higher risk of being victims or perpetrators online. Also, those who experience less teacher social support are more likely to be victims (Furlong & Maynard, 1995). Interestingly, there is a positive correlation between cyberbullying perpetration and students’ perception about teachers’ abilities in stopping bullying at school (Elledge et al., 2013). The majority of students who are bullied do not disclose this to their teachers (Mishna et al., 2012). It may be because students assume that school staff would not help them (Li, 2007).

Young people who have antisocial friends in early adolescence are more likely to engage in both traditional and cyberbullying perpetration (Hemphill et al., 2012). Also, students who lack peer social support are more likely to be cyber victims (Fanti et al., 2012) and traditional victims (Furlong & Maynard, 1995). Peers can be protective as well; victimisation at school may be less likely when peers do not agree with bullying (Denny et al., 2015).

To date, little is known about the extent to which the bullying roles are stable or fluid over time (Gumpel et al., 2014; Ryoo, Wang, & Swearer, 2015). There is some emerging evidence that the majority of youth who report involvement have infrequent experiences (Ryoo et al., 2015). Even for youth who bully others, there seems to be unstable involvement over one or a few years at middle or high school (Lereya et al., 2015). Findings of an ethnographic study conducted with 10th grade adolescents revealed that ‘many roles are fluid’ (Gumpel et al., 2014) in specific situational context. Unfortunately, most evidence regarding factors associated with bullying roles has been derived from cross-sectional studies (Cook, Williams, Guerra, Kim, & Sadek, 2010; Hong & Espelage, 2012). The present study builds upon prior research by Lereya et al. (2015) to examine the extent to which students have stable or unstable in bullying roles over time. We analyse associations between individual characteristics and family, peer and school relationships and the different levels of patterns in bullying roles among adolescents in schools in Vietnam.

Method

Data for this study were derived from two waves of school-based surveys, six months apart at four public middle and high schools in urban areas of Vietnam during the academic year 2014–2015. Details of the study are described elsewhere (Le, Holton, Nguyen, Wolfe, & Fisher, 2016a). Briefly, an invitation to participate in the study was sent to 1668 students, of whom 1539 (92.3%) consented to and completed the questionnaire at the baseline survey (Time 1). Of them, 1460 students were followed up at Time 2. The final sample for analyses included 1424 students, accounting for 92.5% of the initial sample. Of the remainder, 115 students were lost to follow-up or could not be matched at Time 2 due to absence on the survey days, change in school, or inability to visually match questionnaire identification across the two surveys. Male students comprised 45.1% of the sample. Age ranged from 12 to 17 (M = 14.7, SD = 1.87). The majority of students lived with both biological parents (87.6%) and had at least one sibling (89.0%). Most students (90%) said they can access the Internet and spent at least one hour daily online.

Ethical clearance was obtained from the Human Research Ethics Committees of the Queensland University of Technology (No. 1400000713) and Hanoi University of Public Health (No. 279/2014/YTCC-HD3). All respondents in the study provided informed consent prior to survey administration.

Measures

Bullying victimisation and perpetration

Traditional and cyber bullying victimisation and perpetration were measured in relation to six behaviours including hitting/kicking/shoving around, robbing/stealing/damaging properties, threatening/forcing someone to do things they do not want to do, calling mean names/teasing in rude ways, excluding, and spreading rumours. These behaviours were identified from literature review and through 16 in-depth interviews during the pilot phase with and were validated among Vietnamese students (Le et al., 2016a). A definition adapted from Ybarra, Boyd, Korchmaros, and Oppenheim (2012) and Langos (2012) was given to students prior to the completing the questionnaires in order to standardise understanding of bullying. For the victimisation scale, students were prompted with the question ‘How often have you been bullied in any way during the last six months?’ then six items were presented. The bullying perpetration measurement was similar, with prompts to ask the students how often they bullied others. We distinguished traditional bullying from cyberbullying via different modes of communication (in-person or cyber) by which students experienced bullying behaviours. A five-point Likert scale, ranging from 0 = ‘never’, 1 = ‘a few times during the last six months’, 2 = ‘once or twice a month’; 3 = ‘once or twice a week’, to 4 = ‘almost every day’, was used to measure frequency of behaviour, for each mode of communication. The study employed a cut-off point to classify victimisation or perpetration behaviours from ‘1 = a few times’ to more often.

Risk and protective factors

Demographic characteristics of the respondents included age (year of birth), gender (0 = female, 1 = male), family structure (0 = living with both parents, 1 = living with single parent/stepparent/others).

Reaction when seeing bullying events was assessed with an item from the Olweus Bully/Victim Questionnaire (Olweus, 1996). The respondents were asked ‘How do you usually react if you see or understand that a student at your age is being bullied by other students?’ with possible responses: 0 = I have never noticed; 1 = I take part in; 2 = I don’t do anything, but I think the bullying is OK; 3 = I just watch goes on; 4 = I don’t do anything, but I think I ought to help; 5 = I try to help the bullied student in one way or another.

Online activities were measured with 4 items asking respondents about time spent on online activities including communication, social networking and entertainment in the past week. The five-point Likert scale response options were ‘1 = never use’, ‘2 = several times a week’, ‘3 = several times a day’, ‘4 = several times an hour’, ‘5 = all the time’. Summed scores ranged from 4 to 20, with higher score indicating more time spent online.

Parents’ and teachers’ supervision of online activities were assessed by two questions: How often do your (i) ‘parents supervise your online activities?’ and (ii) ‘teachers supervise your online activities?’; using a 5-point Likert scale (1 = none of the time to 5 = all of the time). Parents’ and teachers’ control of Internet and mobile phone usage was measured with these questions: How often do your parents (i) ‘control your access to the Internet?’, (ii) ‘control your use of the mobile phone?’ and how often do your teachers (iii) ‘control your access to the Internet?’, (iv) ‘control your use of the mobile phone?’. Response options were on a 5-point Likert scale ranging from ‘1 = none of the time’ to ‘5 = all of the time’.

Family, friend, and school social support was measured by the Multidimensional Scale of Perceived Social Support (MSPSS) (Zimet, Dahlem, Zimet, & Farley, 1988). The MSPSS comprises 12 items that are equally distributed to measure family support (e.g. ‘My family really tries to help me’), friend support (e.g. ‘My friends really try to help me’), and school support (e.g. ‘There is a school staff member who is around when I am in need’); using a 4-point Likert scale ranging from ‘1 = strongly disagree’ to ‘4 = strongly agree’. These were summed with higher score values indicating higher levels of support. In this study, alpha coefficients for the three subscales respectively were .88 (family support), .91 (friend support), and .90 (school support) at Time 1.

Witnessing parents serious arguing or fighting was assessed by asking students ‘How often have you witnessed your parents having (i) a serious argument with each other? and (ii) physically fighting with each other?; using a 4-point Likert scale ranging from 1 = never to 4 = often. In this study, alpha coefficients were .71 at Time 1.

Conflict with siblings was assessed with the question ‘How often have you had serious conflict (argument, fighting etc.) with your siblings?’ using 4-point Likert scale ranging from 1 = never to 4 = often.

Perceptions of students and teachers trying to stop bullying at school were assessed by asking students ‘How often do (i) teachers/other adults try to stop to it when a student is being bullied at school? and (ii) students at school try to stop to it when a student is being bullied at school?; using a 5-point Likert scale ranging from 1 = almost never to 5 = almost always.

Depressive symptoms were measured with The Centre for Epidemiological Studies-Depression Scale (CESD) (Radloff, 1977). The scale comprises 20 items (e.g. ‘I felt lonely’). Respondents were asked to indicate the frequency with each item that they experienced during the previous week. Response options on the four-point scale were 0 = ‘less than 1 day’, 1 = ‘1–2 days’, 2 = ‘3–4 days’, and 3 = ‘5–7 days’. These were summed with scores ranging from zero to 45, the higher scores indicating the higher level of depressive symptom. In this study, the alpha coefficient for this scale was .86 for Time 1.

Psychological distress was assessed by using The Kessler Psychological Distress Scale (K10) (Kessler et al., 2002) . The scale includes 10 items (e.g. ‘during the last 30 days, about how often did you be tired out for no good reason?’ to measure emotional feelings experienced in the last month with a five-point Likert scale ranging from ‘1 = none of the time’ to ‘5 = all of the time’. A composite score was created so that a higher score indicated higher levels of psychological distress. In this study, the alpha coefficient was .87 at Time 1.

Self-esteem was assessed by The Rosenberg Self-esteem scale (RSES) (Rosenberg, 1965). The scale includes 10 items (e.g. ‘On the whole, I am satisfied with myself’) using a 4-point Likert scale ranging from ‘1 = strongly agree’ to ‘4 = strongly disagree’. Higher score values higher levels of self-esteem. Cronbach’s alpha was .70 at Time 1.

Suicidal ideation was measured with three items adapted from the American School Health Association (Kent, 1989). Respondents were asked ‘During past 6 months, have you ever (i) seriously thought about attempting suicide? (ii) made a specific plan about how you would attempt suicide? and (iii) attempted suicide?’. The responses were categorised as 0 = no, 1 = yes if respondents admitted at least one of these thoughts or behaviours.

Data analyses

Data analyses were performed using Stata (version 11.2). Descriptive analyses explored the prevalence and stability in bullying experiences as well as the characteristics and distribution of other variables. Multinomial logistic regressions were utilised in bivariate and multivariate association analyses. Bivariate analyses examining the associations between each predictor at Time 1 and temporal patterns of bullying roles as victims, bullies, and bully-victims were conducted. All predictors with p value ≤.2 (Bursac, Gauss, Williams, & Hosmer, 2008) were entered in the multivariate models, controlling for other significant variables. Multinomial logistic regressions were utilised in bivariate and multivariate association analyses.

Results

Prevalence of bullying victimisation and perpetration

Table 1 presents prevalence estimates for specific forms of traditional bullying and cyberbullying victimisation and perpetration at Times 1 and 2. Cyberbullying victimisation and perpetration were reported much less often than traditional bullying experiences at both times. A high correlation between traditional and cyberbullying was observed, with very few students experiencing only cyberbullying victimisation or perpetration.

Table 1. Specific forms of traditional and cyberbullying victimisation and perpetration among school adolescents at Times 1 and 2.

	Victimisation N (%)		Perpetration N (%)
Behaviours	Traditional	Cyber	Traditional	Cyber
TIME 1
Hitting	373 (26.2)	NA	247 (17.3)	NA
Robbing	170 (11.9)	NA	43 (3.0)	NA
Threatening	140 (9.8)	78 (5.5)	85 (6.0)	31 (2.2)
Teasing	392 (27.5)	112 (7.9)	202 (14.2)	57 (4.0)
Excluding	97 (6.8)	36 (2.5)	105 (7.4)	35 (2.5)
Spreading rumours	166 (11.7)	83 (5.8)	58 (4.1)	33 (2.3)
Specific form	620 (43.5)	170 (11.9)	400 (28.1)	87 (19.8)
Any or both forms	637 (44.7)	412 (28.9)
TIME 2
Hitting	290 (20.4)	NA	185 (13.0)	NA
Robbing	127 (8.9)	NA	39 (2.7)	NA
Threatening	105 (7.4)	57 (4.0)	64 (4.5)	37 (2.6)
Teasing	308 (21.6)	86 (6.0)	149 (10.5)	41 (2.9)
Excluding	75 (5.3)	24 (1.7)	80 (5.6)	28 (2.0)
Spreading rumours	140 (9.8)	69 (4.8)	53 (3.7)	28 (2.0)
Specific form	461 (32.4)	134 (9.4)	282 (19.8)	66 (4.6)
Any or both forms	472 (33.1)	288 (20.2)

Note: NA = not applicable.

Temporal pattern in bullying roles

The responses from students were divided into four bullying roles at Times 1 and 2. The baseline survey yielded prevalence estimates for each role: not involved at all (48.3%); victims only (22.7%); bullies only (6.9%); and bully-victims (22%). At follow-up six months later, the estimates were: not involved (62.1%); victims only (17.6%); bullies only (4.7%); and bully-victims (15.5%). Subsequently, we measured temporal patterns of each bullying role by following Turner and his colleague’s classification (Turner, Shattuck, Finkelhor, & Hamby, 2015) to categorise the respondents into four levels of bullying involvement: stable-low, declining, increasing, and stable-high. Further details of generating the patterns of bullying roles over Times 1 and 2 have been provided elsewhere (Le et al., 2016b).

The levels of stability in bullying roles across two times were: (i) not involved in any bullying at either time, accounting for 38.9% of the sample, (ii) victims only, accounting for 24% (of these students, 58.2% were stable-low, 17.0% declining at time 2, 14.3% increasing at time 2, 10.5% were stable-high), (iii) bullies only, accounting for 6.6% (of them, 23.4% were stable-low, 36.2% were declining, 40.4% were increasing which included just 4 students who were involved at stable-high level), and (iv) bully-victims, accounting for 40.4% (of these students, 52.8% were stable-low, 19.3% declining, 14.5% increasing, and 13.4% were stable-high).

Determinants of bullying roles

Associations between individual, family, peer and school related factors at Time 1 and levels of stability in bullying roles over Times 1 and 2 were examined. Table 2 presents the results of multivariate multinomial logistic regression of predictors of temporal patterns of victim role. Older students significantly decreased the odds of being in the stable-low, declining, or increasing victimisation groups, compared to the not involved group. Those students who reported higher level of psychological distress, low parental supervision of online activities, and those who perceived that other students infrequently try to stop bullying at school increased the odds of being in the stable-low victimisation group. Students who reported higher psychological distress and perceived that students infrequently try to stop bullying at school were more likely to be in the declining group (e.g. involved high at Time 1 but low at Time 2). These factors were not associated with the likelihood of students being in the group with increasing victimisation, with the exception of younger students. Male students who reported higher psychological distress had increased odds of being frequent victims over time (e.g. stable high victimisation).

Table 2. Multivariate multinomial logistic regression of predictors for intensity of stability in victimisation across Times 1 and 2.

	Intensity of stability in Victimisation (the ref. group: ‘Not-involved’)
Predictors measured at Time 1	Stable low OR (95% CI)	Declining OR (95% CI)	Increasing OR (95% CI)	Stable high OR (95% CI)
Gender
Female	1.0	1.0	1.0	1.0
Male	1.3 (.9–1.9)	1.8 (.9–3.7)	.6 (.3–1.4)	3.5 (1.4–8.4)^**
Age (years)	.8 (.7–.9)^***	.7 (.6–.9)^**	.8 (.6–.9)^**	.8 (.6–1.0)^†
Depressive symptoms	1.0 (.9–1.0)	1.04 (1.0–1.1)^†	1.0 (.9–1.0)	1.05 (1.0–1.1)^†
Psychological distress	1.1 (1.0–1.1)^**	1.1 (1.0–1.1)^*	1.0 (.9–1.1)	1.08 (1.0–1.1)^*
Reaction when seeing bullying events
Never noticed	1.0	1.0	1.0	1.0
Take part in/think bullying is acceptable	.8 (.4–1.7)	1.0 (.2–3.9)	1.6 (.4–5.9)	.8 (.1–4.9)
Think they ought to help	1.1 (.6–1.9)	.8 (.2–2.8)	1.3 (.4–3.9)	1.1 (.3–4.5)
Try to help stop bullying	.9 (.6–1.5)	2.2 (.9–5.5)	1.7 (.7–4.4)	2.4 (.8–7.3)
Time spent online	1.0 (.9–1.1)	1.1 (.9–1.2)	1.1 (.9–1.2)	.9 (.8–1.1)
Parents’ supervise online
Frequent	1.0	1.0	1.0	1.0
Infrequent	1.6 (1.0–2.4)^**	1.0 (.5–2.2)	.7 (.3–1.4)	1.7 (.7–4.4)
Parents’ control Internet access and mobile phone
Infrequent	1.0	1.0	1.0	1.0
Frequent	1.4 (.9–2.0)	1.1 (.5–2.3)	.8 (.4–1.7)	1.8 (.8–4.3)
Family social support
High	1.0	1.0	1.0	1.0
Low	1.5 (.9–2.3)^†	.8 (.4–1.7)	1.0 (.4–2.2)	1.4 (.5–3.7)
Witness parental violence
No/rarely	1.0	1.0	1.0	1.0
Often	1.2 (.8–1.8)	1.6 (.8–3.3)	1.3 (.6–2.7)	1.0 (.4–2.4)
Conflict with siblings
No/rarely	1.0	1.0	1.0	1.0
Often	1.3 (.9–1.9)	1.3 (.6–2.6)	1.2 (.5–2.4)	1.2 (.5–2.9)
Perception of teachers trying to stop bullying
Frequent	1.0	1.0	1.0	1.0
Infrequent	.8 (.5–1.2)	.6 (.3–1.3)	.7 (.3–1.5)	.6 (.3–1.5)
School social support
High	1.0	1.0	1.0	1.0
Low	1.0 (.6–1.5)	1.7 (.7–3.8)	.9 (.4–2.1)	1.3 (.5–3.5)
Friend social support
High	1.0	1.0	1.0	1.0
Low	.9 (.6–1.5)	1.3 (.6–2.8)	1.5 (.7–3.2)	1.3 (.5–3.1)
Perception of students trying to stop bullying
Frequent	1.0	1.0	1.0	1.0
Infrequent	1.8 (1.2–2.9)^**	3.0 (1.3–6.7)^**	2.0 (.9–4.6)^†	1.5 (.6–4.1)

^† p < .10

^* p < .05

^** p < .01

^*** p < .001.

Predictors of temporal patterns of the ‘bully only’ role are presented in Table 3. Only a few factors were significantly associated with perpetration. Students who were not living with both biological parents had higher odds of being in the stable-low perpetration group. Students who witnessed parental violence had higher odds of being in perpetration at Time 1 than at Time 2 (i.e. the declining group), and those who had serious conflict with siblings had higher odds of being involved in the increasing perpetration, compared to students not involved in bullying.

Table 3. Multivariate multinomial logistic regression of predictors for intensity of stability in perpetration across Times 1 and 2.

	Intensity in stability in perpetration (ref group: Not-involved)
Predictors measured at Time 1	Stable low OR (95% CI)	Declining OR (95% CI)	Increasing OR (95% CI)
Gender
Female	1.0	1.0	1.0
Male	.9 (.3–2.6)	1.7 (.7–4.0)	1.6 (.7–3.6)
Age (years)	1.3 (.9–1.8)	.8 (.6–1.1)	1.0 (.7–1.2)
Depressive symptoms	1.0 (.9–1.0)	1.0 (.9–1.0)	.9 (.9–1.0)
Psychological distress	.9 (.8–1.0)	1.1 (.9–1.1)	1.0 (.9–1.1)
Time spending online	1.1 (.9–1.3)	1.1 (1.0–1.3)^†	1.1 (.9–1.2)
Family structure
Living with biological parents	1.0	1.0	1.0
Living with single parent/stepparent/others	4.8 (1.6–14.6)^*	.6 (.1–2.9)	1.6 (.5–5.0)
Family social support
High	1.0	1.0	1.0
Low	2.1 (.7–6.1)	1.3 (.5–3.3)	.8 (.3–2.1)
Witnessed parental violence
No/rarely	1.0	1.0	1.0
Often	.2 (.1–.8)^*	2.6 (1.0–6.7)^*	1.6 (.7–3.6)
Conflict with siblings
No/rarely	1.0	1.0	1.0
Often	2.2 (.8–6.3)	2.0 (.8–4.6)	2.9 (1.3–6.5)^*
Teachers’ supervise online
Frequent	1.0	1.0	1.0
Infrequent	.8 (.3–2.3)	1.4 (.6–3.2)	1.7 (.7–3.8)
School social support
High	1.0	1.0	1.0
Low	2.0 (.6–5.9)	1.4 (.5–3.4)	.9 (.4–2.2)

^† p < .10

^* p < .05

^** p < .01

^*** p < .001.

Table 4 summarises findings for factors associated with temporal patterns of bully-victim role over time. The adjusted models show that the odds of being in any bully-victim group significantly decreased with age, and were higher among males. Adolescents who were not living with both biological parents, or who witnessed parental violence often, and experienced low school social support had higher odds of being in the low level but chronic bully-victim group. Students who reported more depressive symptoms and those who showed their support to perpetrators were more likely to have been involved in the bully-victim group at a higher level at Time 1 than at Time 2 (i.e. the declining group). Interestingly, frequently reported parental control of children’s mobile phone and Internet access was associated with lower odds of being in the declining bully-victim group. Experiencing higher psychological distress increased the odds of being in the increasing or stable-high bully-victim groups. Adolescents who reported higher self-esteem, and those who passively watched and thought bullying was acceptable, and who perceived that teachers infrequently try to stop bullying at school, had higher odds of being high involvement as both victims and bullies over time.

Table 4. Multivariate multinomial logistic regression of associated predictors for changes in Bully-victim status across Times 1 and 2.

	Bully-victim group (Ref. group: not-involved)
Predictors measured at Time 1	Stable low OR (95% CI)	Declining OR (95% CI)	Increasing OR (95% CI)	Stable high OR (95% CI)
Gender
Female	1.0	1.0	1.0	1.0
Male	1.9 (1.2–2.8)^***	1.7 (.9–3.1)	2.2 (1.1–4.2)^*	2.3 (1.1–4.6)^*
Age (years)	.6 (.5–.7)^***	.7 (.6–.9)^**	.7 (.5–.9) ^**	.5 (.4–.7)^***
Depressive symptoms	1.01 (1.0–1.0)	1.07 (1.0–1.1)^**	1.01 (.9–1.1)	1.0 (.9–1.0)
Psychological distress	1.04 (1.0–1.1)	1.04 (.9–1.1)	1.07 (1.0–1.1)^*	1.07 (1.0–1.1)^*
Self-esteem	1.04 (1.0–1.1)	1.1 (1.0–1.2)	1.03 (.9–1.1)	1.1 (1.0–1.2)^**
Suicidal ideation
No	1.0	1.0	1.0	1.0
Yes	1.2 (.7–2.2)	1.1 (.5–2.4)	1.9 (.8–4.3)	1.6 (.7–3.9)
Reaction when seeing bullying events
Never noticed	1.0	1.0	1.0	1.0
Take part in bullying events	.4 (.04–4.4)	8.6 (1.8–40.5)^**	4.3 (.6–28.9)	4.2 (.6–30.5)
Passively watch and think bullying is acceptable	1.8 (.9–3.7)	1.6 (.5–5.1)	2.3 (.8–7.1)	3.4 (1.0–11.2)^*
Think they ought to help	1.7 (1.0–3.1)	1.2 (.5–3.4)	1.2 (.4–3.5)	2.4 (.8–7.4)
Help stopping bullying	1.2 (.7–2.0)	2.2 (.9–4.5)	1.1 (.4–2.8)	1.4 (.5–4.2)
Time spending on online	1.0 (.7–1.5)	.9 (.5–1.7)	.8 (.4–1.6)	1.1 (.5–2.3)
Family structure
Living with biological parents	1.0	1.0	1.0	1.0
Living with single parent/stepparent/others	2.3 (1.4–4.1)^**	.9 (.3–2.3)	.3 (.1–1.6)	1.5 (.5–4.1)
Parents’ supervise online
Frequent	1.0	1.0	1.0	1.0
Infrequent	1.2 (.8–1.8)	1.4 (.7–2.6)	1.9 (.9–3.8)^†	1.3 (.6–2.7)
Parents’ control Internet access and mobile phone
Infrequent	1.0	1.0	1.0	1.0
Frequent	.9 (.6–1.3)	.5 (.3–.9)^*	.8 (.4–1.6)	.9 (.4–1.8)
Family social support
High	1.0	1.0	1.0	1.0
Low	1.1 (.7–1.7)	1.8 (.9–3.6)	.8 (.4–1.8)	1.4 (.6–3.1)
Witness parental violence
No/rarely	1.0	1.0	1.0	1.0
Often	2.9 (1.9–4.4)^***	1.8 (1.0–3.4)	2.0 (1.0–4.2)^*	3.1 (1.4–6.8)^**
Conflict with siblings
No/rarely	1.0	1.0	1.0	1.0
Often	.9 (.6–1.3)	1.3 (.7–2.4)	1.1 (.6–2.2)	1.6 (.8–3.3)
Teacher’s control mobile phone/Internet frequently
Frequent	1.0	1.0	1.0	1.0
Infrequent	1.2 (.8–1.8)		1.6 (.8–3.3)	.6 (.3–1.3)
Perceive that teachers trying to stop bullying
Frequent	1.0	1.0	1.0	1.0
Infrequent	1.1 (.7–1.7)	1.1 (.6–1.9)	1.3 (.6–2.5)	2.2 (1.1–4.6)^*
School social support
High	1.0	1.0	1.0	1.0
Low	1.6 (1.0–2.6)^*	.9 (.4–1.8)	1.8 (.8–3.8)	1.3 (.6–3.0)
Friend’s social support
High	1.0	1.0	1.0	1.0
Low	1.1 (.7–1.7)	.8 (.4–1.5)	.6 (.3–1.3)	1.0 (.5–2.1)
Perceived that students try to stop bullying
Frequent	1.0	1.0	1.0	1.0
Infrequent	1.4 (.9–2.2)	1.3 (.7–2.5)	1.2 (.6–2.3)	1.3 (.6–2.8)

^† p < .1

^* p < .05

^** p < .01

^*** p < .001.

Discussion

This is the first short-term quantitative longitudinal study of bullying in Vietnam and any Southeast Asian country (Sittichai & Smith, 2015) and one of few international studies examining predictors for different levels of temporal stability in bullying roles among school adolescents. It is clear that traditional bullying victimisation and perpetration are much more common than cyberbullying victimisation and perpetration among Vietnamese school students. This pattern is unlikely to be due to limited online activity, as over 90% of students reported using mobile phones and other devices that connect to the Internet and most spend at least one hour daily online. The dominance of traditional bullying victimisation and perpetration and high correlations between traditional bullying and cyberbullying are consistent with previous studies in both Western and Asian countries (Chan & Wong, 2015; Modecki, Minchin, Harbaugh, Guerra, & Runions, 2014; Olweus, 2013). There is growing evidence worldwide that cyberbullying rarely occurs in isolation from traditional forms. Preventive intervention should address all forms of bullying rather than focus heavily on the online environment (Modecki et al., 2014). Anti-bullying programs should include components on cyberbullying within the context of broader efforts to prevent interpersonal conflict and violence.

In relation to stability or change in bullying roles, our findings show that about 60% of students engaged in bullying roles as victim, bully, or bully-victim at some point during the year. Of these students, nearly three in four indicated unstable bullying roles over time. The findings are consistent with other studies indicating that bullying experiences are fluid (Gumpel et al., 2014; Lereya et al., 2015). Among the students who experiences remained stable, the two largest groups were those with stable-low victimisation or stable-low bully-victim status. The largest group of perpetrators also had low level and stable involvement. In contrast, stable high involvement over time was relatively rare for victims, bullies and bully-victims. These patterns are similar to a North American study (Ryoo et al., 2015) that showed victimisation and bully-victim status were more stable at low level involvement. In contrast to the relatively few stable-high victims or perpetrators, students were more likely to change their role over the six-month period (i.e. declining or increasing groups).

Consistent with international trends, age differences were observed among the victimisation and bully-victim groups in Vietnam. Students were more likely to be victimised in early secondary school (sixth and seventh grade) while transitioning from primary school to a new environment (Cook et al., 2010; Olthof, Goossens, Vermande, Aleva, & van der Meulen, 2011). At this time, some children seek social dominance over their peers in a new environment; for many, this places them at risk of being in the victimisation or bully-victim group. This underscores the need to start prevention of bullying early when students are in primary school and especially for students in transition years (Olthof et al., 2011).

The present study found that adolescents who experience poor mental health might be most at risk of becoming victimised, or being bully-victims, over time (Cappadocia, Craig, & Pepler, 2013; Goldbaum, Craig, Pepler, & Connolly, 2003). Analyses of these data indicated that those with the worst mental health had more persistent and frequent involvement as victims or bully-victims. To be effective, anti-bullying efforts should be a core element of mental health promotion in schools rather than addressed in programs that stand alone from mental health promotion efforts (Carta, Fiandra, Rampazzo, Contu, & Preti, 2015).

As in other societies, it is clear that social normalisation of bullying has negative effects among Vietnamese adolescents. Perceptions that peers typically react against bullying are significantly associated with declining victimisation (Denny et al., 2015). This is also consistent with findings from previous research indicating that school students are more likely to be victimised when there are social rewards to the perpetrators (like cheering) and less support to the victims (Denny et al., 2015; Salmivalli, 2014). It is possible that if students are aware of their teachers trying to stop bullying in school, they change their strategy to bully their peers in covert forms, including cyberbullying, where the teachers are less able to monitor the behaviour (Elledge et al., 2013).

There was no significant association between family support and bullying experiences. It is possible that students think parents are unable to help them to solve the problem or it is not discussed if there are wide gaps in communication between children and their parents (Trinh, Steckler, Ngo, & Ratliff, 2009). The situation is further complicated because children’s experience at violence at home influence their bullying experiences outside. Similar to previous studies, Vietnamese students who witness parental conflict or are bullied by siblings are more prone to perpetrate bullying (Hemphill et al., 2012; Hong & Espelage, 2012). It seems that students involved in bullying as perpetrators or as bully-victims are exposed to complex risk environments with many relationship problems with family, school, and peers. Indeed, while bullying events mostly happen in school settings, the majority of predictors of perpetration in this study were related to the family environment. Violence and disharmony at home predict perpetration and bully-victim status. It is recommended that future protective interventions and broader prevention programs need to address family conflict (Hemphill et al., 2012). Parents should be engaged in anti-bullying programs and be educated to recognise the impact of their own behaviours and home environment. In addition, the link between inter-sibling aggression and peer bullying (Wolke, Tippett, & Dantchev, 2015) suggests interventions need to involve other family members.

In conclusion, these novel findings from a Southeast Asian country strongly suggest that anti-bullying programs should incorporate action across a wide range of settings and social relationships, rather than focus primarily on classmate relationships or social skills training (Mitchell, Finkelhor, Jones, & Wolak, 2012). This study has highlighted the need for comprehensive prevention and intervention programs against bullying in schools throughout Vietnam.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

Ms Le was supported by an Australia Awards PhD Scholarship from the Australian Government. This work was supported by Know Violence in Childhood.