Connections Between Online Harassment and Offline Violence among Youth in Central Thailand

Defining online harassment and cyberbullying

Research in the last 10 years has documented the emergence of intentionally hurtful acts committed by youth using digital communications technologies, such as mobile phones and the Internet. These behaviors have variably been conceptualized as cyberbullying (Kowalski, Giumetti, Schroeder, & Lattanner, 2014), digital bullying (Olweus, 2013), online bullying (Microsoft, 2012), Internet harassment (Tokunaga, 2010), electronic aggression (Hertz & David-Ferdon, 2008), electronic bullying (Raskauskas & Stoltz, 2007), cyber aggression or c-aggression (Pornari & Wood, 2010), or as online harassment (Wolak, Mitchell, & Finkelhor, 2007).

These terms reflect differences in how the phenomenon is defined (Tokunaga, 2010). Many studies have defined the phenomenon using the three criteria borrowed from traditional bullying (or offline bullying) literature: intent to harm, power imbalance between victim and perpetrator, and repetition (Dooley, Pyżalski, & Cross, 2009; Gradinger, Strohmeier, & Spiel, 2010; Kowalski, Limber, & Agatston, 2012; Olweus, 2013; Tokunaga). Smith, del Barrio, and Tokunaga (2012) have discussed the issue at length and argued that these three criteria are largely appropriate for the study of cyberbullying.

Others have argued that the criteria may be overly restrictive (e.g., Wolak et al., 2007). Only including repeated acts overlooks single incidents that have lasting negative consequences if the material involved remains online and may be forwarded on and on by others (Dooley et al., 2009; König, Gollwitzer, & Steffgen, 2010; Olweus, 2013). Furthermore, victimizing others online may not require greater power, whereas in the offline world it is often crucial (Dooley et al.; Shariff, 2008). However, some have asserted that the criterion of power imbalance is meaningful for cyberbullying if it is understood as “differences in technological know-how between perpetrator and victim, relative anonymity, social status, number of friends, or marginalized group position” (Smith et al., 2012, p. 36) rather than the more traditional forms of power imbalance, such as those related to bigger body size or greater physical strength.

As in the traditional bullying literature, intentionally hurtful online behaviors comprise a larger phenomenon of intentionally hurtful behaviors per se, and a narrower subset of such behaviors that occur in the context of power imbalances and are typically repeated (Olweus, 2013). Olweus has argued that to avoid confusion, the term “bullying” should be reserved for the narrower subset of behaviors-in-context. Correspondingly, Wolak, Mitchell, and Finkelhor (2007) introduced the term “online harassment” to refer to the larger phenomenon of intentional behaviors to harm others through the Internet or mobile devices that may or may not involve power imbalances and repetition. We follow their usage (see Ojanen et al., 2014), reserving the term “cyberbullying” for intentionally harmful behaviors that occur in the context of power imbalance and repetition, except when referring to works of other authors using the term.

Participants and sampling

Participants were 15-24 year-olds living, studying, and/or working in the district. Youth participants belonged to one of three groups: 1) current university students, 2) current secondary school students including vocational college students, or 3) youth currently not attending any educational institution (out-of-school youth). Students were sampled through educational institutions and out-of-school youth from communities. The secondary school subsample was a probability-based, stratified random sample of students in secondary educational institutions within the district. The out-of-school youth and university student subsamples were essentially convenience samples.

Measures

The research team created a computerized survey instrument to collect mostly numeric data on 1) demographics, 2) online harassment, and offline violence, 3) online and mobile media use, and 4) sexual life and partnerships.

Most items on offline youth violence were adopted or adapted from a recent large-scale Thai study (Pradubmook-Sherer et al., 2008). Many items on online harassment were adopted or adapted from a Thai master's thesis (Rungsakorn, 2011) that used the same set of questions that most quantitative Thai cyberbullying studies have used so far. These two existing instruments formed the basis of our measurement of online harassment and offline youth violence so as to enable us to examine connections between phenomena described by these two streams of previous research in Thailand. Table 1 lists all violence and harassment items asked in the survey. Some of the original items described two distinct behaviors and were split to describe only one distinct behavior each. We also added some items to measure phenomena described by our focus group or interview participants. The questions were behaviorally defined, so the participants' perception of what might constitute violence or harassment was not at issue. The question was “in the past one year, have you had the following experiences”, and the participants were requested to tick the number of times (if any) they had been involved in each type of incident as perpetrators, victims, and/or witnesses.

In this article, any participant who responded in the affirmative about having perpetrated, been victimized, or witnessed at least one type of incident listed in Table 1, at least once in the past one year, was categorized as having perpetrated, been victimized, or witnessed online harassment or offline violence, respectively. Cronbach's alpha for the dichotomized scale for offline violence perpetration was 0.66; for the dichotomized scale for offline victimization it was 0.69. For the dichotomized scale for online harassment perpetration, Cronbach's alpha was 0.75, and for the dichotomized scale for online harassment victimization it was 0.77.

The survey program was created in collaboration with an outsourced programmer. Context-specific animations portraying youth were added to make the task more engaging. The program was pilot tested with youth from educational institutions and local communities. Based on their comments, the order of the questions was changed, some items were reworded, and some were deleted to reduce the length of the questionnaire. The completed survey took from 15 minutes to an hour to complete (depending on the participant's literacy and computer skills).

Procedure

Universities and secondary schools were first requested to participate by telephoning them, then sending them a formal request letter, and when appropriate, making a follow-up visit. All identified educational institutions in the district were contacted, and all except one secondary school participated. At participating secondary institutions, students in the classrooms sampled were given the participant information sheet and a consent form, and asked by their teachers if they would like to participate. Some university students were directly approached in their faculties by research team members to request their participation. All data from educational institutions were collected in the premises of these institutions. Participation in the study was voluntary and all data were kept confidential.

Most out-of-school participants were approached directly within each geographical community or introduced to the research team by other participants. In some cases, community members volunteered to help recruit participants, even if they themselves did not participate. The community data were collected in diverse contexts including various kinds of shops, Internet cafés, snooker halls, factories, food courts, markets, housing developments, private homes, sport venues, district council election sites, a temple fair, and a golf course. Recruitment and the informed consent process usually took place immediately prior to data collection.

In both settings, 18-24 year old participants decided by themselves whether to participate, and gave written informed consent if they did participate. For 15-17 year old participants, written informed consent was requested from their guardians and informed assent was obtained from the participants. All participants were verbally briefed about the study and that all data would be kept confidential. They also received a participant information sheet. Participants completed the computerized survey on their own, using netbook computers provided by the research team, and received further assistance if they requested it. All participants were given an incentive payment of 50 baht (1.7 USD) in cash. The study was reviewed and approved by institutional review boards at Mahidol University and University of Pittsburgh.

Prevalence comparisons with other studies

In this study, online harassment victimization and perpetration rates were similar to offline youth violence perpetration and victimization. This suggests that Thailand may be closer to the general picture of Asian countries like China and Singapore, where cyberbullying (or online harassment) appears to be more common than its offline equivalents (Microsoft, 2012; Zhou et al., 2013), rather than Western countries where traditional bullying is considerably more common than its online equivalents (Olweus, 2013). Kowalski et al. (2014) have interpreted this difference between Western and Asian countries (or research participants) in terms of the individualistic/collectivist culture dimension, and in terms of power distance.

The prevalence of overlapping victimization and perpetration found in this study falls within the range found in other countries. In the present study, 30.3% had perpetrated and 33.6% had been victimized both online and offline. In a recent Irish study (O'Moore, 2012, p. 209), 29.8% reported having been bullied both offline and online, and 24.4% said they had bullied others both online and offline. In a US-based study, 36% experienced both online and offline bullying (Ybarra et al., 2007). Yet, in an Australian study, only 7% were both cyberbullies and traditional bullies (Hemphill, Kotevski, & Tollit, 2012). The 25-country average in the Microsoft (2012) study was 23% for concurrent online and offline victimization and 13% for concurrent perpetration.

The role of definitions in estimating the prevalence of violence, bullying, and harassment

The above figures may not make permit stating that one context or country has more concurrent victimization or perpetration than another, because different age groups and different conceptualizations of the phenomenon are involved (e.g., violence vs. bullying vs. harassment; one or more incidents vs. repeated incidents only). Following Wolak et al. (2007), the present study focused on violence/harassment behaviors rather than on bullying as a behavior-in-context (Olweus, 2013); in other words, repetition and power imbalance between perpetrator and victim were not criteria for being counted as having been involved in a violence/harassment incident. .

Our use of a behavioral checklist (see Table 1) is likely to have captured a relatively higher proportion of those with violence or harassment experiences than if they had been asked about any involvement in “violence,” “bullying,” or “harassment,” which youth might define more narrowly than researchers (Gradinger et al., 2010; Ybarra, Boyd, Korchmaros & Oppenheim, 2012). Our qualitative findings from the same study (Ojanen et al., 2014, p. 12) indicated that youth in our context had narrow definitions of violence, and so, had we asked about “violence” experiences, many of the behaviors we included in our survey as violent behaviors would probably not have been counted had we asked specifically about “violence” experiences.

Correspondingly, a higher proportion of our participants reported offline perpetration and victimization than those of Pengpid and Peltzer (2013), whose data on Thai students on grades 7-10 were collected using a Global School-Based Health survey, which used the word “bullying” and gave a definition of it. In their study, 27.8% had been bullied (though their time period for this variable was also much shorter - the past 30 days), whereas 33.3% had been in a physical fight in the past 12 months. However, the comparison is also complicated by the fact that our sample included out-of-school youth (unlike the sample of Pengpid & Peltzer), who had generally higher levels of violence and harassment involvement both as victims and as perpetrators than students (data not shown).

Within-context cyclicality and cross-context role consistency

The findings of our study correspond to findings from previous studies (Juvonen & Gross, 2008; Kowalski et al., 2014; Kowalski & Limber, 2011, cited in Kowalski et al., 2012, p. 106-107; Microsoft, 2012; Sourander et al., 2010) indicating that those with one kind of involvement in violence, bullying, or harassment are more likely to also have other kinds of involvement. The modeling conducted in this study suggests two key concepts: 1) within-context cyclicality of violence/harassment (within a context, perpetrators also tend to be victims, and vice versa); and 2) cross-context role consistency (victims in one context also tend to be victims in the other; perpetrators in one context also tend to be perpetrators in the other).

These concepts have implications for our understanding of the dynamics of both online harassment and offline violence. One previous German study (König et al., 2010) has argued that taking revenge for offline victimization may be an important motivation behind cyberbullying, as 83.3% of those bullied offline were themselves cyberbullies. A similar pattern was seen in the present study: 57.8% of those victimized offline (versus 27.3% of those not victimized offline) were themselves perpetrators online.

Yet, in the multivariable modeling, the adjusted odds ratios for violence/harassment cross-context and across roles were statistically non-significant. Thus, rather than implying online revenge for offline slights, these figures might be better explained by the other roles the same individuals hold. More specifically, those victimized offline might be more likely to harass others online because they are also being victimized online (online-only cyclicality: AOR 10.1). Alternatively, the harassment they perpetrate online might in fact be a continuation of the violence they are also perpetrating offline (cross-context perpetrator role consistency: AOR 2.7).

The importance of within-context role consistency seen in this study replicates findings from a study conducted in Finland, in which the greatest (and in some cases, only significant) AORs in a similar logistic regression model generally were between the same role across the two contexts (Sourander et al., 2010). Similarly, in a meta-analysis of the cyberbullying literature (Kowalski et al., 2014), the correlations across roles and contexts (i.e., between cybervictimization and traditional bullying perpetration, or cyberbullying perpetration and traditional bullying victimization) were lower than for pairs involving at least the same context (e.g., between cyberbullying perpetration and cyberbullying victimization) or the same role (e.g., cyberbullying perpetration and traditional bullying perpetration). Our findings also correspond to those of Kowalski et al. in the sense that in both studies, the highest correlations were observed between perpetration and victimization within the same context (either online or offline).

However, youth perceptions obtained qualitatively within our project (data not shown) suggest that spillover across the offline and online contexts, and the perpetrator and victim roles sometimes does happen; this is motivated by a desire to take revenge, as suggested by König et al. (2010). Yet, such perceptions also emphasize that spillover is only likely between people who have previous offline contact. This seems to fit Olweus' (2013, p. 767) view that “to be cyberbullied or to cyberbully others seems to a large extent to be part of a general pattern of bullying, where use of the electronic media is only one possible form.” By extension, our findings suggest that there may be a similar continuity between the more inclusively defined youth violence and online harassment (so, online harassment between youth can be seen as a subtype or online equivalent of youth violence).

Witnessing violence/harassment and its consequences

Our participants were also asked about witnessing violence and harassment. This was not only done to assess the extent to which violence or harassment experienced as a third party forms a part of the participants' daily lives, but also to assess associations between involvement as a witness and involvement as a perpetrator or a victim. Being exposed to violence may in itself be detrimental to health (Graham-Bermann & Seng, 2005), but it can also contribute to a culture of violence in which violence (or, by extension, online harassment) is seen as normal and inevitable, and thus more easily perpetrated and tolerated (Galtung, 1990; Pradubmook Sherer & Sherer, 2011).

Our results demonstrate that the majority of youth in our context have witnessed offline violence and online harassment within a one-year time frame, and that witnessing is correlated with perpetration and victimization. However, because witnessing experiences were so common, they would not be particularly useful as predictors of violence/harassment. Culture of violence (or cultural violence) is best seen as a societal-level explanation of violence. On the individual level, culture of violence can arguably be operationalized in terms of moral disengagement and normative beliefs about aggression, which a recent meta-analysis (Kowalski et al, 2014) indicated were associated with cyberbullying perpetration.

The role of sexuality and gender norms in violence, harassment, and bullying

One previous qualitative study conducted in the Thai context has suggested that cyberbullying may particularly often be motivated by sexual jealousy in the offline context (Surat, 2010). Sensitized by this, we included partnership status as one of the controlled variables in our multivariable modeling, because our data exploration indeed indicated that youth with partners were considerably more likely to perpetrate offline violence and online harassment, as well as be victimized in both contexts. This elevated prevalence might have to do with conflicts between partners as well as with possible competitors in their love lives.

We also had to control for the sex of our participants because our findings indicated that male participants were more likely than female participants to perpetrate violence/harassment and be victimized, both online and offline (data not shown). This is likely to do with norms of masculinity and femininity. More generally, previous bullying studies conducted in Thailand also point at the role of sexuality and gender themes in bullying motivations and behaviors: Mahidol University et al (2014) reported that prejudice against same-sex attracted or transgender students was a common motivation of bullying, whereas Pengpid and Peltzer (2013) found that sexual jokes, comments, and gestures were the most common subtype of bullying behaviors.

The study was funded by the Thai government National Research Universities program (http://www.nru.go.th), and University of Pittsburgh Center for AIDS Research. Thomas E. Guadamuz was supported by a grant from the U.S. National Institute of Mental Health (MH085567). Darong Wannaburee designed and programmed the computerized survey as an outsourced consultant. Anusorn Payakkakom, Jetsada Taesombat, Kitsanee Senakao and Chanan Yodhong assisted with quantitative data collection. Wimontip Musikaphan and Penchan Pradubmook-Sherer provided background information and materials.