Introduction

Aggression between adolescents via electronic devices, instant messaging programs, and social networks (cyber-aggression) has been the object of growing interest in recent years (González-Moreno et al., 2020). Nowadays there is great social concern about this problem, because of its prevalence and effects (Machimbarrena et al., 2018). However, the nature of the construct, as distinct from traditional aggression, is still open to debate. Authors such as Olweus (2012) and Mehari et al. (2014), considered that these new forms of aggression were essentially no different to traditional forms, and should therefore not be studied in isolation. The underlying problem would be the same, except that it would happen in two different specific contexts (in-person and online). Both forms of victimization are positively correlated and share psychosocial predictors. Rather than being “new” victims, cyber-victims would generally be no different from those suffering from traditional aggression. It is, instead, a new form of aggression which usually overlaps traditional forms (physical, verbal, relational, etc.). In fact, according to authors like Olweus (2012) and Williford et al. (2018), most of the negative impact attributed to cyber-victimization should in reality be attributed to the overlap of traditional forms of aggression.

If an adolescent suffers from traditional aggression, and also from aggression via electronic devices, the latter would add little significant effect. The negative effects of cyber-victimization can only, therefore, be understood in a generally broader context of aggression, including traditional types. This is important, as it points the way towards the most effective prevention. It hints as to whether it is worth prioritizing specific programs for preventing aggression via electronic media (mainly focused on applications, Internet behaviour, and Internet safety) or whether more general programs for preventing aggression between adolescents would have a positive impact on the prevention of cyber-victimization even without specific elements aimed at this type of aggression. Previous studies suggest that programs aimed at reducing traditional aggression, despite not having specific elements about preventing online aggression, also have a positive effect on the prevention of cyber-victimization between adolescents (Gradinger et al., 2016; Salmivalli et al., 2011).

One argument in favour of this hypothesis, is the positive, significant relationship that is usually found between suffering from both types of victimization (online and offline). Some meta-analyses have found a fairly high correspondence between cyber and traditional victimization ( $\dot{r}$ =.40, Modecki et al., 2014; $\dot{r}$ =.43, Gini et al., 2018), although there can be differences depending on age and the context of the sample (Cosma et al., 2020). However, co-occurrence of these two phenomena is not perfect, since being the victim of traditional aggression is generally more prevalent than being a cyber-victim (López-Castedo et al., 2018; Modecki et al., 2014; Sastre, 2016; Viau et al., 2020).

The possible co-occurrence, the overlap between being a victim of offline and online peer aggression in adolescence, has also been examined in studies that have attempted to identify types of victims according to the extent to which they suffer from the different types of aggression (online and offline). The table shown in Álvarez-García (2020) provides a summary of the main studies in this regard. There have been few studies to date, mostly from the USA, with heterogeneous results. The number of different types of victims varies between three and five, with four being the most common. These approaches generally include a type characterized by null or minimal victimization, one or more types characterized by traditional victimization, and one or more types characterized by dual (cyber- and traditional) victimization. The most common typology includes the following four: non-victims, cyber-victims, traditional victims, and dual victims (Beltrán-Catalán et al., 2018; Gini et al., 2019; Mindrila, 2019). “Mainly cyber-victim” is the type with the least consensus. In some studies, cyber-victimization is included in a category of dual or multiple victimization, and in some of these studies distinct levels of dual or multiple victimization are differentiated (Álvarez-García, 2020).

As indicated above, one problematic aspect of studying cyber-victimization in adolescents is identifying the extent to which its relationships with significant variables (risk factors or consequences) are due to the cyber-victimization or whether they can only be understood in the broader context of ordinary victimization, including traditional types. In this study, we focus specifically on social anxiety and self-esteem. Regarding social anxiety, there is a solid body of evidence about the positive relationship between social anxiety and traditional peer victimization in adolescence (Gómez-Ortiz et al., 2017; Pontillo et al., 2019; Romera et al., 2016; Suárez-García et al., 2020). Some studies have concluded that social anxiety increased the likelihood of being a victim (Acquah et al., 2016; Pabian, &Vandebosch, 2016), others that being a victim increased the probability of having social anxiety (Calvete et al., 2018; Silberg et al., 2016; Van-den-Eijnden et al., 2014).

Studies looking at the relationship between adolescent cyber-victimization and social anxiety have also found a positive relationship between the two, noting that social anxiety is a risk factor for cyber-victimization, while there is less evidence that it is a consequence (Pabian & Vandebosch, 2016; Ruíz-Martín et al., 2019; Van-den-Eijnden et al., 2014). In this regard, some studies have found, in line with the hypothesis posed by Olweus (2012), that traditional (and more specifically, relational) victimization has a more negative impact on adolescent social anxiety than cyber-victimization (Landoll et al., 2015; Van-den-Eijnden et al., 2014); and that adolescents who are victims of both types of aggression do not generally present higher levels of social anxiety than those who are victims of only traditional violence (Cañas et al., 2020).

Regarding self-esteem, meta-analyses have shown that there is a reasonably solid body of evidence about its negative relationship with both traditional peer victimization (Tsaousis, 2016) and cyber-victimization (Chen et al., 2017; Kowalski et al., 2014). However, it is not clear which of the two types of victimization has a greater impact on self-esteem, nor the effect of the possible overlap between the two types of victimization on self-esteem. Previous research in Spain has shown that both victimization and cyber-victimization have negative effects on self-esteem, and that the effect is slightly weaker for cyber-victimization (Álvarez-García et al., 2019). Olweus (2012) found traditional victimization to significantly predict self-esteem, while additional cyber-victimization did not significantly contribute. In addition, recent meta-analyses of longitudinal studies have found results that depend on the type of victimization. Traditional victimization and self-esteem demonstrated a bidirectional causal relationship (Van-Geel et al., 2018). In contrast, there was no causal relationship between cyber-victimization and self-esteem (Marciano et al., 2020).

In light of the above, this study has two objectives. Firstly, to identify profiles of victimization in adolescence, depending on the levels of offline or online peer aggression suffered, as well as determining the prevalence of each victim type, in a sample of Spanish adolescents. Secondly, to analyse the association of the victimization profiles with adolescents’ social anxiety and self-esteem. With regard to the first objective, we expect to find the victim typology most often found in previous studies (non-victims, cyber-victims, traditional victims and dual victims) (Beltrán-Catalán et al., 2018; Gini et al., 2019; Mindrila, 2019), with a similar prevalence to these studies for each type of victim.

With regard to the second objective, we expect that the greater the level of victimization, the greater the adolescents’ social anxiety and the lower their self-esteem, and that traditional victimization will be more strongly related to both variables than cyber-victimization.

Material and methods

Results

Descriptive statistics

The descriptive statistics and Pearson correlations between the observed variables are shown at Table 1. The greater the victimization (traditional and online), the greater the anxiety and the lower the self-esteem. In the case of self-esteem, the correlation with traditional victimization is slight higher than the correlation with cybervictimization, and both are close to medium in size (r=-.237, d=0.49; r=-.198, d=0.41). However, for social anxiety, while its relationship with traditional victimization is close to medium in size (r=.209, d=0.43), the relationship with cyber-victimization is small (r=.096, d=0.19). There was no appreciable relationship between gender and either form of victimization, although there was a relationship with social anxiety and self-esteem.

Latent profile analysis

Selection of the best model

Three latent profile models were fitted. The results are shown in Table 2. Although the LMRT-test indicates that the four-class model is statistically no better than the three-class model, the other statistics suggested selecting the four-class model. The AIC, BIC, and SSA-BIC for the four-class model are all lower than the corresponding values for the three-class model. Besides, the entropy is greater in the four-class model, indicating that this model has a better quality than the three-class model. Equally important is the fact that at a theoretical level, there is more basis for a four-class model.

In addition, the two variables were important in differentiating the four profiles: cyber-victimization (F(3,3116)=3888.209; p<.001; d=3.87) and traditional victimization (F(3,3116)=1095.683; p<.001; d=2.05). All of the comparisons between classes in both variables were statistically significant at p <.001. In short, we selected the four-class model over the three-class model because it was statistically better, because it was theoretically valid, and because it demonstrated better quality than the three-class model for classifying the subjects in the classes. The probability of each subject being classified in their assigned class was excellent, both generally (.917) and for each class individually (class 1=.865; class 2=.974; class 3=.889; class 4=.937).

As noted above, the models were fitted using gender as a covariate. The data indicated that gender did not affect the model fit. More specifically, we found that the assignment of subjects to classes did not change with or without gender as a covariate (only one subject was assigned to a different class), and consequently entropy did not change either.

Description of the selected latent class model

The selected latent class model is made up of four profiles of victimization, depending on the levels of offline or online peer aggression suffered (Figure 1).

As is shown in Table 3, the largest victimization profile (77.76%) was composed of adolescents who had null or very low scores of any kind of victimization, and it was called “non-victims” (50.3% girls). The next largest group (13.49%) is composed of adolescents who, in general, had low scores of victimization, but that had suffered predominantly from cyber-victimization, so it is called “cyber-victims” (46.1% girls).

The third group (4.46%) had mainly suffered from traditional victimization, so it is called “traditional victims” (39.6% girls). The fourth and smallest group (4.29%) is “dual victims” (53% girls). It is made up of students who had suffered both kinds of victimization significantly over the sample mean.

Victimization profiles and health-related outcomes

Non-victims demonstrated the lowest levels of social anxiety and the highest levels of self-esteem. In contrast, traditional victims and dual victims demonstrated the highest levels of social anxiety and the lowest levels of self-esteem (Tables 4 and 5).

In general terms, the data showed statistically significant differences between the profiles in both social anxiety ( ${\chi }^2$ =53.847; p<.001) and self-esteem ( ${\chi }^2$ =84.288; p<.001). Considering the multiple comparisons between classes, all the differences were statistically significant, with the exception of the comparison between traditional victims and dual victims in both variables (Table 5).

Discussion and conclusions

The first objective of this work was to identify profiles of victimization in adolescence, depending on the levels of offline or online peer aggression suffered, as well as determining the prevalence of each victim type, in a sample of Spanish adolescents. We found, as previous meta-analyses have found (Modecki et al., 2014; Gini et al., 2018), a fairly high positive correlation between being a victim of traditional aggression and being a cyber-victim. The size of the correlation obtained in the present study was very similar to the average correlation obtained in these two meta-analyses. The typology of victimization profiles we found agree with the typology found by Beltrán-Catalán et al. (2018), Gini et al. (2019) and Mindrila (2019), which is what has been found most often in the few published studies on this topic (Álvarez-García, 2020). This typology distinguishes between four types of adolescents: non-victims, cyber-victims, traditional victims, and dual victims. Most of the studies we reviewed (Álvarez-García, 2020) distinguished between non-victims, traditional victims, and dual victims. However, there was less agreement on the category of “cyber-victims”. Our results suggest the importance of considering adolescents who are mainly cyber-victims as a specific type of victims.

As in all of the typologies proposed in the studies we reviewed (Álvarez-García, 2020), in our study the most common profile by far was non-victims. The “dual victims” category had the fewest members, which is the same as in the majority of studies we reviewed which included this category. In our study, the percentage of adolescents suffering predominantly from cyber-victimization was larger than the percentage suffering predominantly from traditional victimization, which contradicts previously published results.

In previous studies where a cyber-victim category has been found, the percentage of students in that category has been significantly smaller (Beltrán-Catalán et al., 2018; Mindrila, 2019) or smaller but not significantly (Gini et al., 2019), than the percentage in traditional victim categories. It is difficult to accurately determine the reason for this disparity, due to the variety of studies in terms of methodology, indicators used, age range and context of the sample (Álvarez-García, 2020). However, a possible reason is that in the present study the adolescents grouped in the “mainly cyber-victimization” class present a less severe general level of victimization than those grouped in the “traditional victimization” class. Maybe that is why the “mainly cyber-victimization” group is more numerous. In the study by Gini et al. (2019), in which the difference in the size of both classes (cyber and traditional victimization) hardly exists, both groups are more similar in overall levels of victimization than in the present work.

Our second objective was to analyse the relationships between the victimization profiles we found, and social anxiety and self-esteem. With social anxiety, our results were consistent with our hypothesis, which was based on prior findings (Gómez-Ortiz et al., 2017; Landoll et al., 2015; Pabian & Vandebosch, 2016; Pontillo et al., 2019; Romera et al., 2016; Ruíz-Martín et al., 2019; Van-den-Eijnden et al., 2014): the greater the level of victimization, the greater the level of social anxiety, with a stronger relationship between social anxiety and traditional victimization than with cyber-victimization.

Our correlation analysis demonstrated that social anxiety was positively, statistically, significantly, associated both with traditional and cyber-victimization. The correlation was stronger with traditional victimization than with cyber-victimization. Analyses of the differences in social anxiety depending on the type of victimization suffered showed that non-victims exhibited the lowest levels of social anxiety. Students who were mainly cyber-victims demonstrated greater social anxiety than non-victims. Students who were mainly victims of traditional aggression demonstrated higher levels of social anxiety than those who were mainly cyber-victims, and therefore than non-victims as well.

Students who suffered from both traditional and cyber-aggression exhibited higher levels of social anxiety than those who were mainly cyber-victims, but there was no statistically significant difference in social anxiety between them and the victims of traditional aggression. This result, in addition to supporting results from previous studies (Gómez-Ortiz et al., 2017; Landoll et al., 2015; Pabian & Vandebosch, 2016; Pontillo et al., 2019; Romera et al., 2016; Ruíz-Martín et al., 2019; Van-den-Eijnden et al., 2014), is consistent with the hypothesis put forward by Olweus (2012), that often cyber-victimization appears linked to traditional victimization, and that in order to understand it, it should be analysed in a broader context of mistreatment, including face-to-face aggression. Our results about self-esteem were consistent with our starting hypothesis, which was based on the prior evidence (Álvarez-García et al., 2019; Chen et al., 2017; Kowalski et al., 2014; Marciano et al., 2020; Tsaousis, 2016; Van-Geel et al., 2018). The greater the level of victimization adolescents suffer, the lower their self-esteem, and this relationship is stronger with traditional than with cyber-victimization.

Our correlation analysis showed that self-esteem was statistically significantly, negatively correlated to both traditional and cyber-victimization. The correlation was slightly stronger with traditional than with cyber-victimization. Analysis of the differences in self-esteem depending on the type of victimization adolescents suffered from showed that non-victims had the highest self-esteem. Adolescents who were mainly cyber-victims exhibited lower self-esteem than non-victims. Mainly traditional victims exhibited lower self-esteem than mainly cyber-victims, and thus also lower than the non-victims. Similar to social anxiety, adolescents suffering from both traditional and cyber-aggression exhibited lower self-esteem than those who were mainly cyber-victims, but they did not differ statistically significantly from traditional victims. This result, in addition to supporting results from previous studies (Álvarez-García et al., 2019; Chen et al., 2017; Kowalski et al., 2014; Marciano et al., 2020; Tsaousis, 2016; Van-Geel et al., 2018), is consistent with Olweus’ (2012) hypothesis that often cyber-victimization appears linked to traditional victimization and to understand it, should be assessed in a broader context of mistreatment, including face-to-face aggression. In fact, Olweus (2012) found that being a traditional victim significantly predicted poor self-esteem, whereas being a cyber-victim did not contribute significantly.

Our study makes a contribution to this field. It is one of the few studies published at an international level to date using mixture modelling in an attempt to identify types of victims in adolescents depending on the extent to which they suffered from online or offline peer aggression. As far as we are aware, it is the first such study in Spain. The results have important theoretical and practical implications. From the theoretical perspective, the study contributes to the consideration of cyber-victimization as a construct in and of itself, which is nonetheless related to traditional victimization. To better understand its risk factors and consequences, it should be examined in the broader context of possible mistreatment, including face-to-face contexts.

From a practical standpoint, our results suggest that, firstly, it is important for early detection of peer aggression to be alert in educational and clinical contexts to students with social anxiety and low self-esteem, because of their higher risk of becoming a victim of peer aggression. Secondly, in terms of prevention and treatment, it is important to work on self-esteem and social interaction skills. Lastly, our results do not support the idea that it is sufficient to have general preventive programs against mistreatment among adolescents in order to prevent cyber-victimization. There is an overlap, and the attitudes, values, and skills that are worked on to prevent aggression in face-to-face contexts can have a positive effect on preventing aggression in virtual contexts (Gradinger et al., 2016; Salmivalli et al., 2011). Nevertheless, there are some adolescents who are mainly cyber-victims, and aggression via virtual contexts has some particular characteristics that mean it is advisable to work on it specifically (Del-Rey et al., 2018; Hinduja & Patchin, 2015).

Although this study does represent a contribution to the field, it is important to recognize some limitations. Firstly, the data were collected via self-reports, which can often produce response biases such as social desirability -although we attempted to minimize this during the process by ensuring the anonymity and confidentiality of results. Secondly, we used a large sample which was representative of the population it was taken from, but limited in age and geographical area. Any generalization of these results to other ages or other geographical areas should be done with caution. It would be advisable to replicate this study in other samples to examine the external validity of the results. (1)