During the last decade, cyberbullying has become an increasing concern which has been addressed by diverse theoretical and methodological approaches. As a result there is a debate about its nature and rigorously validated assessment instruments have not yet been validated. In this context, in the present study an instrument composed of 22 items representing the different types of behaviours and actions that define cyberbullying has been structurally validated and its cross-cultural robustness has been calculated for the two main dimensions: cyber-victimization and cyber-aggression. To this end, 5679 secondary school students from six European countries (Spain, Germany, Italy, Poland, United Kingdom, and Greece) were surveyed through this self-report questionnaire which was designed based on previously existing instruments and the most relevant conceptual elements. Exploratory and confirmatory factor analyses were conducted and the global internal consistency was computed for the instrument and its two dimensions. Identical factor structures were found across all of the six subsamples. The results contribute to existing research by providing an instrument, the European Cyberbullying Intervention Project Questionnaire, which has been structurally validated in a wide sample from six different countries and that is useful to evaluate psycho-educative interventions against cyberbullying.
In recent years, society has shown a growing interest in the phenomenon named cyberbullying frequently appearing in the online social relationships among youngsters and adolescents (Fenaughty & Harré, 2013). Nowadays, we are immersed in the process of elaborating a solid theoretical approximation and an agreed definition of the phenomenon (Berne et al., 2013 and Tokunaga, 2010). Thus, one of the main guides to follow is the research developed around traditional bullying (Olweus, 2013) as cyberbullying is defined as bullying developed through electronic media (Vivolo-Kantor, Martell, Holland, & Westby, 2014). Traditional bullying has been defined as physical, verbal, social and/or psychological aggression by a pupil against another, whom is chosen to be a victim of repeated attacks (Olweus, 1993 and Olweus, 1999). Such a negative and intentioned action puts the victim in a situation that is difficult to get out of. Bullying is neither an isolated aggression nor a simple individual behaviour but an interactive phenomenon in which several subjects are involved in at least three roles: bully, victim and bully-victim. Its distinctive characteristics are: the intentionality to hurt someone else, the imbalance of power between the aggressor and the victim and the repetition of the aggressive conducts by the aggressors over their victims. Such scientific evidences have clarified the nature of the bullying phenomenon and determined its standardization, hence, the appearance of instruments to measure it (Greif & Furlong, 2006).However, the nature of the electronic means that characterizes cyberbullying has made it necessary to investigate not only its conceptualization but also in order to provide instruments suitable to its nature with the aim of showing the levels of prevalence among adolescent population (Vivolo-Kantor et al., 2014).
7. Results
7.1. Descriptive analysis
The Kruskal–Wallis test has been performed with the aim of comparing the differences among the participant countries with regard to the two dimensions of the questionnaire, cyber-aggression and cyber-victimization, and to the total questionnaire. Differences among participant countries were found (see Table 1).
Table 1.
Kruskal–Wallis test .
Country N Mean rank Mean Rank Mean Rank
Total sacale Poland 900 2697.09 Cyber agression 2906.57 Cyber Victim 2599.78
Spain 859 2509.55 2681.66 2528.01
Italy 1430 3183.37 3230.88 3072.31
UK 737 2334.24 2312.35 2549.22
Germany 846 2545.05 2830.72 2497.55
Greece 907 3439.48 2745.03 3563.27
χ2 = 350.60 p = .00 χ2 = 247.21 p = .00 χ2 = 363.90 p = .00
Table options
7.2. Exploratory factor analysis
In the EFA with subsample S1 the variables violated the assumption of multivariate normality (see Table 2), Mardia’s coefficient = 777.18. The Kaiser–Meyer–Olkin of Sampling Adequacy (KMO) value was .832 and Bartlett’s Test of Sphericity χ2: 71170.0 (df = 231; p < .001). The factors were subjected to a direct oblimin rotation and the analysis revealed that the extraction of two factors was appropriate (see Table 2). The variance accounted for by the two factors was 65.4%. Also, Bentler’s simplicity (S) index (1977) and the loading simplicity (LS) index ( Bentler, 1977 and Lorenzo-Seva, 2003) were computed with values of .99 and .75, respectively, which means that each item was mainly related to only one dimension, and the overall solution showed high factor simplicity. The inter-factor correlation is .765.
Table 2.
Descriptive univariate analysis and factor loadings.
M SD Skewness Kurtosis Factor 1 Factor 2
CybV1 .27 .645 3.305 13.362 .359
CybV2 .23 .606 3.441 14.298 .498
CybV3 .12 .471 5.347 34.466 .427
CybV4 .13 .492 5.015 30.112 .408
CybV5 .22 .489 3.051 14.493 .520
CybV6 .08 .384 6.101 43.589 .849
CybV7 .10 .448 5.646 36.979 .866
CybV8 .13 .463 4.629 26.313 .698
CybV9 .09 .398 6.318 47.837 .793
CybV10 .11 .424 5.447 36.627 .687
CybV11 .15 .509 4.704 26.398 .883
CybB1 .22 .623 3.934 17.955 .782
CybB2 .17 .562 4.427 22.986 .800
CybB3 .10 .492 6.096 40.367 .945
CybB4 .09 .487 6.284 42.252 .910
CybB5 .09 .471 6.499 46.054 .882
CybB6 .10 .490 6.147 41.200 .828
CybB7 .08 .463 6.788 48.921 .828
CybB8 .09 .443 6.500 47.670 .819
CybB9 .07 .408 7.486 62.431 .859
CybB10 .22 .678 3.881 16.329 .787
CybB11 .08 .443 6.746 49.208 .865
Table options
7.3. Confirmatory factor analysis
According to the EFA results, a two-factor model could be an adequate approximation of the data. The CFA was conducted with the conditions outlined above because of the violation of the assumptions of normality and kurtosis with Mardia’s coefficient = 629.17. The results of fitting the two-factor solution are suitable in sample S2 and in each subsample from the six countries in which the instrument was evaluated (see Table 3, Fig. 1).
Table 3.
Model Fit subsample S2 and six countries.
χ2 df GFI CFI NNFI RMSEA SRMR ECVI
Total Sample S2 1484.15 208 .986 .993 .993 .030 .080 .505
Poland 452.42 208 .992 .994 .993 .010 .058 .664
Spain 375.35 208 .973 .978 .975 .020 .055 .677
Italy 581.63 208 .966 .988 .987 .051 .087 .699
UK 525.71 208 .958 .952 .947 .010 .072 .684
Germany 389.59 208 .989 .996 .994 .047 .021 .600
Greece 395.56 208 .991 .989 .981 .013 .010 .654
Table options
Full-size image (54 K)
Fig. 1.
AFC Graphic solution.
Figure options
The results show that all models have CFI values greater than 0.95 and the ECVI index value does not exceed 0.6. Based on these low values of ECVI and adequacy of CFI values, the suitability of the model for different samples can be assumed. Both these indices are used to measure the comparative fit between two or more models, and the smaller the obtained values the better the fit (Bandalos, 1993). Our study, with the large sample employed and the need to compare more than three groups, is unsuited for the chi square difference test (Browne and Cudeck, 1993 and Satorra and Bentler, 2001). This test has the same limitations as the likelihood ratio test in general, so that very large samples lead to very trivial difference tests. As to the reliability of the instrument, the indices obtained (McDonald’s Omega = .99 and Standardized Cronbach’s alpha = .96) exhibit a suitable over all reliability and also adequate reliability of each of the two factors making up the scale, the cyber-victimization factor with α = .97 and the aggression factor with α = .93.
Finally, for calculating the prevalence of involvement in cyberbullying, subjects were selected on the basis of the above-cited theoretical criterion. Such a selection was developed with the whole sample and with the subsamples comprising each of the participant countries (see Table 4).
Table 4.
Prevalence of cyberbullying.
Country N Victims (N, %) Aggressors (N, %) Bully/Victims (N, %) Total implication (N, %)
Poland 900 55 (6.11%) 61 (6.77%) 36 (4%) 152 (16.88%)
Spain 859 40 (4.65%) 44 (5.12%) 18 (2.09%) 102 (11.87%)
Italy 1430 115 (8.04%) 79 (5.52%) 75 (5.52%) 269 (18.81%)
UK 737 47 (6.37%) 7 (.94%) 15 (2.03%) 69 (9.36%)
Germany 846 35 (4.13%) 58 (6.85%) 26 (3.07%) 119 (14.06%)
Greece 907 92 (10.14%) 71 (7.82%) 56 (6.17%) 219 (24.14%)
Total Sample 5679 384 (6.76%) 320 (5.63%) 226 (3.97%) 930 (16.37%)
