Defensive egotism and self-esteem: A cross-cultural examination of the dynamics of bullying in middle school

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Abstract

The revised compensation model of aggression posits that bullying is driven by defensive personality, not low self-esteem. Supporting research has failed to distinguish bullies vs. bully-victims (i.e., bullies who are also victims). In three studies with middle school students on two continents, pure bullies and bully-victims scored higher in defensive egotism than pure victims and controls but, unexpectedly, did not differ from each other. Low self-esteem was linked with victimization, high self-esteem with defending victims. Boys were higher than girls in bullying and physical aggression; girls were higher in social than physical aggression. Part of the lack of success of anti-bullying programs may be their failure to accommodate different types of bullies and different forms of bullying.

Keywords:

• Defensive egotism
• self-esteem
• bullying
• compensation model of aggression
• bully-victims

Funding

Study 3 was supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under [grant number R15HD075229].

Acknowledgment

We thank Marlene Sandstrom and two journal-provided reviewers for their constructive suggestions and inputs of previous drafts.

Notes

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

1. In all three studies we evaluated the data for the multivariate assumptions of linearity, multicollinearity, multivariate normality, and homoscedasticity using standard procedures. In multilevel modeling, these assumptions must be assessed for each level 2 unit (i.e., class). Thus, it would be burdensome to report these tests for every class in every study reported. In every case, violations were nonexistent or minimal and uninfluential in the full nested regression model for each study. For example, in Study 1, Variance Inflation Factors (VIF, <10) and Tolerance values (>.25) in the regression analyses were within the recommended range for all but one of the six homeroom classes we used. This provides good evidence that multicollinearity was not influential in the full model. To explore for possible sex differences in Study 1, we effect-coded sex (Boys = 0, Girls = 1) and repeated the hierarchical regression but with sex and all two- and three-way interaction terms included in the analysis. It yielded only one additional significant effect, a main effect for sex, b = .17, t (89.99) = 2.15, p < .04. Boys were rated significantly higher in defensive egotism by their peers (M = 1.03, SD = .40) than girls (M = .91, SD = .37).

2. Once again, to explore for possible effects for sex in Study 2, we effect-coded sex (Boys = 0, Girls = 1) and repeated the hierarchical regression with sex and all two- and three-way interaction terms included. There were no significant main or interaction effects involving sex.

3. We compared the four bully/victim groups on each of the secondary dependent variables in Studies 2 and 3 using both MANOVA and hierarchical linear modeling. The MANOVA proved to be the more conservative of the two procedures in terms of protecting against Type I errors, so we report these findings. All of the contrasts declared significant in Tables 3 and 6 using MANOVA were also declared significant with hierarchical linear modeling.

4. Once more, in Study 3 we effect-coded sex (Boys = 0 and Girls = 1) and repeated the hierarchical regression with sex and all two- and three-way interaction terms included. There were no significant main or interaction effects involving sex.

5. To compare the bullying and victim of bullying rates in Study 2 vs. Study 3, we transformed the scores from Study 2 from a 0–2 scale to a 0–4 scale as used in Study 3. MANOVA yielded F (2, 538) = 3.58, p < .03, η² = .01. Students in Study 2 did not differ in peer-rated bullying (M = .61, SD = .69) compared to students in Study 3 (M = .56, SD = .37), F (1, 539) = 1.58, p = .51, η² = .003. However, Students in Study 2 were rated lower in victim of bullying (M = .45, SD = .56) than students in Study 3 (M = .53, SD = .35), F (1, 539) = 3.90, p < .05, η² = .007. This finding should be interpreted cautiously, however, given that in Study 2 the average ratings were based on all members of a class, whereas in Study 3 they were based only on students enrolled in the study.