The social context of school bullying: evidence from a survey of children in South Wales

Abstract

The article presents a descriptive review of the various patterns of association with school bullying that were revealed in a survey of over 26,000 children aged 11–16 in South Wales. The survey examined risks and protective factors for young people and included a question about being a bully and another about being bullied. Following regression analyses, significant associations were found between being a bully and various attitudinal and behavioural factors. There was a modest but ambiguous association with socio‐economic status, and no independent association with minority ethnicity. There was a strong independent association between being a bully and being bullied. Some comment is made about the implications of the findings for policy, including the apparently positive effect (as revealed in the survey data) of schools being perceived by children as having clear rules on bullying.

Keywords:

• bullying
• children
• secondary school
• secondary analysis
• survey

Acknowledgements

We acknowledge the assistance of Communities that Care Cymru, who conducted the survey and offered us this data set for further analysis. In particular we acknowledge the help and encouragement of Anne Fairnington, the former Director of CTC Cymru, Pat Dunmore, its current Director, and Stephanie Lee the former research manager for CTC GB. We are also grateful to Peter Smith for his advice on an earlier version of this article.

Notes

1. Throughout this text responses to these variables are described as ‘reports of being bullied’ and ‘reports of being a bully’. It is important to remember that this is a slight simplification of the original question format, and that not all respondents who gave positive answers to these questions may have been thinking in the more pejorative terms used here.

2. In fact, several alternative indicator variables may have been constructed from these responses. In many social science applications, it is believed that the substantive impact of alternative representations of such Likert‐style data are often minimal (cf. Bryman and Cramer 2005:71). Our own analyses of patterns of correlation and association between different representations of the same responses to these variables revealed strong consistency regardless of the alternative representations (a table of results is available from the authors on request). Additionally, Appendix Table A2 illustrates that the results from regression models which used different representations of these Likert variables as responses were similarly unaltered in any substantial way by the alternative functional forms.

3. In fact, reports of having ‘seen a pupil attack a teacher’ were disproportionately concentrated amongst respondents from Swansea.

4. School exclusion was also related to other differences between respondents – for instance, with females much less likely to report exclusion. However, later analyses suggest that exclusion does indeed have a strong independent association with bullying behaviour.

5. It should be noted that the magnitudes of the coefficient estimates are not strictly comparable between models. However, their sign, size relative to their standard errors, and size relative to each other can be compared. In this regard Table A2 suggests that all five models display similar overall patterns of explanatory variable effects, but with the binary response models showing greater differentiation from the other formats.