School Attachment: Control for Academic Achievement
Academic achievement comes to mind as, perhaps, the most important threat to the interpretation that the association between school bonding and violent behavior is causal. Given the likely association between school attachment and school achievement, there are a surprisingly small number of studies where the authors control for academic achievement in their models. In one study, the association between school attachment and violence is not statistically significant in a very conservative multivariate model that includes academic achievement (McNulty & Bellair, 2003), but in three other multivariate studies with such a control, school attachment remains significantly, negatively associated with violence (Bellair et al., 2003; Johnson, 1979; Wright & Fitzpatrick, 2006).
School Attachment: Control for Parent Education
Another potential confound is parent education because it is strongly associated with violence in many studies, and it is likely to impact children’s school experience as well. Two of the studies which included an indicator of parent education still found that their measure of school attachment was significantly, negatively associated with violence (Dornbusch et al., 2001 who measured “school connectedness” and Gottfredson et al., 1991 who combined school attachment and commitment). Two others do not supply supportive data, though their reports present only a weak antagonism to the school bond- violence association. Martino et al. (2008) controlled for parent education and found that the association between “low academic orientation” and a high physical trajectory course was in the right direction, but it was not statistically significant. This trajectory analysis does not necessarily negate an association between school attachment and violence overall. McNulty and Bellair (2003) control for “parents college graduates” in their model, but as we have mentioned, their model includes a control for prior wave violence, which is probably overly conservative because the effects of parent education are likely to have appeared by that prior wave. In our own analysis for a separate paper, we found that school attachment was still negatively associated with violence in a multivariate model controlling for parent education (school attachment was also associated with nonviolent offending in a similar model) (Savage & Ellis, 2014).