Control for Socioeconomic Status
As happens in many analyses where control variables are entered into models, the associations between intellectual measures and violence are not as consistent in models controlling for socioeconomic status (SES), though we emphasize that in all cases, authors simultaneously added controls for other factors as well. There are enough studies where authors have reported that the associations between IQ and violence withstood a control for SES for us to conclude that the association is not fully mediated by socioeconomic factors (e.g., Bellair & McNulty, 2010; Loeber et al., 2012; NICHD, 2004; Piquero, 2000; Waldorf, 1997; Yeomans, 1996; Zagar, Arbit, Sylvies, Busch, & Hughes, 1990). However, in numerous studies, the associations have been in the predicted direction but not strong. We believe, however, that most of the nonsupportive multivariate studies have used “overspecified” models. Bauer (2000) matched his subjects on SES and reports that violent adolescents had lower scores on EF tests than nonviolent, conduct-disordered adolescents, but the differences are not statistically significant. Brownlie et al. (2004) report that the association between “language impairment,” and not verbal IQ withstood the imposition of this control variable (though we would not emphasize this finding because both of these independent variables were in the model at the same time). Bellair and McNulty (2005) also controlled for SES and did not find an association between the Peabody Picture Vocabulary test and later violence, but their model included a measure of prior violence (thus partial- ling out all of the impact of verbal ability on violent behavior as of last year). Walsh and Petee (1987) found that the significant association between IQ and violence disappeared in their stepwise, multivariate model where they added a measure of social class, but also a measure of “love deprivation.” Using the same sample, Walsh et al. (1987) found that deviation scores (PIQ—VIQ) were still associated in a model controlling for social class.
We conclude that controlling for SES is a good practice for studying the association between intelligence and violence but that SES does not fully account for that association.