MULTIVARIATE MODELS: ACADEMIC ACHIEVEMENT
The association between academic achievement and violence has been robust across models controlling for a wide variety of factors. In analyses of Add Health data presented by Bellair and colleagues, school grades are negatively associated with violence in multivariate models controlling for important potential confounds such as use of alcohol and drugs, and reducing the likelihood of a causal order problem by controlling for prior violence (Bellair & McNulty, 2005); a later analysis also controls for a series of employment factors, family income, school attachment. and family bonds (Bellair, Roscigo, & McNulty, 2003). McNulty and Bellair (2003), who also used Add Health data (waves 1 and 2), found that school grades were significantly negatively associated with serious adolescent violence across all multivariate models (controlling for demographics, substance and alcohol use, family structure variables, community factors, concentrated disadvantage, residential stability, family and school bonding, “interacts with neighbors,” gang membership and exposure to violence). In another analysis of Add Health data, Wright and Fitzpatrick (2006) found that violence was significantly negatively associated with grade average in models controlling for neighborhood income, neighborhood racial distribution, school affiliation, and participation in sports and clubs.
The effect of academic achievement on violence has been robust in multivariate models using other large data sets as well (e.g., Loveland, Lounsbury, Welsh, & Buboltz, 2007; Owens-Sabir, 2007). This includes a study of 10th grade Icelandic adolescents (Bernburg & Thorlindsson, 1999, controlling for parenting factors such as rule setting, monitoring and support, and peer drug use) an Ottawa cohort (Brownlie et al., 2004; controlling for language impairment, verbal IQ, parent conviction and parent distress, among other factors) a sample of several thousand adolescents in Oregon and California (Ellickson & McGuigan, 2000; controlling for numerous factors including peer drug use, personality factors like self-esteem and rebelliousness, and neighborhood SES), the Seattle Social Development Project (Herrenkohl et al., 2001; Kosterman et al., 2001; controls for a series of family problems and peer factors) the National Youth Survey (NYS) (controlling for abuse experience, parent divorce, quality, time with family, importance placed on education, time studying, and other factors; Rebellon
& van Gundy, 2005) and a sample of 3,761 adolescent males from three locations (Salts et al., 1995; controls for age and family structure and an index of individual characteristics, along with problem school behavior, and drug and alcohol use).