The Differential Etiology of Violence
If parental warmth or rejection contribute to a differential etiology of violence, we expect at least one of the following patterns to emerge in the evidence. First, we might expect that these constructs are associated with violent, but not nonviolent, antisocial behavior. It could also be the case that, in studies where both outcomes are reported, the coefficients estimating the association with violent outcomes is larger than the coefficient for nonviolent outcomes. Apart from this, if we see that violent offenders have experienced less warmth or more rejection than nonviolent offenders, this could also be seen as evidence of a differential impact on violence.
Looking at Table 7.1, the rows displaying effects of parental warmth, maternal warmth, and paternal warmth, we see that warmth is associated in the predicted direction with nonviolent offending in most cases. The consistency of these associations is similar to the consistency in studies of violence. The possible exception is that the evidence is divided in studies where “parental warmth” is used to predict nonviolent crime; however, these are small in number (k = 4). At this point, we conclude that an analysis of dual-dependent variables, violent and nonviolent offending, indicates that parental warmth is not consistently predicting one over the other.
The pattern of findings on parental rejection appear to be more supportive of the differential etiology hypothesis, though the small number of studies of nonviolent offending makes this conclusion tentative. While a strong majority of studies (8 of 12) reported a statistically significant PoC for violent crime, only 3 out of 5 did so for nonviolent offending, and 2 studies of nonviolent offending reported a PoC in the opposite direction of the hypothesized relationship.
Looking within studies that provide separate analyses for violent and nonviolent offending, five studies report essentially the same pattern of findings for violent behavior and nonviolent offending (Barnow et al., 2005; Brannigan et al., 2002; McCord, 1996; Shekarkhar & Gibson, 2011; Veneziano, 2003) and one reports a pattern where the association between warmth is consistently significant for nonviolent offending but not violent offending (Scholte et al., 2007). Three studies of parental rejection report the same pattern of findings for violent and nonviolent offending, analyzed separately (Barnow et al., 2005; Brannigan et al., 2002; Brennan et al., 1999). When we compare coefficients, we observe no differences in magnitude, and no indication that formal tests would uncover any. Thus, the overall findings from dual-dependent-variable studies do not indicate that parental warmth is a differential predictor of violence.
The findings provided by Tyler et al. (2011) deserve to be highlighted. They tested our research question most directly. They controlled for general delinquency in their model of violence and found that parental warmth still had a significant direct effect on violence.
Nonetheless, in 3 out of 4 studies comparing violent to nonviolent offenders (not displayed due to the small number divided across rows), the violent offenders reported lower parental warmth or greater parental rejection than nonviolent offenders did (McCord, 1996; Tupin et al., 1973; Walsh et al., 1987). McCord (1996) was very interested in this outcome and emphasized it across her various works on delinquency. She compared “low parental affection” across violent and nonviolent criminals, and her table suggests that while nonviolent criminals reported much lower average parental affection and maternal warmth than noncriminals, the violent offenders reported the lowest of all with parental affection ratings that were less than half the average of those for the nonviolent offenders. Two studies compared rejection between violent and nonviolent offenders and both suggest that violent offenders rate their parents as more rejecting than nonviolent offenders do.