Sexual abuse has been associated with violent behavior in many studies. Savage et al. (2014) recently reported that retrospectively measured sexual abuse victimization was significantly associated with violent offending in a conservative model controlling for a wide variety of factors. In that analysis, the association between sex abuse and violence was of greater magnitude than the association between physical abuse and violence.
In this review, we confirm the familiar pattern. The PoC in 8 out of 9 studies suggests that sexually abused individuals commit more violence than those not abused. The association with nonviolent offending is also consistent, however (6 of 8 studies). In 6 studies, the sexual abuse victimization of violent and nonviolent offenders has been compared. While almost all of these studies suggest that violent offenders are more likely to have endured sexual abuse (k = 5 out of 6), the PoC statistically significant in just one (Hughes et al., 2009). Because of methodological concerns related to some of the estimates, including small sample size and unusual findings reported across the board, more studies would be needed for confident conclusions.
Among studies where the authors report coefficients for associations between sexual victimization and violent and nonviolent offending separately, the majority report similar patterns of findings for violent and nonviolent offending (Benda, 2005; Burgess, Hartman, & McCormack, 1987; Goodkind, Ng, & Sarri, 2006; Herrera & McCloskey, 2003; McClintic, 2003; Swanston et al., 2003; Zingraff et al., 1993). In studies of violent behavior among offenders, where nonviolent offending is implicitly controlled, the findings are mainly in the right direction, but not statistically significant.
There are several studies that report a pattern of findings consistent with the hypothesis that sexual abuse victimization has a special relationship with violence. Cusick et al. (2012) report a significant effect of sexual abuse on violence, but a negative effect on nonviolent offending. Siegel and Williams (2003) report a significant coefficient in the right direction for violence, and a coefficient that is not significant for nonviolent offending. (Their sample consisted of all females.) McDaniels (1998) found that most coefficients estimating the association between sexual abuse and violence were in the right direction, but most estimating the association with nonviolent offending were in the opposing direction.
In this particular area, we find that female offenders are studied almost as much as males. Byrd and Davis (2009) reported that, among female offenders, violence was less common among those who were sexually abused. Goodkind et al. (2006) looked at violent and nonviolent offending among female offenders. We see nothing in the pattern of findings that suggests a special association between sexual abuse and violence and draw the same conclusion about findings reported by Herrera and McCloskey (2003) and McDaniels (1998). Pollock et al. (2006) provide a test very close to that demanded by our hypothesis and compare violent to nonviolent female offenders. They reduced the chance of bias by giving every entering inmate a chance to participate in the study. The simple difference in prevalence of sexual abuse between violent and nonviolent offenders was statistically significant, but in the multivariate model, it was no longer significant, though in remained in the predicted direction. The sample size was 657, so statistical power was high. The control variables included childhood physical abuse and adult physical and sexual victimization, so we believe the model may have created redundancies for the purposes of estimating the effect we are interested in and is likely to underestimate that effect. Siegel and Williams (2003) report the only study where the data display a pattern where sexual abuse is differentially related to violence. In 4 out of 4 comparisons, sexual abuse was associated with violence (and this was supported in a supplementary multivariate model), and this was true in only 2 of 4 comparisons where nonviolent criminal behavior was the outcome. In summary, the evidence does not yet support the proposition that sexual abuse victimization is a differential predictor of violence, but we retain it as a prospect due to strong evidence of an association with violence, and methodological in the contradictory studies.