Undifferentiated Drug Use
Many studies use an undifferentiated category of “drug use” to test associations with violence. Due to conflicting expectations about the impact of various drugs on behavior, one might predict that associations between undifferentiated measures of “drug use” and violence would be weaker or less consistent than studies of specific drugs thought to have pharmacological effects that cause violence.
Table 11.1 shows that about one-third of the coefficients reported across studies were positive and statistically significant for violent offending and that a larger proportion (53%) were positive and statistically significant for nonviolent offending. We see in Table 11.2 that this was not due to outlier studies. At the study level, while 25% reported a PoC in the predicted direction, a number that is not particularly large, that figure was 70% in studies of nonviolent offending. Thus, there are fairly consistent correlations between drug use and crime, and this finding is more consistent for nonviolent than violent crime.
Quite to the contrary, studies that compare offenders usually report that violent offenders have more problems with drugs, in general, than nonviolent offenders. Kennedy (2006) reported that substance abuse was more prevalent among violent than nonviolent offenders. Loeber et al. (2005) report that persistent drug use was more common among their violent offenders. Levi et al. (2010) found that drug use was high among both predatory and “irritable” violent Canadian inmates. In a sample of parolees, Sacks et al. (2009) compared the intake profiles of violent offenders with those with no record of violence. While the likelihood they had some form of drug arrest history was similar, the violent offenders reported a greater frequency of substance use across many types of substances, though they did not report having undergone more treatment. In another study, this one looking at data from probationers in rural Kentucky, the prevalence of “ever using” drugs of many different types was significantly higher among the violent than nonviolent offenders (Webster et al., 2010). Haapasalo and Hamalainen (1996) report that substance abuse and dependence was lower among property than violent offenders in Finland.
There were just two exceptions to this pattern of supportive findings among the offender studies. In a sample of minority adolescents in the Netherlands, Colins et al. (2009) report that property offenders had a much higher prevalence of both alcohol and drug use disorders than violent offenders. Valdez et al. (1997) report that “nonaggressive” arrestees in Texas were more likely to test positive for substance use than violent offenders. This was true in a multivariate model controlling for alcohol use, amount of money spent on drugs last week, age, education, and income (legal and illegal). We wonder if the high percentage of positive drug tests among the “nonaggressive” arrestees (63%) was due to arrests for drug crime.
Turning to dual dependent variable studies of marijuana consumption, we find that marijuana use was positively and significantly related to violent crime in 54% of models and positively and significantly associated with nonviolent crime in 65%. At the study level, the discrepancy is larger; 36% of studies report a PoC in the expected direction for violent crime, while 57% did so for nonviolent crime. Although marijuana appears to be consistently associated with both violent and nonviolent crime, that association is more frequently observed in studies of nonviolent offending, which is evidence that marijuana is not a differential predictor of violent crime.
We were surprised to learn, then, that offender studies suggest that marijuana use is more commonly reported among violent offenders than nonviolent offenders. In a Kentucky study, a higher prevalence of violent probationers had ever used marijuana than nonviolent probationers (Webster et al., 2010). In a Finnish study, the prevalence of cannabis dependence was lower among property offenders than violent offenders, and abuse and dependence on other “street drugs” was significantly lower as well (Haapasalo & Hamalainen, 1996). In one exception, Colins et al. (2009) report that property offenders had much higher prevalence of both marijuana use disorders than violent offenders.
In a very interesting study, Martin and Bryant (2001) found that a dirty test for marijuana was positively and significantly associated with violence in a sample of female and male offenders. Their model was unique as it controlled for alcohol use, cocaine use and other drug use (Martin & Bryant, 2001). The coefficients in the model pointed to marijuana, rather than other types of drugs, as a risk factor for violence.
We might expect, based on decades of hyperbolic rhetoric in the media, that cocaine intoxication would be a strong contender as a differential predictor of violent crime. Oddly, no studies really distill that effect. We also might expect that cocaine use would be correlated with nonviolent crime in the form of drug dealing. Many authors have asserted that illegal drugs are more strongly related to property crime than violence (e.g., Harrison, 1992; Parker & Auerhahn, 1998; Watters et al., 1985). Table 11.1 shows that cocaine use (combining the “cocaine” and crack categories) was positively and significantly related to violence in 11 out of 22 models (50%), and negatively related to violence in 2 out of 22 models (9%). By comparison, combined cocaine consumption was positively related to nonviolent crime in 17 out of 40 models (43%). At the study level, we see the same, slight disproportion (see Table 11.2), but we deem that difference inconclusive.
When we look at studies that specify crack cocaine, we see that 44% of the coefficients indicate a significant association with violence, but only 27% of the coefficients for nonviolent offending were positive and statistically significant. However, at the study level, 2 out of 3 studies reported a PoC that was positive and statistically significant for both violent and nonviolent offending, so the disproportion seen at the comparison level is likely due to a high number of comparisons being reported by one or two studies.
In just two studies, offender cocaine use was compared between violent and nonviolent offenders. In Kentucky, a higher prevalence of violent offenders reported having used powder cocaine and crack cocaine than nonviolent offenders (Webster et al., 2010). In the study by Martin and Bryant (2001), which we highlighted above, being under the influence of cocaine was significantly negatively associated with violence. Recall that this study pits different types of drugs against each other in the same model. Because drug dealing is a form of nonviolent crime, and use of drugs is likely to be correlated with it, estimates of associations between cocaine use and nonviolent crime are likely to confound possible pharmacological influences of the drug with the noncausal correlation between drug dealing and arrest.
Taken together, the findings do not provide evidence that cocaine use, even crack use, has a greater impact on violent than nonviolent offending. This conclusion is based on only a small number of studies, and estimates between cocaine use and nonviolent offending are likely to be elevated, so a firm conclusion cannot be forwarded.