Violent vs. Nonviolent Offenders: Quality Control
Comparisons of violent and nonviolent offenders may be flawed in several ways. Because the groups are not randomly assigned, there may be systematic differences between them that do not stem from a differential etiology of violence. For example, the type of offender who ends up in prison for a nonviolent crime is not likely to be representative of nonviolent offenders as a whole. We would expect, for example, that those who end up in prison in the United States are more likely to be involved in the drug trade and to be heavy drug users; we would also expect that those who are imprisoned for nonviolent offenses have a greater number of recorded offenses than typical nonviolent offenders. Meanwhile, many offenders who commit a serious violent offense are imprisoned without a prior record, and their drug involvement is less likely to be integral to their case. Because these differences are due to enforcement and punishment practices, not the etiology of crime, they create a potential for bias when we compare violent and nonviolent offenders.
An important comment to make on this point, which is central to the present chapter, is that differences between violent and nonviolent offenders on measures of intelligence could also be related to criminal justice practices. Even if violent offenders overall have lower IQ than nonviolent-only offenders, it could be the case that imprisoned nonviolent offenders have lower intelligence than those who are not caught. Clearance for nonviolent offenses in the United States, even for serious ones such as burglary, is quite low, and the vast majority of offenders are not caught. Of course, it may also be the case that “caught” violent offenders have lower intelligence than uncaught ones, but because police spend more resources solving violent crimes, the influence of this source of bias in estimating the intellectual abilities of “violent offenders” as a whole is likely to be smaller. If this is true, then violent and nonviolent offenders who are in prison may have more similar IQ scores than the population of violent and nonviolent offenders would have if we could find a representative sample.
In addition, some studies do not report the method they used to obtain the offender sample. If there are differences in the way violent and nonviolent offenders are recruited for participation, or in their willingness to volunteer, then selection bias may influence estimates of coefficients. If, for example, the smartest violent offenders are less willing to participate in research than other inmates, due to their comprehension of the gravity of their crime and the associated shame, we can see how the differences in IQ between the violent offenders in the study (whose average IQ will be lower if the highest-IQ offenders have not volunteered), and the nonviolent offenders would be reduced. We can only speculate about this problem; we have not seen any published studies about it. The best practice is to try, with some vigor, to recruit random samples from the populations of violent and nonviolent offenders. This will not eliminate the problem, but it is the best way to minimize it when offender samples are used.
Finally, many of those who have been arrested or imprisoned for nonviolent crimes are violent people who just happened to be caught, most recently, for a nonviolent offense. Thus, studies that use the instant offense to operationalize “violent” and “nonviolent” are likely to have a muddy nonviolent group that includes some violent individuals. In this case, particularly when the sample size is small, we imagine that differences between groups in average IQ scores are also likely to be small. The best approach is to use the offenders’ full criminal records, or to use an offender sample, and analyze correlates of self-report violence. This approach effectively controls for the presence of nonviolent offending.
The vast majority of comparisons of violent to nonviolent offenders that we reviewed did not ensure that the nonviolent offenders had no previous history of violence, so we focused on those that made an effort to remedy this problem. Overall, the data are still supportive, with a few very supportive studies, and several that provide modest, mixed support. Syverson and Romney (1985) also made sure that the nonviolent comparison inmates had no record of a violent offense. In this case, all measures of intellectual ability were significantly lower in their violent offender group. Levi, Nussbaum, and Rich (2010) classified offenders based on criminal history, rather than instant offense. They found that violent offenders had lower scores on the vast majority of a wide array of measures of EF. Studies by Walsh (1987) looked at IQ as a correlate of violent offending in a group of offenders, and this design effectively controls nonviolent offending. He found that IQ was negatively associated with violent crime but not property crime. In a later study, using a different sample and similar modeling, Walsh et al. (2004) report significant negative associations between IQ and violence. Mixed support has been provided by Barker et al. (2007); Cornell and Wilson (1992); Hancock, Tapscott, and Hoaken (2010); and Miura (2009).