Mixed Results on “Offense Specialization”
Another reason we believe that the field has not emphasized a search for theories of violence is the conclusion, reached by numerous prominent authors, that “offenders do not specialize.” Klein (1984) reviewed the literature and examined his own data on gang members and drew the analogy of “cafeteria-style delinquency” and concluded that “The evidence is very supportive of a general delinquency factor and extremely weak for offence specialisation [sic],.. .” (p. 191). Farrington (1996) also concluded that criminals do not specialize. In recent years, life-course studies have emphasized offender versatility, based on Moffitt’s (1993) prediction that early onset offenders would be likely to commit many types of crime, including violence. Findings by Mazerolle and colleagues, who analyzed data from the Philadelphia birth cohort study, support this prediction (Mazerolle, Brame, Paternoster, Piquero, & Dean, 2000). If offenders are versatile, and violent offending is merely symptomatic of long-term frequent offending, there is no need for an additional theory to explain it. Deane et al. (2005) summarize the feelings of many authors who have studied offense specialization: most offenders are versatile, and “Extensive versatility in offending is consistent with the notion that general processes cause all forms of criminal behaviors” (Deane et al., 2005, p. 958).
We find that studies where broader definitions of crime types have been employed to test “offense specialization” provide plenty of evidence that violent criminality is not randomly distributed among offenders. Bursik (1980) found evidence for “specialization” of property offenses, and among personal injury offenses for some in his sample. Kempf (1987) also found greater specialization in property versus other types of offenses. In an analysis of data from a large Danish birth cohort, Brennan, Mednick, and John (1989) conclude that “specialization” in violence exists for offenders with more than three arrests. The authors emphasize the utility of knowledge of past violence for the prediction of future violence.
Others who have tested different versions of our own research question have concluded that a unidimensional criminality is not supported by the evidence. Rebellon and Waldman (2003) present a confirmatory factor analysis and decide that “force” and “fraud” cannot be viewed as “manifestations of a single underlying construct” (p. 303). A factor analysis by Thornton, Graham-Kevan, and Archer (2013), using British data, also returns separate factors for violent and nonviolent offenses. Deane et al. (2005) analyzed nine behaviors and “show clearly that violent offenders are more likely to engage in additional violent offenses, nonviolent offenders are more likely to engage in additional nonviolent offenses” (p. 983). Osgood and Schreck (2007) propose a carefully elaborated analytic method for testing offense specialization and, based on their analysis of three data sets, conclude that evidence exists for “substantial levels” (p. 274) of specialization in violence and considerable stability over time.
We assert, however, that it is not necessary that criminals need specialize in violence for there to be a differential etiology of violence. Just because a violent offender commits many crimes does not mean that all offenders who commit many crimes are equally likely to commit a violent one. Further, narrow definitions of offense specialization, for example an empirical test that finds no significant association between committing a robbery and the next offense being a robbery as well, may mislead us into missing broader patterns of violent behavior. The methodological technicalities have made this literature less relevant than it might be for our research question. We have no plan to distinguish between Person A, who has committed three violent offenses, from Person B, who has committed three violent offenses and six nonviolent ones. For those of us interested in the causes of violence and desirous of preventing violence, those two individuals are both of equal interest.