III. Empirical Evidence
As noted above, some key theoretical propositions in criminological research assert different degrees of specialization in offending. This has given rise to a fair amount of empirical research seeking to examine the presence or absence of such properties of offending. These studies have drawn on a variety of methods, and in many cases their differences may be partly attributable to alternative approaches to measurement or analysis.
Most studies tend to find only marginal levels of specialization in offending. Gibbons (1988) suggests that the typological schemes offered by criminologists and practitioners tend to lack support when attempts are made to identify cases that provide a “real world” match to their expectations. In general, the literature on this aspect of the criminal career tends to indicate fairly low levels of specialization among offenders (Piquero, Farrington, & Blumstein, 2003). Relevant early research, for example, found that having information about the number and type of an offender’s previous crimes was not a good predictor of later offenses (Wolfgang et al., 1972).
Farrington, Snyder, and Finnegan (1988) studied the question of specialization in a sample of juvenile offender court records. They found an average coefficient value of .10, which was statistically significant, but quite small relative to the value of 1.0 that would be expected in the case of complete specialization in offending. They also compared the actual number of specialists for each offense to that expected by chance and found that only about 20% could be designated as specialists. While identifying somewhat higher levels of specialization in an adult sample, coefficient values identified by Blumstein et al. (1988) were all below .50 and were generally closer to .20. As in these studies, Paternoster et al. (1998) identified fairly low levels of specialization in juvenile and adults when they analyzed crime patterns for a sample of British offenders.
Using the Philadelphia Birth Cohort data set, Kempf-Leonard (1987) examined five measures of specialization (e.g., specialization coefficient, transition probabilities) and found that, in general, offenders exhibited low levels of specialized criminal activity. Piquero and colleagues (Piquero, Paternoster, Brame, Mazerolle, & Dean, 1999) and Mazerolle and colleagues (Mazerolle, Brame, Paternoster, Piquero, & Dean, 2000) examined the relationship between specialization and versatility of offending and key covariates (e.g., age of onset, gender). Mazerolle et al. calculated diversity index values across three crime types and found observed values ranging from about .28 to .50, depending on the number of offenses, suggesting fairly diverse offending profiles. Piquero et al. utilized several measures (e.g., diversity index, transition matrices) and similarly found fairly low degrees of specialization. For example, the specialization coefficients were below .35 in all cases (with 1.0 representing complete specialization). Focusing on violent specialization, Piquero (2000) likewise found no discernable evidence that certain offenders tend to concentrate on such offenses and suggested that individual volume of offending is the primary correlate of whether offenders are violent or nonviolent.
Some recent evidence suggests that the question of specialization in offending may be more nuanced than first thought. For example, in work drawing on ethnographic accounts, Shover (1996) found that, in contrast to specialization over the entire criminal career, offenders make short-term adjustments in offense patterns that suggest some degree of temporary specialization. Sullivan, McGloin, Pratt, and Piquero (2006) report a similar finding with the diversity index in a sample of incarcerated felons when studying short pockets of time (e.g., months and years). Francis et al. (2004) also found evidence of offender “types” when studying longitudinal crime patterns in a U.K. offender sample. Like Shover, both of these studies also suggested that criminals may shift their offending preferences over time, which would aggregate to versatility over a career.
Using a novel measurement and analytic approach, Osgood and Schreck (2007) report clear and consistent evidence of specialization in violent offending. Their study incorporated self-reported delinquency measures across three prominent studies of juvenile delinquency (e.g., Monitoring the Future). They also found that the propensity to specialize in violent behavior was relatively consistent over time and was systematically related to some expected correlates (e.g., risk seeking). Thus, they assert fairly strong evidence of specialization using their approach.
It is also important to consider different correlates of offending patterns that might be important in connecting this aspect of the criminal career to theoretical propositions and policy considerations. Piquero et al. (1999) and Mazerolle et al. (2000), for instance, attempted to tie the degree of specialization to important aspects of the broader criminal career. Lynam, Piquero, and Moffitt (2004) looked at specialized offender groups and found some differences in the extent of their respective developmental risk factors (e.g., childhood conduct).
Although these findings do suggest some degree of specialization in offending, it is important that they are viewed in the context of a larger group of studies that have identified offenders who tend to commit a variety of different criminal acts over their careers. Britt (1994) suggests that researchers sometimes lose sight of the fact that even studies identifying purported specialized offenders tend to suggest that they represent a fairly small portion of the overall pool. So, in considering empirical results, it is important to keep the findings of generality in criminal careers in mind but also consider potential segments of that career where there might be specialized pockets of offending.