An investigation into the fundamental regularities of co-offending for violent and property crime classifications.

• Author: Andresen, Martin A.
• Position: Canada

Introduction

Co-offending is a term coined by Albert Reiss (1980). This criminal phenomenon refers to offences that are committed with the simultaneous presence of more than one offender. (2) It deserves special research attention because of its extra consequences: each co-offence generates more cases for the criminal justice system than each solo-offence, those involved in co-offences commit more subsequent offences more frequently and at more serious levels (Carrington 2002; Hindelang 1976; Sarnecki 2001), (3) and early co-offenders tend to have more serious and more violent criminal careers (Conway and McCord 2002; McCord and Conway 2002). Moreover, there is reason to believe that co-offending causes extra harm to youth, ethnic minorities, and urban centres (Felson 2003).

Because of the data-intensive nature of co-offending research, (4) much of the classic research on co-offending has been based on relatively small samples. Despite this fact, in his famous review of the co-offending literature, Reiss (1988) reported on two fundamental empirical regularities across studies: co-offending is a youth phenomenon (co-offending declines as youth age into their 20s and beyond), and the number of co-offenders involved in each offence decreases with age--there are, of course, other regularities (see Reiss 1988; Felson 2003).

From a theoretical perspective, these fundamental empirical regularities may be explained using a developmental approach. This approach is common in the co-offending literature. A developmental approach recognizes that youth carry out most of their activities (legitimate and illegitimate) in groups. As youth age, however, the group size decreases. Ultimately, the "final group" is most often a pair-bond. Within the developmental approach, crime is simply one of the activities carried out by youth; as youth age, co-offending and criminal group size decreases, just as with other activities (Carrington 2002; Reiss 1988; Reiss and Farrington 1991; Warr 1996). The question is, 'Does the developmental approach explain the fundamental empirical regularities of co-offending across crime types?'

With sample sizes being relatively small in much of the previous literature, little could be known regarding these empirical regularities for specific crime classifications. Do violent crimes have the same empirical regularities as property crimes? If so, do all violent crimes follow the same pattern? Carrington (2002) was the first study to analyse co-offending using a very large incident-based data set. His Canadian data set (UCR2) includes 2.9 million criminal incidents and 3.4 million alleged offenders, 1992-1999. Because of this large number of observations, he was able to disaggregate data by single year of age (8 to 89) and detailed crime classifications. In his analysis, Carrington (2002) not only confirms--for the first time with a large incident-based data set the two fundamental empirical regularities of co-offending but also shows that the strength of the relationship between the two regularities, measured using Pearson's correlation coefficient, is extremely high, r = 0.997.

In an investigation of specific crime classifications, Carrington (2002) also finds that the percentage of co-offending for some violent crime classifications (assault and sexual assault, level 1) is low when compared to that for most property crimes. Erickson (1971) and Hindelang (1976) also report this result but are based on relatively small samples. However, co-offending is still higher for youth and declines with age. In two more recent analyses also using large incident-based data sets, violent crimes tended to have less co-offending than property crimes (van Mastrigt and Farrington 2009; Carrington 2009), and the mean number of offenders per criminal incident was quite similar across crime classifications (van Mastrigt and Farrington 2009). However, neither of these two studies broke down the individual crime classifications by age--this was not the focus of either of these two research studies.

In this article, we contribute to the co-offending literature through an investigation of the two fundamental empirical regularities disaggregated by crime classification and single year of age. For the individual crime classifications, we focus on those aged 12-29 because there is little change before and after this age range, (5) but the age range of 8-68 is used for a comparison to previous research (Carrington 2002). We find that co-offending...