Co-Offending and the Diversification of Crime Types

Abstract

There is theoretical and empirical support for co-offending being important not only for understanding current offending but also subsequent offending. The fundamental question is—why? In this article, an aggregate analysis is performed that begins to answer this question. Disaggregating solo- and co-offending by single year of age (12-29 years) and crime type in a largely metropolitan data set from British Columbia, Canada, 2002 to 2006, it is shown that the distribution of co-offences is significantly more varied than the distribution of solo offences. This more varied distribution of co-offences favors property crimes during youth but fades as offenders age.

Keywords

co-offending group crime diversification

Introduction

Co-offending, a term coined by Albert Reiss (1980), refers to offences that are committed with the simultaneous presence of more than one offender. Co-offending is very important for crime, especially during adolescence, and might help us understand how crime emerges in the context of daily life. Most co-offending occurs in very small groups, often consisting of two or three youths. Although the term group offending is often used synonymously,¹ it is a mistake to assume that offender groups are necessarily juvenile gangs. Indeed, group offending is normal among nongang youths and a substantial number of nongang adult offenders. Even aside from the relative importance of co-offending, it has extra consequences, thus meriting special research attention. Not only does each co-offence get more people in trouble and generate more cases for the justice system, but co-offending can also easily set the stage for subsequent crime and delinquency (Felson, 2003).

Empirically, scholars find that those involved in co-offences commit more subsequent offences, more frequently and at more serious levels (Hindelang, 1976; Felson, 2003; Sarnecki, 2001; Warr, 2002; Zimring, 1981). More specifically, early co-offenders tend to have a more serious and more violent criminal “career” (Conway & McCord, 2002; McCord & Conway, 2002; Zimring, 1981); “nonviolent offenders” who commit their first co-offence with accomplices who are violent have an increased risk for committing subsequent serious violent crimes (Conway & McCord, 2002), and more serious offences are more likely to be committed by groups (Carrington, 2002). However, Carrington (2002) states that concluding more serious offences are co-offences is an oversimplification; property offences, often considered less serious than violent offences, can have co-offending rates that are significantly higher than violent offences, particularly for offences such as aggravated assault and sexual assault—see Erickson (1971), Hindelang (1976), Reiss (1980), Sarnecki (1986), and Reiss and Farrington (1991) that also show inconsistencies with this generality.

Regardless of these inconsistencies, there is substantial support for co-offending having an impact on offending patterns, more generally. The question, of course, is—why?² This may be explained through a discussion of activity and awareness spaces. Brantingham and Brantingham (1981) use the geometric theory of crime to explain expanded search behavior of offenders that have accomplices—see Figure 1.11 in page 47 in Brantingham and Brantingham. Simply put, if one offender “introduces” another offender to different areas and its respective targets, the second offender will incorporate (at least some of) that new area into his or her activity and awareness space. If one extends the term awareness beyond its geographical context to the awareness of new types of opportunities as well as new areas of opportunities, co-offenders are expected to be more diversified in their criminal activities than solo offenders. McGloin and Piquero (2010) refer to this as nonredundant networks.³

In the context of this article, more diversified refers to co-offenders being distinguished from solo offenders by performing a greater variety of crime classifications. The analysis below uses aggregated level data. As such, it is expected that the set of criminal incidents that involve co-offending are more varied than the set of criminal incidents that involve solo offending: The percentages of co-offending crime classifications are more evenly distributed than the percentages of solo offending. In this article, the diversification of solo- and co-offences is investigated using a large and comprehensive data set that allows for a detailed breakdown of offences by age and offence type. After reproducing previous results to show that there are no peculiarities in our data, it is shown that the distribution of offences for co-offences and solo offences are different. Specifically, it is shown that co-offending is indeed far more diversified than solo offending.

Background on Co-Offending

For nearly a century, since Breckenridge and Abbott (1912) noted the frequency that youth offend together, co-offending has been acknowledged as an important aspect in the etiology of crime. However, as noted by Reiss (1988), most of the literature on this topic is based on sample sizes too small to differentiate types of crime and ages of offenders. In spite of this limitation regarding early work on co-offending, Reiss identified two fundamental empirical regularities of co-offending: It is a youth phenomenon and it declines with age. The earliest studies of co-offending show that co-offending represents a substantial percentage of youth offending: 82% of youth offences in Chicago (Shaw & McKay, 1931); 75% in Flint, Michigan (Gold, 1970); and 85%, based on a review of 11 other empirical analyses (Erickson, 1971). Needless to say, there is substantial evidence that offending, particularly youth offending, is a group phenomenon.⁴

In addition to this early evidence, Felson (2003) argues that co-offending indirectly generates additional crimes and does harm beyond itself: Each co-offence generates additional cases for the criminal justice system and, as stated above, sets the stage for more (serious and violent) crime and delinquency. Felson speculates that given the impact of co-offending on later offending, the total direct and indirect percentage of co-offending is upwards of 80%, without multiple counting.⁵ Needless to say, whether co-offending consists of 50% or 80% of offences (even without multiple counting), understanding the characteristics of co-offending is important not only for the criminal justice system but also for studying the etiology of crime.

Probably, the most convincing empirical analysis of co-offending regularities is Carrington (2002). In studying Canadian incident-based data, Carrington produced the first research paper to analyze co-offending using a large incident-based data set (Uniform Crime Reporting Survey 2 [UCR2]) covering offenders at virtually all ages. Carrington’s data set consists of approximately 2.9 million incidents and 3.4 million alleged offenders for the years 1992 to 1999. Carrington’s data contain police incident data representing 6 of the 13 provinces and territories in Canada, consisting of 41% of crime known to Canadian police—a great leap beyond the relatively small-scale studies that are most often used to investigate the phenomenon of co-offending. Overall, co-offending incidents account for 24% of crimes: 44% for young offenders and 20% for adult offenders. Even for young offenders at the age of 12, he found that 57% of reported incidents are co-offences. These results can be interpreted as contrary to the literature in one sense: Co-offending is less dominant at young ages than found by many smaller studies in the past. Thus, Carrington gives us the strongest reason in almost a century to ask whether co-offending has been exaggerated, though still important. He shows that the frequency of co-offending incidents and the average number of co-offenders per incident essentially decreases monotonically from age 10, leveling off once offenders reach their mid-30s.⁶ His results are not out of line with the larger literature but impel us toward further detailed analysis.

In an analysis of co-offending in the United States, Stolzenberg and D’Alessio (2008) use data from the National Incident-Based Reporting System (NIBRS) to calculate age–crime curves, ages 8 through 68, from 446,311 criminal arrests from seven states. Even more than Carrington (2002), their research contradicts the fundamental empirical regularity that co-offending is substantial for young offenders: Co-offending rates are substantially lower than solo-offending rates for all ages, even for those aged 12 and younger—the rate of co-offending still rises in early adolescence and begins to fall in the late teenage years. This result is consistent across gender, ethnic, and crime classifications (violent vs. property crime). Their conclusion is emphatic:

[T]he group nature of juvenile offending appears to be overstated. . . . [O]ur analysis reveals that solo offending rather than co-offending is the domina[nt] form of criminal activity among juveniles . . . [S]olo offending is the primary form of offending among all ages including juveniles. This is an interesting finding and contravenes nearly all previous research. (Stolzenberg & D’Alessio, 2008, pp. 79-81)

Consequently, if Stolzenberg and D’Alessio (2008) are correct, the common understanding of juvenile offending as a group phenomenon requires revision. Even the Carrington (2002) research suggests downplaying the significance of co-offending.

In another recent study of co-offending that uses a large incident-based data set (105,000 offences), but in the United Kingdom, van Mastrigt and Farrington (2009) find that co-offending is present in a relatively small fraction of offences: 10% when counting offences but 22% when counting participations—participations is another term for multiple counting, discussed in Note 5. However, it should be noted that there is substantial variation across different ages; in fact, the relationship between co-offending participations and age found by van Mastrigt and Farrington is similar to that found by Carrington (2002): Co-offending participations peak at approximately 40% in the early teen years, rapidly decreasing to approximately 15% by the mid-20s. Though youth offending does consist of a significant portion of co-offending—enough to consider co-offending an important component of youth offending—this finding provides further support for the notion that estimates of co-offending may have been exaggerated in previous research that used relatively small samples. These aggregate crime analyses must be interpreted with caution. Both Carrington and van Mastrigt and Farrington find substantial variation in co-offending participation rates across crime classification: Simple assault, sexual assault, and drug crime classifications have low co-offending participation rates. As such, if a data set, no matter how large, is dominated by one or more of these crimes (simple assault, for example, because it is such a common crime type), this will be reflected in the aggregate analysis.

Most recently, two Canadian studies have shown the importance of co-offending, particularly for youth. Andresen and Felson (2010), employing a large incident-based data set (750,000 offences) from British Columbia, Canada, confirm the importance of co-offending found in previous studies using relatively small samples. For example, in aggregate, the co-offending participation rates for youth (11-17 years) are 50.1% but are as high as 66% (11 years) decreasing to 44% (17 years). However, when considering disaggregated crime classifications (commercial burglary, residential burglary, and other burglary), Andresen and Felson find that co-offending participation rates are as high as 92% (other burglary, 12 years) and remain above 74% up to and including 17 years of age (commercial and other burglary).

Finally, Carrington (2009) uses a large incident-based data set comprised of 110,000 participations (approximately 55,000 offenders) and finds results similar to Andresen and Felson (2010).⁷ Carrington finds that burglary, robbery, arson, mischief (property damage), theft of motor vehicle, other property offences, and theft under US$5,000 all have co-offending participation rates for youth (5-17 years) that are greater than 50%—only sexual assault, assault, other offences against the person, and drug-related crimes had co-offending participation rates less than 50%. Consequently, the importance of co-offending cannot be dismissed based on a small number of recent studies, regardless of the size of the data set. This is not to say they are correct, but other studies using large incident-based data sets have confirmed past results. In addition, understanding detailed crime classifications are essential before any bold statements can be made regarding any overstatements made in previous research.

In short, the older studies of co-offending that use relatively small samples consistently show that the prevalence of co-offending is high, whereas the more recent studies using large samples are inconsistent regarding this prevalence. An obvious question to ask here is—why? Stolzenberg and D’Alessio (2008) may not have found a high prevalence of co-offending because they do not disaggregate by crime type, aside from distinguishing between property and violent offences, and/or because they do not count participations, only incidents. However, Carrington (2002) and van Mastrigt and Farrington (2009) do disaggregate by crime type and count participations; Carrington’s (2002) results are much more in line with previous research than those of van Mastrigt and Farrington. If their estimates are incorrect—Carrington’s (2009) most recent research on the prevalence of co-offending confirms past research, using a more comprehensive data set than his 2002 research—the only plausible explanation is their data are somehow not representative. Such a claim can only be made on detailed examination of their raw data. More importantly, this inconsistency highlights the need for more studies using large data sets to investigate the phenomenon of co-offending.

The present analysis contributes to this literature using a large incident-based data set. The purpose of this presentation is twofold. The first is to investigate the “known” stylized facts of co-offending using a large incident-based data set, employing multiple counting. This is to replicate previous work that calls the significance of co-offending into question (Carrington, 2002; Stolzenberg & D’Alessio, 2008; van Mastrigt & Farrington, 2009). The second is to extend the knowledge regarding co-offending focusing on solo- and co-offending diversification for multiple crime classifications over young and adult offenders.

Data and Method

The incident-based data used in the analyses below are extracted from the Royal Canadian Mounted Police (RCMP) Police Information Reporting System (PIRS) for British Columbia. These data include 174 of the 186 police jurisdictions (67% of the provincial population) patrolled by the RCMP. All data are extracted from August 1, 2002, to July 31, 2006. These 4 years are the complete set of incidents dealt with by the RCMP—previous years are available but only for more serious offences such as homicide and sexual assault.

There are approximately 5 million negative police contacts in the PIRS database, with each of those contacts containing information on approximately 9 million individuals (offenders, victims, complainants, and witnesses)—a negative police contact occurs when a crime has taken place or the police are needed because of the potential for a crime resulting from a disruptive person, for example. These data have several advantages to assess co-offending. First, though others have used large-scale data sets to investigate co-offending, the data set employed here is the largest to be used thus far. Each of the 5 million incidents has a list of corresponding individuals associated, listing why they are in the database and their age at the time of the incident. Second, though these data represent 174 different police jurisdictions, they are tabulated from one police agency or reporting body. Consequently, unlike the NIBRS database that literally has thousands of enforcement agencies reporting, there is a lower likelihood of inconsistencies in the reporting of criminal incidents. Finally, the PIRS database covers 4 years such that all statistics are calculated using averages across the years. This prevents the possibility of reporting only on one potentially unusual year of criminal activity.

From this database, co-offending is calculated using the classifications of suspect, chargeable, and charged for those in the age range of 12 to 29.⁸ This age range is used in the analysis because there is an established literature that shows that these ages dominate the volume of crime—see, for example, Hirschi and Gottfredson (1983). The use of these three classifications is more inclusive than Carrington (2002) with the addition of suspect. The addition of this classification is to have data that are comparable with Stolzenberg and D’Alessio (2008), which include those who are arrested. Suspects represent 27.5% of the data and their exclusion does not have any qualitative impact on the results.

The current analysis is based on criminal incidents, considering 12 crime classifications that represent the vast majority of violent and property crimes. Violent crimes include homicide, sexual assault, aggravated assault,⁹ assault, robbery, and armed robbery; property crimes include burglary (residential, commercial, and other¹⁰), theft of motor vehicle, and theft from motor vehicle. In the PIRS database, the classifications of theft and theft from motor vehicle are broken down into above and below CDN$5,000, but these further breakdowns are aggregated because of the very small number of incidents in the above CDN$5,000 classifications. The total number of criminal incidents is approximately 750,000, and the total number of criminal participations is approximately 1.2 million.

The methodology used is straightforward. Aside from some statistical significance testing (two-sample difference of proportions test, p value of 10%) for differences in the diversification of crime classifications in solo- and co-offending, the methodology is descriptive.

Last, co-offending is measured using participations rather than incidents as outlined by Frank and Carrington (2007). If two boys, ages 15 and 16, commit a crime together, they need to be sorted into separate ages. If co-offending rates are based on incidents, each age group is credited with 0.5 incidents. With three co-offenders of different ages, 0.33 incidents is assigned to each age group. Participations become a more practical basis for crime accounting and disaggregation in an era when very large data allow specific age coding. We acknowledge that the decision to use participations does increase offending rates for those crimes that have a greater average number of co-offenders, for example. However, we consider this approach more realistic because counting incidents does not consider how many individuals are involved in each crime.

Results

The Basic Regularities

Table 1 shows that the general finding that co-offending dominates solo offending in younger ages is present for most crime classifications. For the sake of brevity, these results are consolidated in one table; detailed figures by age are available on request. In fact, in a number of cases (commercial burglary, other burglary, homicide, robbery, and armed robbery), co-offending is still highly prevalent. The numbers reported here are similar to those reported by Carrington (2009) for youth. Needless to say, this is further evidence that research using large incident-based data sets in other contexts is necessary to confirm or deny this inconsistency with more recent research that employs large incident-based samples. This current analysis shows that the percentage of co-offending is of greater magnitude than Carrington (2002) and Stolzenberg and D’Alessio (2008). This discrepancy may be explained for Stolzenberg and D’Alessio because they measure incidents, not participations, but Carrington (2002) does measure participations. The more recent research of Andresen and Felson (2010) and Carrington (2009) also confirm the “old fact” that co-offending dominates young offending, but more research is necessary here because in addition to Stolzenberg and D’Alessio, van Mastrigt and Farrington (2009) do not confirm this old fact.

Table 1.

Co-Offending as a Percentage of Crime Participations, Ages 12 to 17 and 18 to 29, British Columbia, Canada, August 1, 2002 Through July 31, 2006

Crime types	Ages 12-17	Ages 18-29
Commercial burglary	77.9	51.9
Residential burglary	61.8	39.9
Other burglary	79.9	48.5
Theft from motor vehicle	65.3	34.4
Theft of motor vehicle	54.3	36.2
Theft	53.2	23.5
Sexual assault	15.0	17.4
Aggravated assault	48.7	41.1
Assault	43.8	34.2
Homicide	83.1	57.7
Robbery	66.8	46.9
Armed robbery	63.9	50.6

Figure 1 reports age-specific solo- and co-offending rates—all denominators in all figures are single-year age specific. This figure shows that co-offending is dominant in younger ages, 8 to 15 years. This figure is a better comparison with the age–crime curves generated by Stolzenberg and D’Alessio (2008) and shows that co-offending dominates solo offending up to and including those aged up to and including 15 years. In addition, when the age–crime curves are calculated based solely on incidents, the method of calculation used by Stolzenberg and D’Alessio, co-offending still dominates but only to those aged up to and including 12 years—the rates of those who are 13 years is almost identical.

Figure 1.

Solo- and co-offending rates per 100,000

The Diversification of Solo- and Co-Offences

Turning to the question at hand in the current analysis, Figure 2 shows the violent crime to property crime ratio for solo- and co-offending. Clearly shown in this figure is that prior to an age of 21 years, solo- and co-offending follow quite different patterns. Though both co- and solo offending are dominated by property crimes because their values are always below unity, co-offending has a far lower ratio than solo offending for all young offender ages, 12 to 17 years of age; at the early ages, property crimes dominate co-offending by a factor of four-to-one but by 19 years the two crime classifications are approximately one-to-one. Therefore, there is a difference in the distribution of crime classifications when comparing solo- to co-offending. This shows that not only is co-offending similar to general crime patterns that have property crime frequencies greater than violent crime frequencies, but co-offending is also far more prone to this relationship (more property crime than violent crime) than is solo offending during youth—no notable differences are apparent for those older than 17. As such, crimes for economic profit are much more the domain of co-offending.

Figure 2.

Ratio of violent crime to property crime

Of the 216 differences between solo- and co-offending (12 crime and 18 age classifications), 44 (20.4%) are statistically significant—the number of statistically significant differences is 19 when considering a Bonferroni correction. The general pattern that emerges from the statistical differences in these distributions is as follows: During the teenage years, theft and (both residential and commercial) burglary have a higher degree of co-offending participations than for solo offending; in the early and mid-20s, co-offending participations dominate for residential burglary, and in the late 20s, only commercial burglary still has co-offending participations dominate. Overall, considering statistical significance, assault and theft are predominantly composed of solo-offending participations whereas burglaries are predominantly composed of co-offending participations. The overall pattern of differences, without paying particular attention to statistical significance, shows that co-offending is far more diversified than solo offending, and the co-offending distribution, in almost every age classification, favors robbery, armed robbery, burglaries, theft of motor vehicle, and theft from motor vehicle.

Another method of showing the differences in the distribution of crime classifications for solo- and co-offending is the Index of Dissimilarity used by demographers (see Duncan & Duncan, 1955a, 1955b). The value of the index is one half the sum of the absolute value differences between two vectors of percentages, each summing to 100%. The value of the Index of Dissimilarity is then interpreted as having to change X percentage of co-offending participations to make its distribution the same as the distribution of solo-offending participations. The Index of Dissimilarity is shown in Figure 3 for the ages 12 to 29. Because the calculation of this index considers all differences, statistically significant or not, an Index of Dissimilarity may be calculated that only considers those differences that are statistically significant.

Figure 3.

Index of dissimilarity

As shown in Figure 3, the degree of change for the distribution of co-offending to be the same as the distribution of solo offending is as high as 48% for the age of 12, falling to 30% by the late teenage years and thereafter. Only considering statistically significant differences, the Indices of Dissimilarity follows the same pattern (r = 0.889), but the values are correspondingly lower: beginning at 30%, falling to 17.5% in the late teenage years and thereafter. This last analysis shows, yet again, that co-offending follows a different pattern than solo offending, and co-offending must be considered when investigating the etiology of crime.

Though instructive, Figures 2 and 3 can only provide a cursory look into the different distributions by crime type for solo- and co-offending. Because of the large number of observations in our data set, the diversification of crime types may be broken down by offending type (solo- or co-offending), by crime classification, and by single year of age (12-29 years). As shown above, the percentages of solo- and co-offending across the various crime classifications prove to be instructive. Figure 4 shows the percentage of crimes that are assaults and theft within each of co- and solo offending. Immediately apparent is that the difference between assault and theft are far greater for co-offending than solo offending. By the late teenage years, assault and theft are similar percentages for both co- and solo offending, but the percentages are greater for solo offending in both cases. Though the differences are not great, they are statistically significant for most of the ages in the range, indicating that solo offenders are more specialized in their crime classifications than co-offenders.

Figure 4.

Distribution of offences, assault, and theft

The diversification of crime classifications for solo offending (Figures 5a and 6a) shows very little change over the age range, aside from sexual assault, aggravated assault, residential burglary, and theft of motor vehicle. Sexual assault decreases its share within the distribution of solo offending whereas the latter three crime classifications increase, all leveling off by the mid-teenage years. Also notable is that aside from aggravated assault, residential burglary, and theft of motor vehicle, all of these crime classifications comprise a relatively small proportion of the distribution for solo offending.

The diversification of crime classifications for co-offending (Figures 5b and 6b) shows similar changes over the age range as solo offending: Aggravated assault, residential burglary, and theft of motor vehicle all increase, primarily through to the mid-teenage years—the increases in aggravated assault and theft of motor vehicle are particularly great. One key distinction emerges in the diversification of crime classifications for co-offending that is different from solo offending. The violent crime classifications (Figure 5b), aside from aggravated assault, all comprise of a relatively small proportion of the distribution for co-offending. However, the property crime classifications (Figure 6b), aside from other burglary, all comprise a moderate proportion of the distribution for co-offending. With the proportions of theft (20%), residential burglary (10%), theft of motor vehicle (10%), commercial burglary (5%), and theft from motor vehicle (4%) all consisting of a significant portion of the distribution for co-offending, co-offending exhibits far more diversification than solo offending, particularly for property crimes.

Figure 5a.

Distribution of offences, violent crimes, and solo offending

Figure 5b.

Distribution of offences, violent crimes, and co-offending

Figure 6a.

Distribution of offences, property crimes, and solo offending

Figure 6b.

Distribution of offences, property crimes, and co-offending

Overall, it would appear that some types of crime classifications are likely to be associated with co-offending. This may be because those who get involved in co-offending are simply exposed to more opportunities for committing different types of crime classifications. With more exposure/opportunity comes learning more skills and making more (deviant) social contacts. Therefore, the teenager who shoplifts alone is less likely than the teenager who shoplifts in a group to move into different crime classifications.

Discussion and Directions for Future Research

Co-offending is committing an offence with at least one accomplice. Most of the research investigating co-offending shows youth co-offending to be extensive, groups of co-offenders tend to be small, those that co-offend tend to commit crimes more frequently and more serious crime, and that co-offending group membership is not stable. Though this phenomenon has been noted for close to a century (see Breckenridge & Abbott, 1912), relatively little research on co-offending exists. Some of this more recent research that employs large-scale, incident-based data sets calls the significance of co-offending into question, potentially forcing the criminological community to reevaluate the state of knowledge on co-offending. Accordingly, the current analysis has evaluated these most recent findings using a large-scale, incident-based data set (n = 5 million) from the RCMP in British Columbia. Specifically, in the current analysis, the incidence of co-offending for those aged 12 to 29 and the specific characteristics of co-offending for a variety of crime classifications exhibit two particularly interesting results. First, previous analyses of co-offending that employed relatively small-scale analyses are robust with regard to the volume and rates of co-offending, as high as 60 or 70, employing multiple counting. Second, co-offending is dominantly composed of property crime classifications and is significantly more diversified than for solo offending, particularly for youth. Consequently, support is found for co-offending to impact the awareness spaces of offenders, in an aggregate context. As such, to the extent that diversification is the result of nonredundant networks, these findings are consistent with the idea that nonredundant networks (McGloin & Piquero, 2010) are more common than redundant networks for those who co-offend.

The implications for these results are twofold. First, with regard to the issue of the importance of co-offending it is shown that not only is co-offending the dominant form of criminal activity for youth but it is also higher (almost one third) for adults than previously thought. Consequently, the phenomenon of co-offending cannot be ignored because it is shown to be present in small-scale studies and now in a number of large-scale studies. As discussed above, some of these large-scale studies are not comparable because of differences in measurement methodology. Moreover, though the degree of co-offending may still be up for debate, co-offending is still a substantial portion of criminal activity for youth (regardless of the study) and needs to be better understood. Second, the variation of co-offending is significant across not only different age categories but also crime classifications. Though this result is not new, the nature of this relationship from youth to adults is shown to vary significantly from crime classification to crime classification.

These results are not without their limitations. Two limitations are immediately apparent. First, all the standard limitations of using police data are present here: Issues such as only analyzing crimes reported to the police and any potential biases in police enforcement—see Sherman, Gartin, and Buerger (1989) for a full discussion of these issues. There is little that can be done regarding the first issue, but the second issue is mediated by using data from 174 police jurisdictions over 4 years. As such, any potential biases in police enforcement (a short-term crackdown on drug crime in response to an identified problem, for example) are minimized. Second, the present results are based on an aggregate analysis. It is possible that these aggregate patterns are not present when analyzing individual offenders. Or, more likely, there are important nuances within these characteristics of co-offending that may only be identified in an individual-level analysis. This is a direction for future research.

Directions for future research can be organized along more aggregate analyses of these data as well as analyses of individuals and the patterns of co- and solo offending. At the aggregate level of analysis, further research is necessary to understand the mobility of co-offending versus solo offending. As found by Sarnecki (2001), young offenders tend to find their co-offenders in their immediate neighborhoods, with older offenders being more willing to find co-offenders as far away as different municipalities. Such analyses should be confirmed using North American data as well as conducting the analysis beyond the ages studied by Sarnecki, up to the age of 21. Also at the aggregate level of analysis, the nature of the offending groups themselves may be analyzed. Most research finds that co-offending groups tend to be small and decrease with age (Carrington, 2002; Reiss & Farrington, 1991; Zimring, 1981). Confirming this relationship with another large incident-based data set provides significant value, but investigating the distribution of crime classifications for larger and smaller groups may also be analyzed. Last, age homogeneity within groups (see Sarnecki, 1986, 2001; Warr, 1996) may be investigated to not only confirm or deny the said relationship but also to investigate the differences in the distribution of crimes (if any) between the groups that contain offenders of similar ages and the groups that contain members of significantly different ages. Indeed, as shown by Shaw and Moore (1931), some prolific and serious criminals tend to commit their offences with older offenders.

Future research on patterns of co-offending at the individual level using this (or another) large incident-based data set needs to investigate the experience levels of co-offenders. Carrington’s (2002) UCR2 data set does not provide such an opportunity but our data set contains information that allows us to follow offenders (through coded unique identifiers, not actual identities) for the available years. Because of this added feature in the current data set, the research of McGloin, Sullivan, Piquero, and Bacon (2008) may be extended on the short-lived nature of offending groups and the lack of reusing offenders for different crimes that extends into adulthood. Finally, the characteristics of co-offending studied in the present analysis at the aggregate level may be performed at the individual level to investigate the volume of co-offending over different ages, the number of co-offenders used, and the distribution of crime classifications as offenders age. Because of more recently available large-scale incident-based data sets that allow for the analysis of co-offending and the shown importance of this phenomenon, it is now time to investigate this phenomenon in great detail.

Footnotes

Acknowledgements

We would like to thank the 3 anonymous referees for their comments that significantly increased the quality of this article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Andresen

M. A.

Felson

(2010). The impact of co-offending. British Journal of Criminology, 50, 66-81.

Blumstein

Cohen

Farrington

D. P.

(1988). Criminal career research: Its value for criminology. Criminology, 26, 1-35.

Blumstein

Cohen

Roth

J. A.

Visher

C. A.

(Eds.). (1986). Criminal careers and “career criminals” (2 vols.). Washington, DC: National Academy Press.

Blumstein

Farrington

D. P.

Moitra

(1985). Delinquency careers: Innocents, desisters, and persisters. In Morris

Tonry

(Eds.), Crime and justice (Vol. 6, pp. 187-219). Chicago, IL: University of Chicago Press.

Brantingham

P. L.

Brantingham

P. J.

(1981). Notes of the geometry of crime. In Brantingham

P. J.

Brantingham

P. L.

(Eds.), Environmental criminology (pp. 27-54). Prospect Heights, IL: Waveland Press.

Breckenridge

S. P.

Abbott

(1912). The delinquent child and the home. New York, NY: Russell Sage.

Carrington

P. J.

(2002). Group crime in Canada. Canadian Journal of Criminology, 44, 277-315.

Carrington

P. J.

(2009). Co-offending and the development of the delinquent career. Criminology, 47, 301-335.

Conway

K. P.

McCord

(2002). A longitudinal examination of the relation between co-offending with violent accomplices and violent crime. Aggressive Behavior, 28, 97-108.

10.

Duncan

O. D.

Duncan

(1955a). A methodological analysis of segregation indexes. American Sociological Review, 20, 210-217.

11.

Duncan

O. D.

Duncan

(1955b). Residential distribution and occupational stratification. American Journal of Sociology, 60, 493-503.

12.

Erickson

(1971). The group context of delinquent behavior. Social Problems, 19, 114-129.

13.

Felson

(2003). The process of co-offending. Crime Prevention Studies, 16, 149-167.

14.

Frank

Carrington

P. J.

(2007). Estimation of offending and co-offending using available data with model support. Journal of Mathematical Sociology, 31, 1-46.

15.

Gold

(1970). Delinquent behavior in an American city. Belmont, CA: Brooks/Cole.

16.

Hindelang

M. J.

(1971). The social versus solitary nature of delinquent involvement. British Journal of Criminology, 11, 167-175.

17.

Hindelang

M. J.

(1976). With a little help from their friends: Group participation in reported delinquent behaviour. British Journal of Criminology, 11, 109-125.

18.

Hirschi

Gottfredson

(1983). Age and the explanation of crime. American Journal of Sociology, 89, 552-584.

19.

McCord

Conway

K. P.

(2002). Patterns of juvenile delinquency and co-offending. In Waring

Weisburd

(Eds.), Advances in criminological theory (Vol. 10, pp. 15-30). New Brunswick, NJ: Transaction Books.

20.

McGloin

J. M.

Piquero

A. R.

(2010). On the relationship between co-offending network redundancy and offending versatility. Journal of Research in Crime and Delinquency, 47, 63-90.

21.

McGloin

J. M.

Sullivan

C. J.

Piquero

A. R.

Bacon

(2008). Investigating the stability of co-offending and co-offenders among a sample of youthful offenders. Criminology, 46, 155-187.

22.

Piquero

A. R.

Farrington

D. P.

Blumstein

(2007). Key issues in criminal career research: New analyses of the Cambridge Study in Delinquent Development. Cambridge, UK: Cambridge University Press.

23.

Reiss

A. J.

Jr. (1980). Understanding changes in crime rates. In Fienberg

S. E.

Reiss

A. J.

Jr. (Eds.), Indicators of crime and criminal justice: Quantitative studies (pp. 11-17). Washington, DC: Bureau of Justice Statistics.

24.

Reiss

A. J.

Jr. (1988). Co-offending and criminal careers. In Morris

Tonry

(Eds.), Crime and justice (Vol. 10, pp. 117-170). Chicago, IL: University of Chicago Press.

25.

Reiss

A. J.

Jr. Farrington

D. P.

(1991). Advancing knowledge about co-offending: Results from a prospective longitudinal survey of London males. Journal of Criminal Law and Criminology, 82, 360-395.

26.

Sarnecki

(1986). Delinquent networks. Stockholm, Sweden: National Council for Crime Prevention.

27.

Sarnecki

(2001). Delinquent networks: Youth co-offending in Stockholm. Cambridge, UK: Cambridge University Press.

28.

Shaw

C. R.

McKay

H. D.

(1931). Report on the causes of crime (Vol. 2). Washington, DC: Government Printing Office.

29.

Shaw

C. R.

Moore

M. E.

(1931). The natural history of a delinquent career. Chicago, IL: University of Chicago Press.

30.

Sherman

L. W.

Gartin

P. R.

Buerger

M. E.

(1989). Hot spots of predatory crime: Routine activities and the criminology of place. Criminology, 27, 27-56.

31.

Stolzenberg

D’Alessio

S. J.

(2008). Co-offending and the age–crime curve. Journal of Research in Crime and Delinquency, 45, 65-86.

32.

van Mastrigt

S. B.

Farrington

D. P.

(2009). Co-offending, age, gender and crime type: Implications for criminal justice policy. British Journal of Criminology, 49, 552-573.

33.

Warr

(1996). Organization and instigation in delinquent groups. Criminology, 34, 11-37.

34.

Warr

(2002). Companions in crime: The social aspects of criminal conduct. Cambridge, UK: Cambridge University Press.

35.

Zimring

F. E.

(1981). Kids, groups and crime: Some implications of a well-known secret. Journal of Criminal Law and Criminology, 3, 867-885.