Skip to main content
SearchLoginLogin or Signup

The Search for Suitable Homicide Co-Offenders among Gang Members

Published onJan 01, 2015
The Search for Suitable Homicide Co-Offenders among Gang Members
·

Abstract

Little is known about homicide co-offending networks at the individual gang member level. Of particular interest is whether and to what degree gang members who are selected to participate in murders are different from those who are not. The current study constructed the co-offense network of eighteen subjects from the Incarcerated Serious and Violent Young Offender Study who were identified as members of a prominent gang within British Columbia, Canada, referred to as the BC Gang. This gang started to form not long before seven offenders together committed a homicide that was orchestrated by the founder and leader of the BC Gang. After this offense, these seven offenders became some of the most central actors within a large network of co-offenders (n = 137) that was measured at four time periods over a 14 year period. Over this period, a second murder, like the first, was orchestrated by the leader of the BC Gang, offering a rare glimpse into the co-offending recruitment decisions made by a high ranking gang member for two separate homicides. Although only 25% of the 137 co-offenders are BC gang members (n = 35), 100% of the offenders selected for a homicide were members of this gang (n = 13). The network contained 8 separate components at the final measurement period, but all 13 homicide offenders were connected to the same network component of 48 individuals.

Keywords

Co-offending; gang homicide; gang structure; homophily; social network analysis

Introduction

Criminal networks and co-offending are complementary lines of empirical study that can further theory related to the processes leading to co-offending (McGloin & Nguyen, 2011). Yet, few researchers have examined the topic of criminal networks and co-offending simultaneously (for exceptions see Malm, Nash, & Vickovic, 2011; Morselli, 2009; McGloin & Piquero, 2010), fewer have examined this topic in the context of gang networks (see Bouchard & Konarski, 2013; Papachristos, 2009), and to our knowledge no studies have examined gang criminal networks and homicides involving co-offenders at the individual level. Existing research has focused on the “gang” as the unit of analysis (e.g., Papachristos, 2009), case studies (e.g., Morselli, 2003), or homicide event characteristics (Rosenfeld, Bray, & Egley, 1999). The current study addresses the paucity of research on the co-offending networks of street gangs by examining individual involvement in gang-related homicide involving co-offenders.

The point of departure for this study arose from Tremblay’s (1993) discussion of the search for suitable co-offenders. Tremblay reasoned that because most offenses require accomplices, an important aspect of successful crime completion involves the participation of offenders who possess the skills needed to commit the offense. To date, Tremblay’s (1993) suitable co-offender concept has yet to be empirically tested, and we argue that network methods are necessary to operationalize Tremblay’s idea. If there is indeed a “search” for co-offenders, this search will exclude a number of potential accomplices that could most easily be identified if the social network around an offender is mapped. Given the stakes involved in co-participating in murder, the search is assumed to be especially revealing in the case of premeditated gang-related homicides, such as the ones examined in this study.

Data were collected on the co-offending network of 18 members of what we will refer to as the BC Gangi. The founder and leader of this gang orchestrated two homicides. At the time of the first homicide, the BC Gang was just one of many small youth-level street gangs operating within the Lower Mainland of British Columbia. However, after this homicide, the gang increased in size and in influence in the Lower Mainland drug trade. Despite a lengthy period of incarceration, the BC Gang leader maintained his leadership role and was responsible for orchestrating a second homicide in adulthood. The data offer a rare glimpse into a high ranking gang member’s recruitment decisions for two separate homicides, one inter-gang and one intra-gang. As the homicides occurred approximately ten years apart, the network trajectory of the homicide offenders could be mapped over time. Questions can therefore be addressed regarding whether those recruited for the first homicide were re-recruited for the second homicide, as well as whether recruitment for the first homicide impacted their network trajectory.

Recruiting co-offenders

Co-offender selection for some offenses involves a recruiter actively seeking co-offenders, the joiners, who possess characteristics that would be valuable for the commission of a particular offense (McGloin & Nguyen, 2011; Tremblay, 1993; van Mastrigt & Farrington, 2009). As part of this co-offender selection process, a recruiter selects an individual based on their perception that he/she is more suitable than others who were available from the recruiter’s pool of criminal resources (Tremblay, 1993) or social network. Therefore, the individual selected represents only a fraction of the offender’s potential co-offenders.

The concept of suitability, however, has scarcely been measured empirically, and is likely to vary for one crime to the next. For gang members, fellow “members” are likely to be prime candidates for selection. Yet, co-offending with outsiders may sometimes be necessary, even rewarding (Tremblay, 1993). In two different studies using SNA to examine the co-offense networks of gangs, Bouchard and Konarski (2013) and Morselli (2009) both found that some gang members co-offended primarily with other gang members, but other gang members had non-gang members as key contacts to facilitate their criminal activities. These studies provide evidence that gang members select co-offenders from within a broader social network of both gang and non-gang members, which reflects the importance of a network approach to studying gang co-offending. This is especially important given the blurred boundaries of gang membership in the first place (Bouchard & Konarski, 2013; Waring, 2002).

Reasons for co-offending only with members of one’s own gang or with a mix of gang and non-gang members may vary based on individual tendencies and preferences, but it may also vary based on offense type. Offenses such as drug trafficking require a large number of offenders to maintain operations. Some roles can be quite menial (e.g. drug runner) and can be managed outside the gang, while others (e.g. wholesaler) may require social, human or economic capital that gang members simply do not possess (Malm, Bichler, & Nash, 2011; Morselli, 2009). The search for co-offenders in drug trafficking ventures is more likely to be driven by the nature of the drug market and by factors associated with criminal achievement such as entrepreneurial ability and criminal social capital (McCarthy & Hagan, 2001; Waring, 2002). In contrast, because homicide offenses have the most deleterious consequences if detected, recruiters searching for suitable homicide co-offenders may be less concerned with entrepreneurial ability and more concerned with the selection of “strong ties” and trusted associates (Tremblay, 2011).

Suitability in the context of gang homicides

The principle of homophily assumes that individuals who share similar traits are more likely to trust one another (McPherson, Smith-Lovin, & Cook, 2001; Weerman, 2003). This principle has been used in co-offending studies to explain why offenders are more likely to co-offend with individuals who share certain characteristics, such as ethnicity and other demographic characteristics (e.g., Malm, Nash, & Vickovic, 2011; van Mastrigt & Carrington, 2011; Warr, 2003). Membership in the same gang is another measure of homophily (Weerman, 2003) that may be seen by offenders as an indicator of co-offender suitability, especially for homicides and other retaliation-related crimes (Tremblay, 2011). For certain offenses; however, if other attributes are not also present, shared gang membership may be insufficient for suitability. For example, in the Gambino crime family, lethal violence could be orchestrated only by members who had earned the status of ‘Capo’, a sign that they had ‘paid their dues’ and were deeply embedded within the gang (Morselli, 2003). A recruiter’s emphasis on trust and mutual embeddedness may vary by offense-type. These qualities seem particularly important in homicide offenses because detection has severe consequences. A recruiter’s emphasis on trustworthiness may also vary depending on the type of gang-related homicide being committed.

Broadly, gang members commit two types of homicides: (1) inter-gang homicides against rival gang members or others in conflict with the gang and (2) intra-gang homicides against members within the same gang. Inter-gang homicides tend to be reactionary and therefore may require a lower threshold of trust between recruiter and joiner. For intra-gang homicides, recruiting co-offenders who are also members of this gang may be particularly risky if these gang members are connected to and have the ability to warn the targeted victim. This is true especially of potential joiners who may have strong ties to the victim through shared gang membership. Thus, the recruiter must be able to trust the joiner to not warn the targeted victim. In a recruiter’s ideal intra-gang homicide recruitment decision, co-offenders will be selected who have strong ties to the recruiter, but weak or no ties to the targeted victim. In other words, it is expected that strong ties are favored over weak ties for intra-gang homicides as long as the co-offenders being recruited do not also have strong ties with the targeted victim. Both intra and inter-gang homicides are represented in the homicides analyzed in this study.

Data and Method

Sample

Approximately 1,400 young offenders were interviewed between 1998 and 2011 within custody facilities in British Columbia (BC), Canada. A sub-sample (n = 18) were identified as members of the BC Gang. A number of requirements for what constitutes a gang were met, including durability, specific symbols, territory, and criminal activity (Curry and Decker, 1998; Klein and Maxson, 2006). This sub-sample was referred to as the “initial 18 BC Gang members” because they represented the initial 18 seeds from which the current study collected network information. Demographic and other characteristics associated with these seeds are outlined in Table 1. The sample was highly involved in criminal activity, averaging 18.2 (SD=14.0) general convictions and 3.7 (SD=2.9) violent convictions. Half of the initial 18 seeds had been convicted of drug trafficking. Five of the initial 18 seeds were involved in three separate homicides. The current study focuses primarily on the two homicides orchestrated by the leader of the BC Gang, but also looks at general differences between all homicide and non-homicide offenders.

--Insert Table One about Here--

Procedure

Corrections Network (CORNET) is an integrated system used to track all provincial offenders in BC. CORNET information included court orders indicating whether an offender had a co-accused for each offense they committed, and if so, the name of that co-accused. This information was used to code co-offending associations made between the initial 18 members of the BC Gang as well as between these 18 initial offenders and ‘other’ offenders. It was not possible to search CORNET for co-offending associations held by these other offenders due to ethical restrictions. However, if two offenders were associated with one another through a mutual co-offense with one of the initial 18 members, then the tie between these two members was coded. The number of co-offenses and co-offenders was likely underreported due to reliance on official data. However, this method still yielded a relatively large network of 137 co-offenders. Of the 119 co-offenders connected to the initial 18 seeds, 17 were BC Gang members.

The co-offending network was measured at four waves of data collection: T1- all co-offenses prior to the first homicide, T2- all co-offenses in T1, the first homicide, and all co-offenses prior to the second homicide (not orchestrated by the gang leader), T3- all co-offenses in T1 and T2, the second homicide (not orchestrated by the gang leader), and all co-offenses prior to the third homicide (orchestrated by the gang leader), and T4- includes the third homicide (orchestrated by the gang leader) as well as all co-offenses recorded from T1 until the end of data collection (Spring 2013).ii The time between each homicide was not equal across the four waves. T1 spanned six years, T2 and T3 spanned five years, and T4 spanned six years.iii The final cumulative co-offending network (T4) contained eight different components and is displayed in Figure 1. The largest (main) component is super-imposed in the top right hand corner.

--Insert Figure 1 about Here--

Measures

Individual attributes were obtained through self-report interviews and therefore were only available for the initial 18 BC Gang members. In addition to interviews, CORNET and media reports were used to identify gang membership status. CORNET also provided information about each offender’s criminal history.iv Criminal histories were used to identify an offender’s network position at each wave of data collection. Criminal history and network position were used to examine homicide offenders and compare them to non-homicide offenders in the full sample. The study design allowed changes in an individual’s network position to be monitored after selection as a homicide co-offender. Network measures included degree and betweenness centrality. Degree centrality measures the total number of ties held by an actor and is indicative of an individual who is highly visible in a network and may carry a certain level of prestige, or of potential risk when the measure is constructed from official crime data. In contrast, betweenness centrality captures an individual’s ability to connect actors who previously were unconnected.

Analytic strategy

Co-offense networks were constructed using UCINET 6 (Borgatti, Everett, & Freeman, 2002). The analyses focus primarily on GM1’s search for co-offenders within a network comprised of BC Gang and non-BC Gang offenders. The results of these analyses are presented in four stages. First, details of GM1’s first homicide are discussed, including examining which offenders he recruited and how this recruitment impacted network trajectories. Second, using cluster analysis (optimization methodv), sub-components within the main co-offending component are interpreted to help understand who GM1 recruited or did not recruit for his second homicide. Third, to set up comparisons of the characteristics of homicide and non-homicide offenders, a third homicide offense, which GM1 was not a part of, is discussed. Fourth, characteristics of homicide and non-homicide offenders are compared on a number of characteristics at the bivariate level.

Results

The founder of the BC Gang, named GM1, maintained leadership of this gang throughout the study period by improving his gang’s status by establishing drug lines throughout the Lower Mainland of British Columbia. His prominence in the network can be seen in Figure 1, which contains the full co-offending network as constructed using 14 years of conviction data from 18 members of the BC Gang. As a teenager, GM1 orchestrated the first homicide by targeting a rival gang. This homicide appeared to facilitate the BC Gang’s ascent from a small youth street gang to one of the most powerful gangs in the Lower Mainland. GM1’s second homicide differed in that GM7, a prominent member of the BC Gang (see Figure 1) and one of GM1’s co-offenders for his first homicide, was targeted. Ancillary reports suggested that GM7’s divergence from the gang and association with offenders outside the gang may have contributed to his death.

Background details on the co-offenders selected for these two homicides are provided below before examining, importantly, whether they are any different than other offenders in the network. An important note about this approach to the question of selection is that the “why him” question cannot be fully answered with this research design. Data on trusted relationships that would be required to satisfactorily address this question were unavailable. Yet, although the puzzle of co-offending selection is not fully solved, homicide offenders can be compared to non-homicide offenders in a number of ways, including network trajectory, gang membership status, and location within the larger network structure.

The co-offenders involved in GM1’s two homicides

GM1 committed his first homicide at age fifteen when he recruited six fellow BC Gang members (GM3, GM4, GM7, GM33, GM34, and GM35; see the middle of the network in Figure 2a) in response to being attacked by a rival street gang. GM1’s recruitment of these six offenders could not be explained by his centrality in the co-offense networkvi, because he had yet to be convicted of any crime.vii Of the offenders recruited, only GM3 was a member of the co-offending network measured in T1. The six co-offenders involved in the first homicide can be divided into two categories, those who benefitted from participation and persisted in the gang (the winners), and those who did not re-appear in the network after this offense (the desisters).

The winners. Three recruits of this early homicide appeared to have an upward trajectory after being released from custody: GM3, GM4, and GM7. The winners co-offended with twenty-two different offenders after their initial homicide offense. GM3 and GM4 committed later homicides in T4 (Figure 1) and T3 (Figure 2b), respectively, and GM7 developed his own dial-a-dope network. In terms of characteristics of the winners, GM3 was 16 at the time of the first homicide. GM3’s previous co-offenders at T1 were not members of the BC Gang, suggesting that although GM3 was a member of the BC Gang, he did not have known co-offending connections with other members prior to this homicide offense. Ancillary reports that GM3’s role in the BC Gang was to work as an enforcer over drug debts were supported by evidence of his propensity for violence (six violence convictions; Table 1), which may have been traits that GM1 viewed as suitable for recruitment. GM4 was also just a teenage member of the BC Gang at the time of the first homicide and had no connections in the co-offending network prior to his involvement in this offense. GM4 appeared to benefit from his homicide with GM1 as he was later recruited into GM1’s dial-a-dope operation. Lastly, GM7 was also just a teenage member of the BC Gang at the time of this offense. Ancillary reports and interview information suggested that, similar to GM1, GM7 came from a well-off family and that their crimes were not financially motivated (see Table 1). GM1 and GM7 were also of the same ethnicity (Asian). Taken together, GM1’s recruitment of GM7 could be related to homophily; although, because GM1’s other co-offenders were not of the same ethnicity, some characteristics of homophily may have been sufficient but not necessary for recruitment.

The desisters. GM33, GM34, and GM35 were three other members of the BC Gang recruited by GM1 and are referred to as the desisters because, despite their early involvement in the main component in the network, they did not co-offend with GM1 or any of the other 17 BC Gang members in our seed sample after this homicide. As they were not part of the initial 18 seeds, information was limited. Ancillary reports did suggest that the desisters were teenagers at the time of the homicide. Also similar to GM1, GM4, and GM7, these offenders were not involved in the network prior to the homicide. In all likelihood, given the rate at which other BC Gang members in the network were convicted over the 14 year period under study, these three may have desisted from the BC Gang at some point after the murder. This does not mean that they desisted from crime altogether, which is a question that cannot be answered with the data.

--Insert Figure Two about Here--

Prior to the first homicide (T1), the co-offending network of the initial 18 seeds was small, did not include GM1, and did not leave much to be interpreted. In contrast, the main components in T2 (Figure 2a) and T3 (Figure 2b) had grown to 38 and 43 offenders, respectively. The size of the network allowed for cluster analysis (optimization method) to be used to identify sub-components consisting of structurally equivalent actors within the main component. Four different subcomponents were identified. The white and darkest gray node represented offenders on the periphery of the network, black nodes were primarily those who were involved in GM1’s first murder, and the lighter gray nodes represented GM1 and GM7’s dial-a-dope networks. The different subcomponents are examined to help understand whether an offender’s network structure was influential in GM1’s decision to recruit or not recruit.

At T3, despite GM1 having direct connections with sixteen offenders and being at most two steps away from 42 offenders, only one of these offenders, GM3, was recruited for GM1’s second homicide. Again, GM3’s role as an enforcer may again have led to his selection. The other five homicide co-offenders for this offense were new to the network in T4: GM2, GM5viii, GM11, GM12, and GM 101. Like the first homicide, all co-offenders were BC Gang members, suggesting again that homicide is a “members only” offense. Although access to the motivation behind this homicide is unavailable, it is informative that GM7, the victim, was GM1’s most structurally equivalent actor in the network. From T1 until T4, GM1 and GM7 had nearly identical betweenness and degree centrality scores (Figure 3a). When GM7 expanded his co-offending network by co-offending with non-BC Gang members unconnected to other offenders within the main component (Figure 2), degree centralization within this component consistently declined (see Table 2). Although speculative, because ancillary reports indicated that the BC Gang was unhappy with GM7 dealing drugs with non-members, GM1’s homicide against GM7 may have been committed to exert greater control over the network. Betweenness centralization increased from T2-T4 (Table 2), which supports our assertion that there was a growing rivalry due to GM7 having access to a large group of co-offenders that GM1 did not have access to. The three cluster analyses in Figures 1 (T4), 2a (T2), and 2b (T3) also indicated that GM1 and GM7 were located in the same sub-component of the main network component. The sub-components were therefore potentially helpful in informing why GM1 appeared to avoid recruiting offenders within his immediate network (with the exception of GM3).

--Insert Table 2 about Here--

--Insert Figure 3 about Here--

Offenders not selected by GM1. GM1 could be more selective of co-offenders in his second homicide because of his central position in the network, and yet did not recruit any co-offenders from three of the four sub-components identified in the cluster analysis in T3. Two of these sub-components (the white and darkest gray nodes in Figure 2b) contained only three BC Gang members, giving GM1 limited options considering the apparent emphasis on recruiting only gang members for homicides. GM1 also did not recruit from the subcomponent that he and the victim, GM7, belonged to, possibly to reduce the risk of GM7 being made aware of his plans. For example, thirteen offenders in this subcomponent were connected to the network only through their co-offending allegiance with GM7 (see the extreme left-hand portion of Figure 2b). The fourth sub-component, denoted by black nodes, contained offenders that had co-offended with GM7 as part of GM1’s first homicide offense. The inability to avoid detection in the previous homicide may have deterred GM1 from re-recruiting them. GM3 was still recruited from this component, possibly because his propensity for violence made him a valuable asset.

The ‘other’ homicide

Between GM1’s first and second homicide, another homicide, occurring in T3, was committed by offenders in the main component, GM4 and another BC Gang member, GM230. Based on collateral information, this offense was a targeted shooting against a rival gang member who was expected to testify in court as part of the trial of another BC Gang member. Although this offense was unsuccessful (the shooter missed), it is included in the current study because the intention of the offense was the same as the other two offenses. Like GM1’s two homicides, this ‘other’ homicide involved recruitment of an offender (GM230) who was not previously connected to the network.

Homicide offenders: Are they any different?

The three homicide offenses are together used to examine whether homicide offenders differ from non-homicide offenders. In each wave of data collection that GM1 was active, normalized betweenness and degree centrality scores were compared between GM1’s homicide and non-homicide co-offenders (see Table 3). Homicide co-offenders averaged higher betweenness centrality scores only at T2. In the latter two columns of Table 3, the centrality of GM1’s co-offenders was compared to the centrality of all other offenders in the network (excluding GM1) and revealed that after T1, GM1’s co-offenders had significantly higher degree and betweenness centrality scores. Thus, it seems that any co-offense with him would improve an offender’s visibility in the network.

--Insert Table 3 about Here--

Referring back to Table 1, comparisons were made between homicide and non-homicide offenders who were part of the initial 18 seeds. Notwithstanding the lack of statistical power due to sample size, chi-square analyses indicated that non-homicide offenders were more likely to be born in Canada and were more likely to report that money motivated their involvement in crime. T-tests indicated that homicide offenders averaged significantly higher scores on betweenness centrality compared to non-homicide offenders. However, Figure 3b indicates that homicide offenders only became central after the first homicide in T2, which suggests that the homicide may have helped increase their status in the gang, rather than the other way around.

Discussion and Conclusion

Although there is an increased interest in co-offending patterns, rarely is there an opportunity to consider the larger pool of accomplices available for specific crimes and compare these accomplices with those who were deemed suitable and chosen for the job (Tremblay, 1993). This line of research may be especially important for high stakes crimes such as homicide. The current study examined, from the point of view of a gang leader (GM1), the selection of co-offenders for two planned homicides occurring approximately ten years apart. The first homicide was committed against a rival gang member whereas the second homicide was committed against a member of the gang leader’s own gang. The current study hopes to have incited research on the process of co-offending selection as a way forward in the field. Yet, our contribution in this regard can only be presented as partial because the network was based on official convictions, offering a limited view of the full array of choices available to GM1. Caution should be used before generalizing to other gangs, especially considering that the BC Gang is located in an area of Canada with little racial segregation and low levels of poverty that may differ from the neighborhoods of gangs operating in the United States.

Although the research requires replication by studies with information on both detected and undetected offenses, we believe that the network of 137 offenders – including 35 BC Gang members – created around the conviction histories of 18 BC Gang members over a period of close to 20 years represents a reasonable approximation of the criminal social structure around the BC Gang leader. This network illustrates the complexity of the selection process and the diversity of co-offenders invited to participate in gang homicides, suggesting that a variety of roles were played, some of them relatively minor (e.g. “driver”) compared to others (e.g. “pulling the trigger”). Albeit imperfect, the data design revealed a series of important findings: (1) shared gang membership may be a homophily characteristic that influences the recruitment of co-offenders; (2) the vast majority of offenders recruited by GM1 were previously unconnected to the co-offending network; (3) all homicide offenders were in the same co-offending component, and; (4) by the final measurement period, the five homicide offenders ranked amongst the highest in betweenness and degree centrality within the initial 18 seeds.

Starting with the latter point, the most well-connected and established BC Gang members were not recruited by GM1 for his two homicide offenses, possibly because GM1 had a purposeful strategy of avoiding the selection of offenders who would be more likely to be on the police’s radar, making them riskier selections when considering detection avoidance. Recruiting members within the same gang for external homicides is less of a risk as these members share mutual interest in disrupting rival gangs. In contrast, for internal homicides a recruiter must be aware of within-gang allegiances. For GM1’s second homicide, he may have avoided recruiting the most central offenders in the network due to possible positive personal relationships with the victim, GM7.

Prior literature on the selection of homicide co-offenders is rare, although Morselli (2003) noted that the Gambino crime family required individuals to be of a specific status before becoming more deeply involved in lethal violence. To a certain extent, our findings resemble the Gambino crime family’s policy in that the founder and leader of the BC Gang was the one to orchestrate two of the murders. At the same time, homicide co-offenders in the current study included individuals who had no previous connection to the co-offending network. If we assume that trust was a factor in co-offender recruitment, nothing in the official data that we used could have helped establish this. Interviews with offenders about their social network, friendship data, or wiretap/surveillance data could be helpful in this case and may help support the importance of trusted personal relationships in recruitment (e.g., Tremblay, 2011; Waring, 2002).

Characteristics of homophily, such as gang membership, may be one indicator of trust (e.g., Weerman, 2003). Gang members never comprised more than 50% of the main component and yet were the only offenders recruited for homicide offenses. In effect, GM1 would recruit BC Gang members outside of his immediate co-offending network before recruiting non-BC Gang members within his own immediate network. Gang membership; however, was the only characteristic unanimously shared amongst all BC Gang members. GM1’s co-offenders were ethnically heterogeneous, which is dissimilar from the ethnic homophily found in other co-offender studies (e.g., Malm, Nash, & Vickovic, 2011). Although of different ethnic backgrounds, four of the five homicide offenders had been born outside of Canada.

Limitations and future research

Of the limitations mentioned above, three are worth reiterating. First, reliance on official data meant that the number of actual co-offending events was likely under-represented. The offenders selected by GM1 who appeared to be new to the network may have committed prior offenses with GM1 for which they were simply not caught. If this were the case, network proximity may have been important in GM1’s co-offender recruitment. Future research should investigate the relative “success” of a co-offending dyad and how that success may impact the likelihood of re-using the same co-offender. Studies using official data under-estimate the re-use of the same co-offenders for more than one crime, and over-emphasize the impression that co-offending groups are ephemeral. The truth is most likely somewhere in between. Second, the ties that individuals who were not part of the initial 18 BC Gang members had with other co-offenders could not be examined in the current study because of ethical restrictions. The inability to have the full network of BC Gang members might help explain why some of the individuals who were involved in homicide were selected as co-offenders despite appearing to have no prior connections to any of the initial 18 seeds. Third, with the exception of gang membership status, we only had access to the offender characteristics of the initial 18 seeds, which limited our interpretations and the type of analyses that could be done.

Involvement in gang homicides appear deeply rooted in the social networks in which gang members evolve. This assertion may seem obvious to the gang or network scholar, but it remains important to realize that this is not necessarily true of other types of homicides (not to the same extent, at the very least). Importantly, uncovering the social structure around gang homicide offenders naturally leads to practical implications, notably in helping narrow down suspects in gang-related homicides. Following the criminal trajectories of adolescents convicted of homicides post-release is also a worthwhile endeavour, as the murder for which they were convicted of may serve as a key status enhancement event that facilitates their transition into successful adult criminal careers.

Endnotes

_____________________________________

i To maintain confidentiality of the participants, the actual name of the gang could not be revealed.

ii Exact dates omitted to avoid identification of the gang and its members.

iii Due to the discrepancy in the number of years examined during each wave of data collection, instead of examining total increases in, for example, the number of gang members in the network, we looked at overall change in the percentage of gang members in the network, which would not be affected by differences in the number of years examined for a particular wave.

iv Murder charges rather than convictions were used because some offenders’ offenses had not been fully processed through the judicial system. Attempted murder was included as a murder offense because the intent of both offenses is the same; only the result is different.

v Correlation used as fit criterion. After multiple runs with cluster solutions of sizes 2 to 6, the best solution was chosen based on the highest Eta score.

vi There may have been others recruited who were not convicted; but the research design in the current study meant that only official convictions could be examined.

vii GM1’s ability to recruit without having any prior connections to his co-offenders, nor having any status in the network itself, may speak to his leadership abilities and influence he had over his co-offenders. Collateral reports from CORNET indicated that GM1 came from a wealthy family, traveled to many exotic locations, and drove a high-end sports car. These types of status symbols may have helped GM1 to recruit co-offenders.

viii GM5 was part of the initial 18 seeds. However, until T4 he had virtually no involvement in the co-offense network. He committed a few minor crimes in adolescence and these crimes did not involve any co-offenders (weapons possession). Thus, GM5 was a relatively inexperienced and unconnected actor in the network. Collateral information from CORNET suggested that GM5 was in a type of non-offending role that helped him facilitate this homicide.

References

Borgatti, S.P., Everett, M.G. and Freeman, L.C. (2002). Ucinet for Windows: Software for Social Network Analysis. Harvard, MA: Analytic Technologies.

Bouchard, M., & Konarski, R. (2013). Assessing the core membership of a youth gang from its co-offending network. In C. Morselli (Ed.). Crime and networks: Criminology and justice series. New York, NY: Routledge.

Curry, G.D., & Decker, S.H. (1998). Confronting gangs: Crime and community. Los Angeles, CA: Roxbury.

Everton, S. F. (2012). Disrupting dark networks. Cambridge, UK: Cambridge University Press.

Klein, M., Maxson, C. (2006). Street gang patterns and policies. New York: Oxford University Press.

Malm, A., Nash, R., & Vickovic, S. (2011). Co-offending networks in cannabis cultivation. In T. Decorte, G.R. Potter, & M. Bouchard. World wide weed: Global trends in cannabis cultivation and its control. (pp. 127-144). Surrey, UK: Ashgate Publishing.

Malm, A., Bichler, G., & Nash, R. (2011). Co-offending between criminal enterprise groups. Global Crime12(2), 112-128.

McCarthy, B., & Hagan, J. (2001). When crime pays: Capital, competence, and criminal success. Social Forces79(3), 1035-1060.

McGloin, J. M., & Nguyen, H. (2011, October). The importance of studying co-offending networks for criminological theory and policy. In Proc. Third Annual Illicit Networks Workshop, Montréal, Québec.

McGloin, J. M., & Piquero, A. R. (2010). On the relationship between co-offending network redundancy and offending versatility. Journal of Research in Crime and Delinquency47(1), 63-90.

McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social networks. Annual review of sociology, 415-444.

Morselli, C. (2003). Career opportunities and network-based privileges in the Cosa Nostra. Crime, Law and Social Change39(4), 383-418.

Morselli, C. (2009). Inside criminal networks. New York, NY: Springer.

Papachristos, A.V. (2009). Murder by structure: Dominance relations and the social structure of gang homicide. American Journal of Sociology115(1), 74-128.

Rosenfeld, R., Bray, T. M., & Egley, A. (1999). Facilitating violence: A comparison of gang-motivated, gang-affiliated, and non-gang youth homicides. Journal of Quantitative Criminology15(4), 495-516.

Tremblay, P. (1993). Searching for suitable co-offenders. In R.V. Clarke and M. Felson (Eds.), Routine activity and rational choice: Advances in criminological theory. New Brunswick, NJ: Transaction Publishers.

Tremblay, P. (2011). Beauvoir Jean, Le recit du vétéran. Montreal, QC: Liber Editions.

van Mastrigt, S.B., & Carrington, P.J. (2011). Sex and age homophily in co-offending networks: Opportunity or preference? In C. Morselli (Ed.), Crime and networks: Criminology and justice series. New York, NY: Routledge.

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.

Waring, E.J. (2002). Co-offending as a network form of social organization. In E. Waring & D. Weisburd (Eds.), Advances in criminological theory (pp. 31-48). New Brunswick, NJ: Transaction Publishers.

Warr, M. (1993). Age, peers, and delinquency. Criminology31(1), 17-40.

Weerman, F. (2003). Co-offending as social exchange: Explaining characteristics of co-offending. British Journal of Criminology, 43, 398-416.


Table 1. Offender characteristics for the initial 18 seeds

Offender Type

Age

Born in Canada

Ethnicity

Convictions

Violent Convictions

Age offense onset

Currently incarcerated

Drug Trafficking

Financial Motive

Cumulative NormalizedDegree Centrality

Cumulative Normalized Betweenness Centrality

Homicide Offenders

GM1

28

No

Asian

4

2

15

Yes

Yes

No

.154

.059

GM3

29

Yes

White

25

6

12

Yes

Yes

.

.118

.029

GM4

30

No

M- Eastern

13

5

16

Yes

No

.

.059

.006

GM5

33

No

Asian

4

3

16

No

No

No

.044

0

GM7

28

No

Asian

5

2

15

No††

Yes

No

.140

.051

Group x\ \overline{x}(sd)/n (%)

29.6 (2.1)

1 (20%)*

.

10.2 (9.1)

3.4 (2.1)

14.8 (1.5)

3 (60%)

3 (60%)

0 (0%)*

.103 (.047)

.029 (.026)*

Non-homicide offenders

GM6

29

Yes

White

43

3

15

No

Yes

Yes

.051

.002

GM8

31

Yes

Asian

24

4

15

Yes

No

No

.118

.016

GM9

25

No

Asian

9

3

16

Yes

Yes

Yes

.096

.016

GM10

29

Yes

Asian

17

2

16

No

Yes

Yes

.059

.001

GM13

23

Yes

White

5

1

14

No

No

Yes

.029

.010

GM21

28

Yes

M- Eastern

17

2

14

No

Yes

Yes

.037

.004

GM32

27

Yes

White

20

10

12

Yes

No

.

.051

.002

GM37

24

Yes

White

27

4

15

Yes

No

Yes

.037

0

GM70

25

No

Asian

1

1

15

No

No

Yes

.066

0

GM105

31

Yes

White

41

4

12

No

Yes

Yes

.184

.035

GM115

26

Yes

White

2

1

17

No

No

No

.022

0

GM121

32

Yes

White

39

11

12

No

Yes

Yes

.022

.003

GM126

22

Yes

White

32

3

15

Yes

No

Yes

.015

0

Group x\ \overline{x}(sd)/n (%)

27.1 (3.2)

11 (84.6%)*

.

21.3 (14.6)

3.8 (3.2)

14.5 (1.6)

5 (38.5%)

6 (46.2%)

10 (83.3%)*

.060 (.047)

.007 (.01)*

Middle Eastern

†† Deceased

Note. Significance tests compared homicide to non-homicide offenders using chi square or t-test depending on the variables level of measurement. * p <.05

Tables and Figures

Table 2. Description of the main co-offender component at each wave of data collection

 

T1

T2

T3

T4

Size of the full network

45

109

124

137

Network Density

0.05

0.04

0.04

0.04

Network Density-Main Component

0.36

0.2

0.16

0.15

Size of the main component

10

38

43

48

Main Component- Degree Centralization

80.95%

41.84%

30.98%

31.27%

Main Component - Betweenness Centralization

78.91%

23.60%

44.25%

44.49%

BC Gang - Main Component

1 (10%)

14 (37%)

19 (42%)

24 (50%)

BC Gang -Remainder of Network

8 (18%)

10 (14%)

10 (13%)

11 (12%)

Two components had ten nodes, we selected component with highest density

Table 3. Normalized centrality for GM1's homicide and non-homicide co-offenders and co-offenders and non-co-offenders

GM1’s homicide co-offenders

GM1’s non-homicide co-offenders

GM1’s co-offenders

Non-GM1 co-offenders

T2

Degree centrality

.07 (.04)

.06 (.03)

.075 (.05)*

.037 (.03)*

Betweenness centrality

.01 (.00)*

.00 (.00)*

.013 (.02)*

.000 (.00)*

T3

Degree centrality

.08 (.05)

.08 (.03)

.075 (.05)*

.029 (.02)*

Betweenness centrality

.01 (.02)

.00 (.00)

.013 (.02)

.000 (.00)*

T4

Degree centrality

.06 (.03)

.07 (.02)

.066 (.05)*

.029 (.02)*

Betweenness centrality

.01 (.02)

.00 (.01)

.011 (.02)*

.000 (.00)*

Note. * Indicates significant (p <.001) differences based on independent t-test. Used normalized degree and betweenness centrality

MURDER 1 COMPONENT NETDRAW.jpg

Figure 2a. Main component T2, four cluster solution. Clusters marked based on node color.

New W3.jpg

Figure 2b. Main component T3, four cluster solution. Clusters marked based on node color.

[CHART]

Figure 3a. Normalized betweenness and degree centrality for GM1, GM7, and the average for the other initial sixteen seeds

[CHART]

Figure 3b. Normalized betweenness and degree centrality for the initial eighteen seeds, separated by homicide involvement

Acknowledgements: An earlier version of this paper was presented at the Fifth Annual Illicit Networks Workshop, October 2013, held in Los Angeles, California. The authors would like to thank Pierre Tremblay, Gisela Bichler, the anonymous reviewers of JCCJ, and the participants of the 2013 Illicit Networks Conference in Los Angeles who provided helpful comments on an earlier draft of this paper. This work was supported by the Social Sciences and Humanities Council of Canada [410-2004-1875]. The authors gratefully acknowledge the assistance of the British Columbia Ministry of Children and Family Development. The views expressed herein are those of the authors and do not necessarily reflect the views or policies of the agencies that funded or supported the research.

Comments
0
comment
No comments here
Why not start the discussion?