Racial Differences in Co-Offending: Correlations With Concurrent Juvenile Police Contacts and Predictions of Future Adult Police Contacts

Abstract

The purpose of this study was to determine whether the relationship between co-offending and offense seriousness varied by race and whether similarities in age (juvenile, adult) and race (white, non-white) augmented the frequency and severity of future offending in co-offending males. Analyzing 15,059 incidents of police contact involving male juvenile participants from the Second Philadelphia Birth Cohort (PBC II) and the records of 7,420 male participants from the PBC II, a stronger co-offending–offense seriousness relationship was noted in the juvenile police contacts of non-white participants, whereas similarity between co-offenders led to increased adult police contacts in non-white but not white participants. These results suggest that juvenile co-offending may operate along social learning lines in non-white, if not white, youth.

Keywords

co-offending offense seriousness social learning

The demographic correlates of co-offending are, for the most part, well-known. Research has established, for instance, that co-offending peaks during adolescence and decreases during adulthood (Carrington, 2009; McGloin et al., 2008; Zimring & Lacqueur, 2015). It has also been reported that co-offending is more common in females than in males (Koons-Wit & Schram 2003; van Mastrigt & Farrington, 2009) and that it is more prevalent in certain types of offenses, such as robbery and burglary (Alarid et al., 2009; van Mastrigt & Farrington, 2011). A cross-national survey of studies conducted on co-offending in Canada, Great Britain, and the United States identified several points of similarity in co-offending across the three countries in terms of demographic correlates and group characteristics (Carrington & van Mastrigt, 2013). One demographic variable that has received only passing attention in the research literature on co-offending is race or ethnic status. In one of the few studies to address the effect of race on co-offending, Lantz and Wenger (2020) discovered that black co-offenders were more likely to be arrested for their actions than their white co-offending counterparts. Results from a second study showed that co-offending was more common in racially homogeneous groups than in racially heterogeneous groups (Schaefer et al., 2014). Quite obviously, there is a need for more research on the issue of race and co-offending.

Co-Offending and Crime Seriousness

A review of the literature on co-offending and crime severity indicates that crime incidents marked by co-offending tend to have more serious consequences for victims, while posing a greater threat to the community, than crime incidents in which offenders act alone. Findings from several early studies, in fact, revealed that co-offending generated higher levels of property loss and victim harm than solo offending (Carrington, 2002; Conway & McCord, 2002; Felson, 2003). Reviewing data from the National Incident Based Reporting System (NIBRS), Tillyer and Tillyer (2015) discovered that co-offending produced greater total profit over the course of a robbery, but that profit per offender was lower and the odds of arrest higher for those who engaged in co-offending than for those who engaged in solo offending. Lantz (2020) also used NIBRS data to evaluate the relationship between co-offending and crime severity and discovered that co-offending was associated with higher levels of serious victim injury. The introduction of a male into a small group of co-offenders significantly augmented the risk of serious victim injury, although groups of five or more offenders produced high rates of serious victim injury, regardless of the gender composition of the group. There is no research, however, on whether race moderates the co-offending–crime seriousness relationship, a gap the current study sought to fill.

Co-Offending as a Social Learning Process

With the advent of Sutherland’s (1947) differential association theory of crime, peers are seen as an important correlate, if not cause, of crime (Warr, 2002). In fact, a number of peer-related factors have been found to predict future delinquency, to include delinquent peer associations (Gifford-Smith et al., 2005), gang membership (Esbensen & Huizinga, 1993), and unsupervised routine activities (Haynie & Osgood, 2005). Co-offending may also belong in this group. Dynes et al. (2015), for example, discovered that co-offending moderated the relationship between peer delinquency and juvenile offending such that youth peer delinquency was a stronger correlate of juvenile offending when concurrent co-offending was high. Likewise, McCuish et al. (2015) determined that gang members frequently selected one another as crime partners. McGloin (2012), for her part, notes that unsupervised routine activities in a group can lead to delinquency independent of the level of criminality in one’s peer network. Controlling for delinquent peer associations, gang membership, and unsupervised routine activities, Walters (2020a) observed a significant indirect association between co-offending and delinquency via an increase in antisocial cognition. The latter finding suggests that social learning processes may be at work in connecting peer-related factors like co-offending to delinquency. The social learning effect or process most likely involved in co-offending is observational learning or modeling (Bandura, 1986).

Present Study

There was a two-fold purpose to this study. The first purpose or objective was to determine whether race moderated the well-documented relationship between co-offending and offense severity. The results of a recent study by Lantz (2020) revealed that white co-offender status was associated with decreased levels of minor and serious injury. Based on these findings, it was speculated that race may be capable of moderating the correlation between co-offending and offense severity, with the relationship being significantly stronger for incidents involving non-white as opposed to white participants. In testing this hypothesis, the number of persons victimized and the use of a firearm during the incident were controlled. This was done in order to account for aspects of the situation that might confound co-offending and incident severity with increased opportunities for offending (e.g., need more perpetrators to control a larger group; choosing to co-offend with an older individual because they have access to a firearm). It was hypothesized that the relationship between co-offending and incident severity would be stronger in non-white juveniles than it is in white juveniles after controlling for the socioeconomic status of the youth’s home of origin, the number of victims, and the use of a firearm.

The second objective pursued in this study was to determine whether similarity in age and race between a youth and his co-offenders contributes to a rise in future police contacts. Bandura (1994) and Schunk (1987), among others, have speculated that model–observer similarity (MOS) may enhance the effect of social and observational learning. Sex is the demographic variable most often examined in research on the MOS, but the results of most of these studies have either been negative or inconclusive (Bussey & Bandura, 1984; Gallupe et al., 2019; Hoogerheide et al., 2018; Schunk et al., 1987; Walters, 2020b). Consequently, two other demographic variables that are less frequently found in research on the MOS were examined in the current study: age (juvenile, adult) and race (white, non-white). The second hypothesis consequently proposed that high similarity between co-offenders (juvenile/co-offender match on both age and race) would do a significantly better job of predicting future police contact than moderate similarity between co-offenders (juvenile/co-offender match on either age or race), and that moderate similarity between co-offenders would do a significantly better job of predicting future police contact than low co-offender similarity (juvenile/co-offender match on neither age nor race).

Method

Participants

All male members of the Second Philadelphia Birth Cohort 1958 to 1988 (PBC-II: Figlio et al., 2006) were selected to serve as participants in this study. These 13,160 individuals were born in 1958 and maintained residence in Philadelphia, Pennsylvania from at least age 10 to age 18. The racial/ethnic breakdown of the male portion of the PBC-II was 47.2% white (n = 6,216) and 52.8% non-white (n = 6,944). The standardized continuous score for the socioeconomic status of the home where the juvenile was residing at the time of the police contact ranged from −2.24 to +3.04 (M = 0.02, SD = 0.99) and annual family income fell into the following five categories: <$4,000 (10.2%), $4,001 to $4,500 (5.4%), $4,501 to $5,783 (30.5%), 5,784 to $6,779 (30.1%), and >$6,779 (23.8%).

Measures

All variables in the PBC-II were constructed from information found in official police, court, and school records. The current study focused on two variables, in particular: police contacts and co-offending. Police contacts were assessed using a severity scale that took into account the degree of injury, damage, financial loss, forcible entry, intimidation, and physical or sexual violation involved in the specific incident. In the first set of analyses, the severity score for each juvenile police contact served as the dependent variable. In the second set of analyses, the sum of severity scores across all adult police contacts for a single individual served as the dependent variable.

Co-offending was rated as present (1) if one or more co-offenders were identified in the police files and absent (0) if no co-offenders were identified in the police files. Irrespective of co-offender age (adult, juvenile), race (white, non-white), or number, co-offending was coded 1 if there were one or more co-offenders mentioned in the police file for a particular incident and 0 if there were no co-offenders mentioned in the police file for a particular incident. For the second set of analyses, co-offending with adult white males, adult non-white males, juvenile white males, and juvenile non-white males were scored dichotomously (1 = present, 0 = absent) to reflect whether the participant had a co-offender in each of these categories during adolescence.

Control variables for the first set of analyses included age at time of police contact, socioeconomic status (SES) of the home of residence, total number of victims, and use of a gun during the incident. The latter two variables were designed to control for opportunity factors in co-offending (i.e., use of co-offenders to control a large group of people and solicitation of older co-offenders with access to firearms). The control variable for the second set of analyses was the severity and frequency of police contacts (summed score) during adolescence.

Procedure

There were two data collection periods. The juvenile period included all police contacts up to age 17. The small number of police contacts in the juvenile file that occurred at ages 18 to 19 were removed from the analysis. Police contacts during the adult period occurred when participants were between the ages of 18 and 26. Similarity between participants and co-offenders was classified as high (white juvenile participants and white juvenile co-offenders; non-white juvenile participants and non-white juvenile co-offenders), moderate (white juvenile participants and non-white juvenile co-offenders or white adult co-offenders; non-white juvenile participants and white juvenile co-offenders or non-white adult co-offenders), and low (white juvenile participants and non-white adult co-offenders; non-white juvenile participants and white adult co-offenders).

Data Analytic Plan

Descriptive statistics, correlations, and multicollinearity were computed with SPSS, Version 26 (IBM Corporation, 2019) and regression analyses were performed with MPlus 8.3 (Muthén & Muthén, 1998–2017). Because police contacts were nested within participants (i.e., some participants had more than one police contact) in the first set of analyses, a complex multiple regression analysis was performed with participants serving as the cluster variable. Non-independence was not a problem in the second set of analyses because participants contributed only one line of data to the analysis. As such, standard regression analysis was computed using a maximum likelihood (ML) estimator. Comparisons between coefficients in the same regression equation were made by calculating a Z-test of the difference between coefficients as described in Paternoster et al. (1998).

Missing Data

Over three-quarters of the incidents from the first set of analyses (77.6%) had complete data on all seven variables; 20.5% were missing data on one variable, 1.6% were missing data on two variables, and 0.3% were missing data on three variables. It should be noted that one in five incidents (20.5%) were missing co-offender information. The vast majority of participants in the second set of analyses had complete data on all eight study variables (97.4%). Nearly all of the participants for whom data were missing (2.5%) were missing data on the four co-offending measures. Missing data in both analyses were handled with full information maximum likelihood (FIML), a procedure that estimates population parameters and standard errors from relationships between non-missing data (Allison, 2002).

Propensity Score Matching

For the second set of analyses (juvenile co-offending as a predictor of adult police contacts), white and non-white participants were paired using propensity score matching. Propensity scores were calculated by regressing racial status (white, non-white) onto six background/demographic variables found in participants’ school (family SES, annual family income, number of changes in residence while in school, number of years of school completed, IQ score) and police (age at time of first police contact) records. Prior to conducting the logistic regression analysis, missing data (range = 0.0%–37.5%) were imputed using expectation maximization, a procedure similar to FIML. The resulting propensity scores were then used to match white and non-white participants using the SPSS Propensity Score Fuzzy Matching program. A match tolerance of .2 was selected because it was the first tenths decimal to match over half of all participants (n = 7,420 or 56.4% of the total sample of 13,160 participants).

Results

Police Contacts in Adolescence

Table 1 lists descriptive statistics for all the variables included in the first set of analyses in which official incidents of police contact were analyzed for all male members of the PBC-II. Numbers of incidents, means, standard deviations, and ranges are organized into separate columns in Table 1, with white and non-white male participant data presented and analyzed separately. All told, there were 3,907 incidents leading to police contact for 2,022 white male participants (range = 1–37) and 11,335 incidents leading to police contact for 4,263 non-white male participants (range = 1–53).

Table 1.

Descriptive Statistics for Police Contact Incidents for Male White and Non-White Male Juvenile Participants.

	White males				Non-white males
Variable	N	M	SD	Range	N	M	SD	Range
Age	3,872	14.93	2.01	2–17	11,187	14.67	2.06	3–17
SES	3,872	0.28	0.80	−1.9–2.1	11,187	−0.71	0.79	−2.2–2.1
Total victims	3,793	0.39	1.83	0–40	10,866	0.42	1.19	0–78
Gun	3,862	0.01	0.11	0–1	11,138	0.04	0.20	0–1
Co-offenders	3,268	1.71	2.66	0–35	8,584	1.31	2.29	0–34
Seriousness score	3,872	4.64	5.77	0–113.5	11,187	5.83	7.12	0–87.3

Note. Age = chronological age at time of police contact, SES = socioeconomic status, Total Victims = total number of victims, Gun = use of a gun during incident, Seriousness Score = seriousness score for incident, N = number of incidents, M = mean, SD = standard deviation, Range = range of scores in White and Non-White samples.

Analyzing police contacts recorded for males in the PBC-II, it was noted that race and the presence of one or more co-offenders interacted (see first set of columns in Table 2). Consequently, analyses were performed on white (second set of columns in Table 2) and non-white (third set of columns in Table 2) participants separately. The results of this analysis revealed that having a co-offender correlated positively with the estimated severity of the alleged criminal act in both white and non-white participants, although the effect was significantly stronger in non-white participants, as represented by a steeper slope for the non-white regression line compared to the white regression line in Figure 1.

Table 2.

Predicting the Severity of Incidents Leading to Police Contact for All Male, White Male, and Non-White Male Juveniles in the PBC-II.

	All males (n = 15,059)			White males (n = 3,872)			Non-white males (n = 11,187)
Predictor	b	se	Z	b	se	Z	b	se	Z
Age	0.354	0.027	13.25**	0.202	0.040	5.00**	0.391	0.040	9.69**
SES	−0.212	0.074	−2.87*	−0.331	0.126	−2.63*	−0.162	0.089	−1.82
Total victims	1.237	0.298	4.15**	0.832	0.194	4.30**	1.583	0.626	2.53*
Gun	7.096	0.548	12.95**	3.798	1.405	2.70*	7.187	0.669	10.74**
Race	0.847	0.149	5.68**
Co-offender	1.642	0.115	14.30**	0.805	0.194	4.14**	1.851	0.160	11.56**
Race × CO	0.235	0.054	4.37**

Note. Outcome Variable = rated seriousness of incident leading to police contact, Age = chronological age at time of police contact, SES = socioeconomic status, Total victims = total number of victims, Gun = use of a gun during incident, Race = racial status (1 = White, 2 = non-White), Co-offender = one or more co-offenders present during incident (1) vs. no co-offenders present during incident (0), Race × CO = race × co-offender interaction, N/n = number of police contacts, b = unstandardized coefficient, se = standard error, Z = Wald’s Z test.

p < .05, **p < .001.

Figure 1.

Interaction between race and co-offender status on seriousness of juvenile police contact incidents.

Predicting Adult Police Contacts With Juvenile Co-Offenders

Propensity matched white and non-white participants (n = 3,710 each; match tolerance = .2) were used in this second set of analyses. Table 3 provides a breakdown of the proportion of white and non-white juvenile male participants from the PBC-II who had white adult male, non-white adult male, white juvenile male, and non-white juvenile male co-offenders. These results demonstrate that white juvenile males were nearly four times as likely to co-offend with another white juvenile male than with the next most common group of co-offenders (i.e., white adult males) and that non-white juvenile males were four times more likely to co-offend with another non-white juvenile male than with a non-white adult male, the next most common category of co-offender for non-white juvenile males.

Table 3.

Number and Proportion of White and Non-White Male Juveniles With White Adult Male, Non-White Adult Male, White Juvenile Male, and Non-White Juvenile Male Co-Offenders.

	White male juveniles(n = 3,710)		Non-white male juveniles(n = 3,710)		Difference
Co-offender group	n	%	n	%	χ² (1 df)
White adult males	156	4.3%	23	0.6%	99.61**
Non-white adult males	22	0.6%	216	6.0%	166.43**
White juvenile males	612	16.8%	89	2.5%	424.76**
Non-white juvenile males	113	3.1%	858	23.9%	672.78**

Note. Co-offenders represent the rows whereas white and non-while male juveniles represent the first two sets of columns; n = number of white and non-white male juveniles with co-offenders who were white adult males, non-white adult males, white juvenile males, and non-white juvenile males; % = proportion of white and non-white male juveniles with co-offenders who were white adult males, non-white adult males, white juvenile males, and non-white juvenile males; χ² (1 df) = chi-square test with one degree of freedom, white and non-white male juveniles selected using propensity matching (percentages are based on an N < 3,710 because of missing data).

p < .001.

A variable measuring the severity and frequency of adulthood police contacts was constructed and then regressed onto prior white adult, non-white adult, white juvenile, and non-white juvenile co-offenders, after controlling for the severity and frequency of juvenile police contacts. A preliminary analysis revealed that race interacted with adult non-white co-offenders (Z = −3.33, p < .001) and juvenile non-white co-offenders (Z = −2.17, p = .030). As such, the second set of analyses (i.e., those pertaining to the second hypothesis) were conducted separately for white and non-white participants (see Table 4). Collinearity diagnostics were performed, the results of which failed to disclose evidence of multicollinearity in either of the two regression equations (tolerance = .667–.923, variance inflation factor = 1.083–1.500).

Table 4.

Predicting Adult Police Contacts with Four Categories of Juvenile Co-Offending and Juvenile Police Contacts.

	White males (n = 3,710)				Non-white males (n = 3,710)
Predictor	b	se	Z	β	b	se	Z	β
Adult white co-offender	−1.402	1.189	−1.18	−0.019	−3.509	4.079	−0.86	−0.014
Adult non-white co-offender	12.858	2.934	4.38**	0.068	4.173	1.391	3.00*	0.050
Juvenile white co-offender	0.404	0.662	0.61	0.010	4.679	2.123	2.20*	0.037
Juvenile non-white co-offender	3.448	1.402	2.46*	0.041	5.042	0.836	6.03**	0.109
Juvenile police contacts (F + S)	0.437	0.019	22.59**	0.408	0.216	0.015	14.25**	0.260

Note. White and non-white participants selected using propensity matching, Outcome Variable = Adult Police Contacts (frequency + seriousness), Adult White Co-Offender = presence of one or more adult white co-offenders in one or more juvenile incidents (1) vs. no adult white co-offenders in one or more juvenile incidents (0), Adult Non-White Co-Offender = presence of one or more adult non-white co-offenders in one or more juvenile incidents (1) vs. no adult non-white co-offenders in one or more juvenile incidents (0), Juvenile White Co-Offender = presence of one or more juvenile white co-offenders in one or more juvenile incidents (1) vs. no juvenile white co-offenders in one or more juvenile incidents (0), Juvenile Non-White Co-Offender = presence of one or more juvenile non-white co-offenders in one or more juvenile incidents (1) vs. no juvenile non-white co-offenders in one or more juvenile incidents (0), Juvenile Police Contacts (F + S) = juvenile police contacts (frequency + seriousness).

p < .05. **p < .001.

A comparison of coefficients under the White Male column in Table 4 failed to produce a single outcome consistent with Hypothesis 2. Prior co-offending with white juveniles failed to predict adult police contacts better than prior co-offending with white adults, Z = 1.25, p = 211, or prior co-offending with non-white juveniles, Z = −1.29, p = .197, and it did a significantly worse job of predicting adult police contacts than co-offending with non-white adults, Z = −2.49, p = .012. Moreover, prior co-offending with non-white juveniles was no more predictive of adult police contacts than prior co-offending with non-white adults, Z = −1.16, p = .246, and while the difference between prior co-offending with white adults and non-white adults was significant, it was in the opposite direction of what had been predicted, Z = −3.85, p < .001. The fact that co-offending with non-white juveniles was significantly more predictive of adult police contacts than co-offending with white adults, Z = 2.58, p = .010, was also inconsistent with predictions.

In contrast to the uniformly negative results obtained with white participants, non-white participants attained results uniformly consistent with predictions. Prior co-offending with non-white juveniles did a significantly better job of predicting adult police contacts than prior co-offending with white juveniles, Z = 2.90, p = .004, prior co-offending with non-white adults, Z = 2.38, p = .017, and prior co-offending with white adults, Z = 4.96, p < .001. Likewise, prior co-offending with white juveniles, Z = 2.12, p = .034, and prior co-offending with non-white adults, Z = 2.67, p = .008, correlated significantly better with adult police contacts than prior co-offending with white adults. Also consistent with predictions, prior co-offending with white juveniles and prior co-offending with non-white adults were equally predictive of adult police contacts, Z = 0.54, p = .589.

Because white and non-white participants differed significantly (p < .001) on all six social/background variables used to match participants, the next largest sample in which there were no significant differences (p > .05) between white and non-white participants on these six variables was analyzed. This required a match tolerance of .00001 and a sample of 812 participants (406 whites, 406 non-whites; 6.2% of the total sample). There were no significant predictive effects in the regression equation conducted on white participants, whereas non-white juvenile co-offending achieved a significant predictive effect in non-white participants, Z = 3.51, p < .001. In the latter analysis, prior co-offending with non-white juveniles achieved a significantly stronger predictive effect than prior co-offending with white juveniles, Z = 2.12, p < .05, and prior co-offending with white adults, Z = 3.71, p < .001, whereas prior co-offending with non-white adults was significantly more predictive of adult police contacts than prior co-offending with white adults, Z = 2.05, p < .05. The other two effects were in the predicted direction—non-white juveniles > non-white adults and white juveniles > white adults—but both effects fell short of significance, Z = 1.61, p = .11.

Discussion

The first hypothesis tested in this study held that race would moderate the relationship between juvenile co-offending and the severity of concurrent incidents of police contact. The results of a complex multiple regression analysis revealed that while co-offending was associated with incident severity in both white and non-white juvenile participants, the effect was significantly stronger in non-white participants. This may have something to do with the fact that guns were four times more likely to be present in incidents involving non-white as opposed to white participants and seriousness scores were 25% higher in the non-white group. Race played a leading role in the second hypothesis as well, given that the similarity pattern predicted for co-offending (i.e., high similarity to co-offenders > moderate similarity to co-offenders > low similarity to co-offenders in predicting the frequency and severity of future police contact) emerged in non-white but not white respondents regardless of where the match tolerance was set. In white participants, the low similarity pattern (adult non-white co-offenders) correlated better with future police contacts than with several higher similarity patterns, though not always significantly, despite the fact it was the least prevalent pattern (0.6%) in white youth.

The Psychology of Co-Offending

In the current study, co-offending was found to increase the severity of individual police contacts regardless of participant race. Incidents marked by co-offending were significantly more serious in terms of injury, intimidation, physical violation, property damage, and financial loss than incidents marked by solo offending. Based on these and other findings, however, it would appear that co-offending provides few benefits to the perpetrator and more harm to the victim. For the perpetrator, it increases the odds of arrest and reduces the profit that each offender can expect from their involvement in a specific criminal event or antisocial act (Tillyer & Tillyer, 2015). For the victim, it means an increased likelihood of injury and greater potential for lost or damaged property (Lantz, 2020; Tillyer & Tillyer, 2015). The increased likelihood of victim injury and augmented financial harm means that the offender is likely to receive a more severe sanction in the event they are caught, convicted, and sentenced for the offense. From an evolutionary standpoint, one has to wonder how co-offending has survived this long, given that it is a flawed survival strategy that should have been abandoned long ago. One possibility is that co-offending has survived because of the psychological benefits it provides rather than because it is an efficient predatory or acquisitive strategy.

One psychological benefit of co-offending is that it gives youth, many of whom would not commit a delinquent act on their own, the impetus to engage in criminal behavior. In presenting his theory of delinquency and co-offending, Reiss (1988) emphasized the role of co-offending in recruiting and socializing youth into antisocial peer networks. An added psychological benefit of co-offending is diffusion of responsibility. The prospect of reduced felt responsibility for antisocial behavior not only attracts youth to co-offending groups and gangs, it also serves to increase crime severity (Lantz, 2018). Diffusion of responsibility reflects cognitive strategies designed to neutralize (Sykes & Matza, 1957) and morally disengage (Bandura et al., 1996) from the negative consequences of one’s criminal behavior. The current results indicate that co-offending correlates with offense severity in both white and non-white youth, although the effect appears to be stronger in non-white youth. Given that non-white youth engaged in more serious offending than white youth, it may be that they required co-offending more than white youth to neutralize the negative consequences of their antisocial actions. It is also possible that co- and participant offending are reciprocally related, such that an increase in one, leads to an increase in the other.

Co-Offending and Social Learning

When juvenile co-offending was used to predict the frequency and severity of adult police contacts, after the frequency and severity of juvenile police contacts had been controlled, there was evidence of model–observer similarity, although it was restricted to non-white participants. The presence of model–observer similarity is consistent with early MOS research (Bandura, 1994; Schunk, 1987) but inconsistent with studies that have assessed similarity based on sex (Bussey & Bandura, 1984; Gallupe et al., 2019; Hoogerheide et al., 2018; Schunk et al., 1987; Walters, 2020b). It could also be argued that outcomes obtained with non-white participants reflect an exposure effect in that the high similarity group (juvenile non-white males) was the group with whom non-white juvenile participants co-offended most often and the low similarity group (adult white males) was the group with whom non-white juvenile participants co-offended least often. Differential exposure, it should be noted, is as consistent with a social learning interpretation of the current results as the MOS effect and embodies aspects of two concepts central to social learning theories of crime: that is, differential association and differential reinforcement (Akers, 1998; Sutherland, 1947). Walters (2020a), in fact, discovered that moral disengagement, another construct congruent with a social learning interpretation of delinquency development, mediated the relationship between co-offending and participant offending in a sample of serious male delinquents, 81% of whom were non-white.

The results attained with white participants were significantly different from those achieved with non-white participants. Although there were only two significant differences between low, moderate, and high MOS levels, neither of which was in the predicted direction, the group to which white participants had the least exposure and lowest similarity (adult non-white males) predicted future police contacts better than the group to which white participants had the most exposure and highest similarity (juvenile white males). In fact, the two groups to which white juveniles had the least exposure (juvenile and adult non-white males) correlated highest with future police contacts. Neighborhood context has been found to be important in the development of youth co-offending, in that co-offending is at its height when neighborhoods are less disadvantaged, more stable, and less racially and ethnically diverse (Schaefer et al., 2014). This may explain why white youth associated more often with white juvenile and adult male co-offenders, but it does not explain why non-white juvenile and adult male co-offenders had a stronger impact on white youth’s propensity for future police contact. Certain possibilities nonetheless exist. First, this may reflect police bias to the extent that police were more likely to stop white juveniles while in the company of non-white adults and adolescents based on studies showing racial bias in police stops (Borooah, 2011; Chanin et al., 2018; Hayle et al., 2016). Second, given that youth tend to co-offend with those with whom they are in close proximity (i.e., others from their neighborhood), it may be that these white youth lived in racially mixed neighborhoods where they were the minority and thus felt compelled to prove themselves through more extreme forms of criminal behavior.

Limitations

There are several limitations to this study that should be noted. First, race was only broken down into two categories: white and non-white. It would be informative to know if these effects relate equally well to Hispanic and African American youth or if they are more applicable to one race/ethnicity than another. The study was also limited by the fact that it took place in one U.S. city, Philadelphia, Pennsylvania. This raises questions about how well the current results generalize to populations in other U.S. cities and states or to other countries, for that matter. Similarly, the PBC-II data are between forty and fifty years old and are thus limited by the fact that offending patterns may have changed significantly in the meantime. A fourth limitation of this study is that the variables in the PBC-II were based exclusively on official police, court, and school records. There is a large dark figure of unreported crime that makes the use of official data potentially problematic (Lynch & Addington, 2007), despite the thoroughness and relative completeness of the PBC-II data. A fifth limitation of this study is that there was no information in the PBC-II on the nature of the relationships between youth and their co-offenders. It is generally agreed that co-offending relationships, like gang membership, is often brief, ephemeral, and transitory, although some of these associations are more long-lasting (Reiss & Farrington, 1991). More than just the presence and number of co-offenders, it would be helpful to have information on the nature, extent, and durability of these relationships in interpreting some of the findings from this study.

Conclusion and Future Directions for Research

Co-offending was associated with increased severity of concurrent offending incidents regardless of race, although the effect was stronger for non-white youth than it was for white youth. Race also played a role in the relationship between juvenile co-offending and future police contact. Additional research is required, however, if we hope to develop a clear picture of how these processes contribute to delinquency development and the transition from juvenile delinquency to adult criminality. Social learning factors may be of cardinal significance in explaining the effect of co-offending on future participant criminal involvement, although the effect seems more straight-forward in non-whites than it is in whites, at least when it comes to studying the similarity and differential exposure effects central to social learning theories of crime. One direction for future research would be clarification of the social learning process that gives rise to juvenile offending. This could then be supplemented by research on the role of social learning factors in the transition from juvenile delinquency to adult offending. We know that co-offending decreases as people age, particularly when they transition from adolescence to adulthood. Could this be partly the result of changes occurring in the sequence of social learning influences that support adult offending separate from juvenile delinquency?

Footnotes

Declaration of Conflicting Interests

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

Funding

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

ORCID iD

Glenn D. Walters

Author Biography

Glenn D. Walters, PhD, is a professor in the Department of Criminal Justice at Kutztown University where he teaches classes in corrections, criminology, research methods, and substance abuse and crime. His principal research interests include delinquency development, mediation analysis, and the relationship between peers and early offending behavior.

References

Akers

R. L.

(1998). Social learning and social structure: A general theory of crime and deviance. Northeastern University Press.

Alarid

L. F.

Burton

V. S.

Hochstetler

A. L.

(2009). Group and solo robberies: Do accomplices shape criminal form? Journal of Criminal Justice, 37, 1–9.

Allison

P. D.

(2002). Missing data. Sage.

Bandura

(1986). Social foundations of thought and action: A social cognitive theory. Prentice-Hall.

Bandura

(1994). Self-efficacy. In Ramachaudran

V. S.

(Ed.), Encyclopedia of human behavior (Vol. 4, pp. 71–81). Academic Press.

Bandura

Barbaranelli

Caprara

G. V.

Pastorelli

(1996). Mechanisms of moral disengagement in the exercise of moral agency. Journal of Personality and Social Psychology, 71(2), 364–374.

Borooah

V. K.

(2011). Racial disparity in police stop and searchers in England and Wales. Journal of Quantitative Criminology, 27, 453–473.

Bussey

Bandura

(1984). Influence of gender constancy and social power on sex-linked modeling. Journal of Personality and Social Psychology, 47, 1292–1302.

Carrington

P. J.

(2002). Group crime in Canada. Canadian Journal of Criminology and Criminal Justice, 44, 277–316.

10.

Carrington

P. J.

(2009). Co-offending and the development of the delinquent career. Criminology, 47(4), 1295–1329.

11.

Carrington

P. J.

van Mastrigt

S. B.

(2013). Co-offending in Canada, England, and the United States: A cross-national comparison. Global Crime, 14, 123–140.

12.

Chanin

Welsh

Nurge

(2018). Traffic enforcement through the lens of race: A sequential analysis of post-stop outcomes in San Diego, California. Criminal Justice Policy Review, 29, 561–583.

13.

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.

14.

Dynes

M. E.

Domoff

S. E.

Hassan

Tompsett

C. J.

Amrhein

K. E.

(2015). The influence of co-offending within a moderated mediation model of parent and peer predictors of delinquency. Journal of Child and Family Studies, 24, 3516–3525.

15.

Esbensen

F.-A.

Huizinga

(1993). Gangs, drugs, and delinquency in a survey of urban youth. Criminology, 31, 565–589.

16.

Felson

(2003). The process of co-offending. In Smith

M. J.

Cornish

D. B.

(Eds.), Theory for practice in situational crime prevention (pp. 149–167). Criminal Justice Press.

17.

Figlio

R. M.

Tracy

P. E.

Wolfgang

M. E.

(2006). Delinquency in a Birth Cohort II: Philadelphia, 1958–1988. Inter-university Consortium for Political and Social Research [distributor], 2006-01-12. https://doi.org/10.3886/ICPSR09293.v3

18.

Gallupe

LaLonde

Nguyen

Schulenberg

(2019). A randomized trial testing deviant modeling, peer gender, and theft: Replication and extension. Journal of Experimental Criminology, 15, 421–439.

19.

Gifford-Smith

Dodge

K. A.

Dishion

T. J.

McCord

(2005). Peer influence in children and adolescents: Crossing the bridge from developmental to intervention science. Journal of Abnormal Child Psychology, 33, 255–265.

20.

Hayle

Wortley

Tanner

(2016). Race, street life, and policing: Implications for racial profiling. Canadian Journal of Criminology and Criminal Justice, 58, 322–353.

21.

Haynie

D. L.

Osgood

D. W.

(2005). Reconsidering peers and delinquency: How do peers matter? Social Forces, 84(2), 1109–1130.

22.

Hoogerheide

van Wermeskerken

van Nassau

van Gog

(2018). Model-observer similarity and task-appropriateness in learning from video modeling examples: Do model and student gender affect test performance, self-efficacy, and perceived competence? Computers in Human Behavior, 89, 457–464.

23.

IBM Corporation. (2019). IBM SPSS Statistics for Windows, Version 26.0. IBM Corporation.

24.

Koons-Witt

B. A.

Schram

P. J.

(2003). The prevalence and nature of violent offending by females. Journal of Criminal Justice, 31, 361–371.

25.

Lantz

(2018). The consequences of crime in company: Co-offending, victim–offender relationship, and quality of violence. Journal of Interpersonal Violence. On-line first: https://doi.org/10.1177/0886260518786497

26.

Lantz

(2020). Co-offending group composition and violence: The impact of sex, age, and group size on co-offending violence. Crime and Delinquency, 66, 93–122.

27.

Lantz

Wenger

M. R.

(2020). The co-offender as counterfactual: A quasi-experimental within-partnership approach to the examination of the relationship between race and arrest. Journal of Experimental Criminology, 16, 183–206.

28.

Lynch

J. P.

Addington

L. A.

(Eds.). (2007). Understanding crime statistics: Revisiting the divergence of the NCVS and the UCR. Cambridge University Press.

29.

McCuish

E. C.

Bouchard

Corrado

R. R.

(2015). The search for suitable homicide co-offenders among gang members. Journal of Contemporary Criminal Justice, 31(3), 319–336.

30.

McGloin

J. M.

(2012). Delinquency balance and time use: A research note. Journal of Research in Crime and Delinquency, 49(1), 109–121.

31.

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(1), 155–187.

32.

Muthén

(1997–2017). Mplus user’s guide (8th ed.). Muthén and Muthén.

33.

Ouellet

Boivin

Leclerc

Morselli

(2013). Friends with(out) benefits: Co-offending and re-arrest. Global Crime, 14(2–3), 141–154.

34.

Paternoster

Brame

Mazerolle

Piquero

(1998). Using the correct statistical test for the equality of regression coefficients. Criminology, 36, 859–866.

35.

Reiss

A. J.

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

Tonry

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

36.

Reiss

A. J.

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.

37.

Schaefer

D. R.

Rodriguez

Decker

S. H.

(2014). The role of neighborhood context in youth co-offending. Criminology, 52, 117–139.

38.

Schunk

D. H.

(1987). Peer models and children’s behavioral change. Review of Educational Research, 57, 149–174.

39.

Schunk

D. H.

Hanson

A. R.

Cox

P. D.

(1987). Peer-model attributes and children’s achievement behaviors. Journal of Educational Psychology, 79, 54–61.

40.

Sutherland

E. H.

(1947). Principles of criminology (4th ed.). J.B. Lippincott.

41.

Sykes

G. M.

Matza

(1957). Techniques of neutralization: A theory of delinquency. American Sociological Review, 22, 664–670.

42.

Tillyer

M. S.

Tillyer

(2015). Maybe I should do this alone: A comparison of solo and co-offending robbery outcomes. Justice Quarterly, 32, 1064–1088.

43.

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.

44.

van Mastrigt

S. B.

Farrington

D. P.

(2011). Prevalence and characteristics of co-offending recruiters. Justice Quarterly, 28, 325–359.

45.

Walters

G. D.

(2020a). Moral disengagement as a mediator of the co-offending–delinquency relationship in serious juvenile offenders. Law and Human Behavior, 44, 437–448.

46.

Walters

G. D.

(2020b). Parental control as a stimulus for youth control: Mediating the parental knowledge-child delinquency relationship with cognitive impulsivity. Crime and Delinquency. On-line first: https://doi.org/10.1177/0011128720962449

47.

Warr

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

48.

Zimring

F. E.

Laqueur

(2015). Kids, groups, and crime: In defense of conventional wisdom. Journal of Research in Crime and Delinquency, 52, 403–413.