Differentiating between delinquent groups and gangs: moving beyond offending consequences

Abstract

Even when controlling for high levels of delinquent peers, gang youth differ from their nongang counterparts on a variety of attitudinal and behavioral measures. Researchers have argued that differences can be attributed to the group processes present in the gang setting. This study explores the extent to which differences between youth in a gang and those in a delinquent group can be explained by Bandura’s social cognitive theory. Much of the prior research in this arena has relied on cross-sectional data; in this study, we expand on this prior research using fixed-effects modeling strategies with a multi-site panel of youth. The results comparing time periods when youth were in a gang versus a delinquent peer group indicate that gang-involved youth are more violent and have fewer conventional bonds. This work is able to advance our knowledge on attitudinal and behavioral differences between gangs and other types of peer groups.

Keywords:

• Gang
• delinquent peer group
• fixed-effects modeling
• social cognitive theory

Acknowledgment

The opinions, findings, and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the Department of Justice. This research was made possible by seven cities including the School District of Philadelphia. Opinions contained in this manuscript reflect those of the author and do not necessarily reflect those of the seven cities.

Notes

1. It is important to note that youth who belong to a typical adolescent group may participate in delinquency with some of their friends, but they did not indicate that they participated in illegal activities as a group.

2. Due to an under-representation of African American youth in Chicago schools obtained in the 2006 sampling effort, two additional schools were included in the evaluation, beginning during the 2007–2008 school year (Esbensen et al. 2012).

3. We examined differences between the excluded sample and our analysis sample. Demographically, the excluded sample includes a slightly greater proportion of males as well as youth categorized as black or other race and a smaller percentage of white youth. In terms of group involvement, the excluded sample represents more observations during gang involvement (excluded = 14.2%, final analysis = 9.8%) and less delinquent group membership (excluded sample = 7.3%, final analysis = 9.7%). The excluded sample has lower agreement with hitting neutralizations, less empathy, more self-centeredness, less negative peer commitment, fewer prosocial peers, greater commitment to school, and are less delinquent in terms of both violent and nonviolent offenses. Thus, while these differences indicate that excluded youth are more likely to be gang involved, display reduced tendencies for delinquent attitudes and norms, less delinquency, and less delinquent group involvement.

4. It is important to note that the use of multiple measures of gang involvement is common in prior literature (Alleyne et al. 2016; Lachman, Roman, and Cahill 2013).

5. For this and all other mean scale variables, scales were created if at least half of the included items had nonmissing values.

6. We recognize that this last item, involvement in gang fights, could potentially drive differences when comparing periods of gang membership and delinquent group membership because youth are more likely to be involved in gang fights during periods of active gang membership. We compared the findings reported in this paper to results using only the first four items and the results do not change substantively.

7. Although variety scales are preferable over logged frequency outcomes because they are not driven by less serious high-frequency items and have been found to possess high reliability and validity (Sweeten 2012), negative binomial regression analysis cannot be used for true fixed-effects modeling because it cannot control for time-stable variables (see Allison and Waterman 2002). We examined the robustness of our results using a negative binomial fixed/random-effects hybrid model with a count nonviolent delinquency outcome. These supplemental analyses are discussed in subsequent endnotes.

8. These percentages are based on time points for which we have available data. For example, youth who were in a gang at two waves, but only have available data for two waves, are in a gang for 100% of their time points. While this may overinflate the percentage of time youth appear to be gang involved, the alternative is to drop any youth who have fewer than six waves of data, which limits our sample substantially.

9. Because the delinquency outcomes are log-transformed, the results are interpreted as a percent change in delinquency based on the following equation: (exp(b) − 1) × 100%.

10. Because a logged frequency measure of delinquency is not ideal (see footnote 7), and negative binomial fixed-effects for a count outcome cannot control for time-stable variables, we examined the results using a negative binomial hybrid model. While this model could not be examined for violent delinquency due to convergence issues related to variability in the count outcome, we present the fixed- and random-effect results for the nonviolent delinquency outcome here, interpreted as incident rate ratios. As compared with periods of typical group membership, periods of delinquent group membership are associated with a 1.26 times greater delinquency variety (b = .23, SE = .03, p < .001). Meanwhile, periods of gang membership are associated with a 1.27 times greater variety of nonviolent offenses (b = .24, SE = .03, p < .001). The coefficients that represent the random effects in the hybrid model indicate that the between-person differences are also significant: delinquent group members report nonviolent delinquency variety at a rate 1.56 times greater than youth in a typical peer group (b = .44, SE = .08, p < .001), while gang members report 1.34 times more nonviolent acts (b = .29, SE = .07, p < .001). The complete results are available from the corresponding author.

11. Again, hybrid model results are available only for the nonviolent delinquency variety outcome, but consistent with our OLS fixed effects model, the within-person effect of gang membership, as compared with delinquent group involvement, is not significant (b = −.002, SE = .03, p = .962). Meanwhile, the random-effects or between-person coefficient indicates that youth in a gang offend at a rate of .86 times that of delinquent group members (b = −.16, SE = .08, p = .05).