Need, Connections, or Competence? Criminal Achievement among Adolescent Offenders

Abstract

Abstract Variations in criminal performance have been much less explored than other parameters of criminal careers. We explore the factors associated with differential criminal achievement in a sample of 154 adolescent offenders involved in cannabis cultivation. Drawing from theories of earnings attainment, we examine the role of drug use, criminal social capital and criminal human capital in providing either (a) monetary, or (b) in kind (cannabis) rewards from crime. Results reveal that criminal social capital and criminal human capital are related to performance while drug use explains little of the variation. Their effects, however, differ between outcomes: young offenders who are mainly connected to adult growers tend to be paid in kind, whereas respondents connected to a majority of other young growers tend to receive money. Criminal human capital is crucial to earning money but insignificant to obtaining larger payments in cannabis. Implications for criminal career and desistance research are discussed. Keywords: juvenile delinquencysocial capitalcriminal capitalcriminal achievementcriminal careers Acknowledgments The authors would like to thank Pierre Tremblay, John Laub, Jean McGloin, Eric Beauregard, Owen Gallupe, Jonathan Basamanowicz and the three anonymous reviewers for their helpful comments on earlier versions of this article. Notes 1. Learning theories originate from the assumption that just like any behavior, delinquency is learned. More specifically, criminal behavior is learned from interacting in intimate personal groups (Sutherland, Citation1947) or also secondary groups, like school and church (Akers, 1998). The learning is more salient depending on (1) sense of length of time and (2) absolute and relative amounts of time spend in different associations (in addition to frequency, duration, priority, and intensity). Once the initial criminal act is committed, behavior is reinforced through the principals of operant conditioning. This dictates whether behavior is subsequently maintained or extinguished. According to Akers (1998) reinforcement can be in the form of social reinforcements (i.e. peers) and non-social reinforcements (i.e. monetary rewards from crime). 2. Differential earnings attainment is one aspect of criminal achievement. Other indicators of criminal achievement/success can include prestige (see Matsueda, Gartner, Piliavin, & Polakowski, Citation1992) and cost avoidance (Bouchard & Nguyen, Citation2010; Kazemian & Leblanc, Citation2007). 3. To be sure, young offenders can have both conventional and criminal social capital. We focus exclusively on criminal social capital herein referred to simply as social capital. 4. A third possible type of relationship between adults and youth can also be one of equality. This dynamic however, is not as common as the other two since the power differentials between adults and youth are seldom equal. 5. The situation can be looked at in reverse. The discontinuation of drug use also affects illegal income. Sheerin, Green, Sellman, Adamson, and Deering (Citation2004) assessed participants of Maori decent involved in a methadone maintenance treatment in New Zealand and found that after 18 months of treatment, drug use, drug expenditures, and illegal income were significantly reduced. 6. To be fair, Uggen and Thompson (Citation2003) do control for criminal embeddedness, operationalized as having an unemployed deviant friend, which is a general measure that does not specify the number of deviant friends or the type of deviance. 7. Effective size is the number of nonredundant contacts in a social network. 8. Uggen and Thompson's sample consisted of 2,268 ex-offenders, 1,395 ex-addicts, and 1,241 youth dropouts. They did not find any significant differences between the sub-samples so they pooled the groups together. 9. Indicators are based on 2006 census figures from Institut de la Statistique du Quebec (http://www.stat.gouv.qc.ca). 10. Some examples of invalid questionnaires include forms that were blank, forms that contained extreme missing values, and forms that contained the same selection for every question. 11. The most common reason for excluding a participant was missing data. We systematically excluded respondents who did not answer a majority of the cultivation-related questions. 12. Involvement in cannabis cultivation can take many forms and includes any role that one can occupy in cultivation; from maintenance to harvest (we refer to all participants as "growers" for the remainder of the paper). 13. We do supplementary analysis that include zero earners and results are substantively similar (see Appendix B and C). 14. Freeman (Citation1996) found that among a group of Boston youth who make money from crime, occasional offenders, and weekly offenders earned $250 and $448, respectively. Viscusi (Citation1986) surveyed inner-city youth (15–24 years) from Boston, Chicago and Philadelphia and found that the average monthly illegal income was $272. Among their sample of homeless youth in Toronto and Vancouver, McCarthy and Hagan (Citation2001) found that the average daily earnings of participants in the drug trade was $101. 15. Conversely, using inmate survey conducted by the RAND Corporation, Wilson and Abrahamse (Citation1992) found that offenders' self-reported income did not correlate with the self-reported number of crimes they committed. Wilson and Abrahamse also observed that the offenders also reported the similar incomes across crimes (e.g. robbery and drug dealing each produced a net gain of $480). Yet, Tremblay and Morselli (Citation2000) reanalysed the same data and logged the distribution (for self-reported income and rates of offending) rather than omitting cases that were too high or too low (as Wilson and Abrahamse did) and found a moderate correlation between income and number of crimes committed (r = .47, p < .001). 16. We used regression tree analyses (CHAID) to test whether the cutting point (15) is a good one. First, it can identify one, or multiple cutting points, if they are found to differ significantly from each other in relation to both dependent variables. Second, both the original and the dichotomized variable can be entered simultaneously in the CHAID model, allowing us to test which variable structure the most suitable for the analysis. 17. Respondents were asked if they ever committed the crime in question. To ensure understanding of the crime type, several examples were provided for each of the crime types. For example, "mischief" was clarified by (vandalism, disturbing the peace, break and enter) and prostitution, pimping, and rape were given as examples of "sexual offenses". 18. We also tested for the pattern of missing values. Little's MCAR test revealed that there was no significant deviation from a pattern of values that are "missing completely at random" (p > .05). Because our sample size is relatively small, we estimated the missing values using multiple imputation, creating five imputed datasets and using the pooled estimates, standard errors and p-values as our results (Allison, Citation2002; Rubin, Citation1987; Schafer & Olsen, Citation1998). We included all independent and dependent measures in the imputation procedure, but we included only respondents with valid data on our outcome measures in our analyses. 19. There are few assumptions of ordinal regression but an important one is that there are proportional or parallel odds. That is, the independent variables have the same effect for each of the categories in the dependent variable, which allows one model to be sufficient to explain the relationship between the dependent variable and a set of predictors (O'Connell, Citation2006; Hosmer and Lemeshow, 2000). Each logit has its own threshold values (intercept) but the same co-efficient, which means that each predictor variable is the same for the different logit functions. Rejecting parallelism implies that there may be an interaction with the predictor variables and the splits in the dependent variable. Tests for parallel lines indicate that parallelism was achieved in all our models (p > .05). 20. Although the participants reporting high earnings are our inspiration for inquiry, we test to see if our analyzes were influenced by extreme outliers. We entered all our models into OLS regressions and examine the Mahalanobis distances and compute each of their probabilities. There appeared to be no cases that had an unusual combination of values that had less than p < .01. Therefore, there were no extreme outliers. 21. SPSS 17.0 does not support CHAID analyses with pooled results so we conducted the analyses with the original data and the five imputed datasets separately with comparable results. The results presented are the ones with the original data. 22. These cutting points were chosen by comparing the chi-square value of each of the respective outcome categories to the previous category against the independent variables. The cut-off between the chosen cut-off points and the categories below them have the most statistically significant differences. 23. We removed commercial sites from the logistic regression models because all the high earners and obtainers participated in commercial sites. 24. We did test for interactions between our social and criminal human capital measures in standard regression models but were not able to detect any effects. This is not entirely surprising however; the relatively small sample size and data nonlinearity make detecting interaction effects extremely difficult (see McCelland and Judd (Citation1993) for a discussion on the difficulty in detecting theorized interaction and moderating effects in non-experimental designs). We therefore use CHAID analyses, a nonparametric method which performs well with non-linear, highly skewed variables, or those with an ordinal structure (Lewis, Citation2000).