Youth gangs, delinquency and drug use: a test of the selection, facilitation, and enhancement hypotheses Uberto Gatti, Richard E. Tremblay,2 Frank Vitaro, and Pierre McDuff University of Genoa, Italy; University of Montreal, Canada Background: Three different explanations have been given for the observation that adolescent gang members report more delinquent behaviour than their counterparts who do not affiliate with gangs: a) adolescents who commit more crimes join gangs (selection hypothesis); b) gang membership facilitates deviant behaviour (facilitation hypothesis); c) selection and facilitation work interactively (enhancement hypothesis). The aim of this study was to test these hypotheses, while controlling for self-reported delinquency, friends' delinquency, and individual as well as family characteristics. Method: The sample included 756 boys first assessed when they attended kindergarten in disadvantaged areas of Montreal. Gang membership was assessed at the ages of 14, 15 and 16 years. Delinquency and drug use data were collected from self-reports and court files at the same ages. Results: Gang members displayed far higher rates of delinquent behaviour and drug use than non-gang members. The results support the facilitation model for transient gang members (i.e., youths in a gang during only one of the three periods considered) and the enhancement model for stable gang members (i.e., youths in a gang for at least two of the periods considered), for person and property offences. The association between gang membership and delinquency persisted after introducing the control variables. Additional analy- ses showed that the effect associated with belonging to a gang was beyond that of simply having delinquent friends. Conclusion: Preventing the creation and participation in such gangs should reduce the frequency of antisocial behaviour during adolescence. Keywords: Adolescence, antisocial behaviour, delinquency, peer relationships, substance use, gangs. A large number of studies conducted in the United States have revealed that youths who belong to a gang commit more crimes than those who do not, and that this difference is greater with regard to serious crimes and crimes of violence (Battin-Pear- son, Thornberry, Hawkins, & Krohn, 1998; Fagan, 1989; Huff, 1998; Klein, 1995; Spergel & Curry, 1993; Taylor, 1990; Thornberry & Burch II, 1997; Vigil, 1988). Outside of the United States, less importance has been attached to the problem of youth gangs. Some earlier European studies found that belonging to a gang did not elicit higher rates of delinquency (Morash, 1983), while others concluded that the stereotype of the violent gang could aggrav- ate or induce deviant behaviour in groups that were not originally involved in delinquency (Cohen, 1980). More recently, however, youth gangs have come under greater scrutiny outside the USA, and researchers are now wondering whether the phe- nomenon has been hitherto underestimated (Klein, 1997; Klein, Kerner, Maxson, & Weitekamp, 2001). Longitudinal studies are especially suited to test- ing three models that, following Thornberry, Krohn, Lizotte, and Chard-Wierschem (1993), can be con- sidered: the selection model, the facilitation model, and the enhancement model. According to the selection model, the high level of delinquency among gang members does not depend on the influence of associating to a gang, but rather is due to the fact that gangs attract youths who are already Association for Child Psychology and Psychiatry, 2005. committing crimes. In contrast, the facilitation model holds that it is the process of associating to a gang which facilitates the use of deviant behaviour in youths who, before joining the gang, are no different from other young people. The enhancement model combines these two views: youths who join gangs already display a higher level of delinquency, but joining the gang exacerbates their deviant behaviour. Using the data from the Rochester Youth Devel- opment Study, a panel study of 1,000 males and females that over-represented adolescents at high risk for serious delinquent behaviour and drug use, Thornberry et al. (1993) studied whether delin- quency and drug use were more frequent among youths who belonged to gangs than among those who did not and, if so, what causal mechanism underlay this difference. They investigated the evo- lution of each phenomenon over time, by analysing data from three successive periods; the results of this study, which included only 15% of white par- ticipants, showed that, before joining a gang, mem- bers did not display higher rates of delinquency or drug use than non-members, that such behaviours increased upon entry into the gang, and that delin- quency declined once the individual left the gang. However, Thornberry later (1998) found that, in addition to a strong facilitation effect, a weaker selection effect was also present. Strong evidence for the facilitation model and partial support for selection were also provided by Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA Table 8 Effects of gang membership, deviant friends and previous risk factors on general delinquency (N = 742) Standardised regression coefficients (B) Total delinquency at 15 years Gang membership (current year) Delinquent friends (current year) Prior delinquency (11-12-13 year) Non-intact family structure (11-12-13 year) Occupational prestige: 11-12-13 years Disruptiveness (11-12-13 years) School difficulties(11-12-13 years) Parental supervision (11-12-13 years) Total delinquency at 14 years .21*** .21*** .45*** .05* .04 -.01 -.03 -.04 R adj. .45 F= 76.80*** .35*** .20*** .33*** .03 .03 .02 -.01 -.03 R adj. .45 F= 75.88*** Total delinquency at 16 years .32*** .24*** .25*** .03 .01 .07** .02 -.08** R adj. .42 F= 68.08*** *p < .10; **p < .05; ***p < .001. and additive effect on delinquency, and that the gang therefore exerts a specific effect that goes beyond the condition of having delinquent friends. Discussion We found that a high percentage (37%) of male adolescents in this French-Canadian sample repor- ted having been in a gang at some time or other when they were between 14 and 16 years of age. This rate is probably an underestimate, since boys with behaviour problems in kindergarten and boys from families with lower socioeconomic levels were over- represented in those who were missing when the gang membership data was collected. Although the sample was chosen from schools in low socioeco- nomic areas to increase delinquency rate, it is somewhat surprising that the rate of gang member- ship for this study is similar to the one observed in Rochester, NY (i.e., Thornberry, 1998). First, violent crimes are much less frequent in Canada than in the United States (e.g., Hagan, 1991; Lipset, 1990). Second, the Montreal sample included only Cana- dian-born, French-speakers, who form the majority of the Montreal population; minority groups, who are often considered more likely to belong to gangs (Covey, Menard, & Franzese, 1997; Curry, Ball, & Decker, 1996; Spergel, 1990), were therefore exclu- ded. The results also revealed that many youths belong to gangs for relatively short periods of time (Lacourse et al., 2003; Thornberry & Burch, 1997) and there- fore the importance of understanding the facilitative impact of affiliating with a gang, but also the life course of these youths once they have left the gang. Results appeared to differ, depending on the years considered. Youths who belonged to a gang dis- played rates of delinquent behaviour that were two or three times higher than those who did not belong to a gang; drug use was three or four times higher, police arrests were four times higher, and there were four times more appearances in the Juvenile Courts at 15 years of age, and seven times more at 16 years of age. Despite important differences in the cultural characteristics of the participants, our findings are similar to the results obtained in the Rochester Youth Development Study (Thornberry et al., 1993), the Seattle Social Development Project (Hill, Lui, & Hawkins, 2001), and the Denver Youth Study (Es- bensen & Huizinga, 1993). One notable difference between these studies and ours, however, lies in the types of offences committed by youths in general and by gang members in particular; while property of- fences were much more prevalent among the par- ticipants in the Montreal study, a higher frequency of violent crimes was recorded in the Rochester and Seattle studies. One of the differences, therefore, between the USA and French-Canadian youth gangs seems to be that the former show a greater propen- sity for violent behaviour, while the latter are more inclined towards other types of deviant behaviour, and towards theft in particular. Another important contribution of the present study was to show that the higher rates of deviant behaviour among gang members were observed through both self-reported data and official data. Concerning the time-course relationships between belonging to a gang and the frequency of deviant behaviours, our results validate the facilitation model for transient gang members, and the enhancement model for stable gang members, with regard to offences against persons and property of- fences. It may therefore be claimed that transient gang members do not display greater previous involvement in deviant behaviours than youths who have never belonged to a gang, while gang member- ship for at least two consecutive years is associated with a higher level of previous delinquency. Thus, while gang membership is associated with increased delinquency in every group, the youths who display higher levels of previous delinquency are more likely to remain in the gang for at least two consecutive years. The patterns for drug sales and drug use are less evident, especially because one-year gang members did not systematically differ from non-gang members and because ex-gang members remained high on these behaviours after leaving the gang. These results support Loeber's conclusion that the pathway for substance use is different from the pathways for other disruptive-delinquent behaviours (Loeber, 1988). They also suggest that the substance pathway is less affected by proximal social agencies such as gangs and more under developmental or societal norms as revealed by the general increase of drug use with age and the increase in prevalence rates during the mid-nineties in North America when the data in the present study were collected (John- ston, O'Malley, & Bachman, 2003; Adlaf & Paglia, 2001). While both behaviours were more frequent among stable gang members, we were unable to ascertain clearly which explanatory model was involved. In conclusion, for non-immigrant Caucasian ado- lescents in the largest French-speaking city in North America, youth gangs appear to play a role which is both important and similar to that observed in the USA, often among African-Americans. Moreover, the study reveals that the higher delinquency rates among gang members are largely linked to the experience of the gang itself, rather than to the social deficiencies that characterise its members, and that the apparent effect exerted by the gang is specific and goes beyond simply having delinquent friends. Stable gang members (those who remain in the gang for at least two of the three periods considered) dis- play higher rates of property offences, violent beha- viour, drug abuse and drug selling in each of the years considered when compared not only with never gang members but often also with transient gang members (those belonging to a gang in only one of the periods). This appears to demonstrate that not only entry into the gang, but also prolonged mem- bership is associated with a greater risk of deviant behaviour, and suggests the importance not only of preventing gang formation, and youths from joining gangs, but also of acting to reduce the duration of their membership. Finally, it must be noted that these conclusions are based on correlational data. We need experi- mental data to test causal mechanisms (Lacourse et al., 2002; Tremblay, 2003; Vitaro, Brendgen, & Tremblay, 2001). The next generation of studies on the effects of gangs should implement randomised experiments that attempt to prevent at-risk youths from entering gangs. Long-term follow-up of youth in such experiments will give a better indication of the effect of gang membership than longitudinal studies without randomised experiments. Acknowledgements This study was made possible by grants from Qubec's CQRS and FCAR funding agencies, Can- ada's Research Chair Program, NHRDP and SSHRC funding agencies, the Canadian Institute for Advanced Research, and the Molson foundation. We thank Hlne Beauchesne and Lucille David for supervising the data collection, Lyse Desmarais- Gervais and Muriel Rorive for managing the data bank, and Ted Barker and Katia Maliantovitch for revising the text. Correspondence to Richard E. Tremblay, GRIP, University of Montreal, 3050 Edouard-Montpetit, Suite A-210, Montreal, Quebec, Canada, H3T 1J7; Tel: (514) 343-6963; Fax: (514) 343-6962; Email: grip@umontreal.ca References Adlaf, E., & Paglia, A. (2001). Drug use among Ontario students 1977-2001. Findings from the OSDUS. Tor- onto, Ont.: Report of the Centre for Addiction and Mental Health. Battin, S.R., Hill, K.G., Abbott, R.D., Catalano, R.F., & Hawkins, J.D. (1998). The contribution of gang membership to delinquency beyond delinquent friends. Criminology, 36, 93-115. Battin-Pearson, S.R., Thornberry, T.P., Hawkins, J.D., & Krohn, M.D. (1998). Gang membership, delinquent peers, and delinquent behavior. Washington, DC: US Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention. Blishen, B.R., Carroll, W.K., & Moore, C. (1987). The 1981 socioeconomic index for occupations in Can- ada. Canadian Review of Sociology and Anthropology, 24, 465-488. Cohen, S. (1980). Folk devils and moral panics: The creation of the Mods and Rockers. Oxford: Martin Robertson. Covey, H.C., Menard, S., & Franzese, R.J. (1997). Juvenile gangs (2nd edn). Springfield, IL: Charles C. Thomas Minoranze. Curry, G., Ball, R.A., & Decker, S.H. (1996). Estimating the national scope of gang crime from law enforce- ment data. In C. R. Huff (Ed.), Gangs in America (2nd edn, pp. 21-36). Thousand Oaks, CA: Sage. Esbensen, F., & Huizinga, D. (1993). Gangs, drugs, and delinquency in a survey of urban youth. Criminology, 31, 565-589. Esbensen, F., Peterson, D., Freng, A., & Taylor, T.J. (2002). Initiation of drug use, drug sales, and violent offending among a sample of gang and non-gang youth. In C.R. Huff (Ed.), Gangs in America (3rd edn, pp. 37-50). Beverly Hills, CA: Sage. Fagan, J.A. (1989). The social organization of drug dealing among urban gangs. Criminology, 27, 633- 669. Hagan, J. (1991). The disreputable pleasures: Crime and deviance in Canada. Toronto, Canada: McGraw-Hill. Hill, K.G., Howell, J.C., Hawkins, J.D., & Battin- Pearson, S.R. (1999). Childhood risk factors for adolescent gang membership: Results from the Se- attle Social Development Project. Journal of Research in Crime and Delinquency, 36, 300-322. Hill, G.H., Lui, C., & Hawkins, J.D. (2001). Early precursors of gang membership: A study of Seattle youth. Juvenile Justice Bulletin, Washington, DC: US Department of Justice, Office of Justice Program, OJJDP. Howell, J.C., Egley, A., & Gleason, D.K. (2000, Novem- ber). Youth gangs: Definitions and the age-old issue. Paper presented at the meeting of the American Society of Criminology, San Francisco, CA. Huff, C.R. (1998). Criminal behavior of gang members and at-risk youths. Washington, DC: US Department of Justice, Office of Justice Programs, National Insti- tute of Justice. Johnston, L.D., O'Malley, P.M., & Bachman, J.G. (2003). Monitoring the future: National survey results on drug use, 1975-2002. Volume I: Secondary school students (NIH Publication No. 03-5375). Bethesda, MD: National Institute on Drug Abuse, 520 pp. Klein, M.W. (1995). The American street gang: Its nature, prevalence, and control. New York: Oxford University Press. Klein, M.W. (1997). Gangs in the United States and Europe. European Journal on Crime Policy and Research, 4, 63-80. Klein, M.W., Kerner, H.-J., Maxson, C.L., & Weitekamp, E.G.M. (2001). The Eurogang paradox: Street gangs and youth groups in the U.S. and Europe. Dordrecht: Kluwer Academic Publisher. Lacourse, E., Ct, S., Nagin, D.S., Vitaro, F., Brendgen, M., & Tremblay, R.E. (2002). A lon- gitudinal-experimental approach to testing theories of antisocial behavior development. Development and Psychopathology, 14, 909-924. Lacourse, E., Nagin, D.S., Tremblay, R.E., Vitaro, F., & Claes, M. (2003). Developmental trajectories of boys' delinquent group membership and facilitation of violent behaviors during adolescence. Development and Psychopathology, 15, 183-197. Lipset, S.M. (1990). Continental divide: The values and institutions in the United States and Canada. New York: Routledge. Lipsey, M.W., & Derzon, J.H. (1999). Predictors of violent or serious delinquency in adolescence and early adulthood: A subthesis of longitudinal research. In R. Loeber & D.P. Farrington (Eds.), Serious and violent juvenile offenders: Risk factors and successful interventions (pp. 86-105). Thousand Oaks, CA: Sage Publications. Loeber, R. (1988). The natural histories of juvenile conduct problems, substance use and delinquency: Evidence for developmental progressions. In B.B. Lahey & A.E. Kazdin (Eds.), Advances in clinical child psychology (vol. 11, pp. 73-124). New York: Plenum Press. Morash, M. (1983). Gangs, groups, and delinquency. British Journal of Criminology, 23, 309-335. Spergel, I.A. (1990). Youth gangs: Continuity and change. In N. Morris & M. Tonry (Eds.), Crime and justice: An annual review of research (pp. 171-275). Chicago: University of Chicago Press. Spergel, I.A., & Curry, G.D. (1993). The National Youth Gang Survey: A research and development process. In A. Goldstein & C.R. Huff (Eds.), Gang intervention handbook (pp. 359-400). Champaign-Urbana, IL: Research Press. Taylor, C.S. (1990). Gang imperialism. In C.R. Huff (Ed.), Gangs in America (3rd edn, pp. 103-115). Beverly Hills, CA: Sage. Thornberry, T.P. (1998). Membership in youth gangs and involvement in serious and violent offending. In R. Loeber & D.P. Farrington (Eds.), Serious and violent offenders: Risk factors and successful inter- ventions (pp. 147-166). Thousand Oaks, CA: Sage Publications. Thornberry, T.P., & Burch, J.H., II. (1997). Gang members and delinquent behavior. Washington, DC: US Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Preven- tion. Thornberry, T.P, Krohn, M.D., Lizotte, A.J., & Chard- Wierschem, D. (1993). The role of juvenile gangs in facilitating delinquent behavior. Journal of Research in Crime and Delinquency, 30, 55-87. Thornberry, T.P, Krohn, M.D., Lizotte, A.J., Smith C.A., & Tobin, K. (2002). The toll of gang membership: Gangs and delinquency in a developmental perspec tive. New York: Cambridge University Press. Tremblay, R.E. (2003). Why socialization fails: The case of chronic physical aggression. In B.B. Lahey, T.E. Moffitt, & A. Caspi (Eds.), Causes of conduct disorder and juvenile delinquency (pp. 182-224). New York: Guilford Publications. Tremblay, R.E., Loeber, R., Gagnon, C., Charlebois, P., Larive, S., & LeBlanc, M. (1991). Disruptive boys with stable and unstable high fighting behavior patterns during junior elementary school. Journal of Abnormal Child Psychology, 19, 285-300. Vigil, J.D. (1988). Barrio gangs: Street life and identity in Southern California. Austin, TX: University of Texas. Vitaro, F., Brendgen, M., & Tremblay, R.E. (2001). Preventive intervention: Assessing its effects on the trajectories of delinquency and testing for media- tional processes. Applied Developmental Science, 5, 201-213. Zhang, L., Welte, J.W., & Wieczorek, W.F. (1999). Youth gangs, drug use, and delinquency. Journal of Criminal Justice, 27, 101-109. Manuscript accepted 27 September 2004 Appendix 1 List of items in self-reported delinquency and drug use indexes: A) General delinquency (16 questions) 1) carried a weapon (such as a chain, knife or gun) 2) used a weapon during a fight with someone 3) threatened someone to force him to do something 4) beat up someone who hadn't done anything to you 5) thrown things like a rock or a bottle at people 6) broken or destroyed things that did not belong to you, on purpose 7) set fire to a store or other place, on purpose 8) entered a place without paying the entrance fee 9) broken into some place to steal something Zhang, Welte, and Wieczorek (1999), through their analysis of the data from the Buffalo Longitudinal Survey of Young Men, a five-year panel study on a sample of 625 males aged 16 to 19. Changes in the behaviour of gang members over time were also investigated in the Denver Youth Survey, a longit- udinal study of families in 'high-risk' neighbour- hoods. The results revealed that youths who belonged to gangs displayed higher rates of street crime and other serious offences even before they joined the gang, but that their prevalence and indi- vidual offending rates were higher during the actual years of membership (Esbensen & Huizinga, 1993). Results from another longitudinal study, the Seattle Social Development Project, also revealed that, in each age group, respondents who were gang mem- bers always had the highest frequency of delinquent behaviour and drug use during the period they were part of the gang (Hill, Howell, Hawkins, & Battin- Pearson, 1999). These results suggest that the criminogenic effect is not due simply to contact with other delinquent youths, but that it is also linked to the influence of the gang itself. That serious forms of drug use, drug dealing and violence are also more frequent among gang mem- bers was confirmed by a longitudinal study con- ducted on students from over 100 schools (middle and high schools) in six American cities, who were questioned for four consecutive years about their belonging to a gang and their involvement in deviant behaviour (Esbensen, Peterson, Freng, & Taylor, 2002). In that study, the enhancement model seemed to fit better with the results observed. On comparing gang members, non-gang youths with delinquent friends and non-gang youths who did not have delinquent friends, Battin et al. (1998) found that gang membership predicted self-reported and officially-recorded delinquency beyond the effects of having delinquent friends and prior delinquency. In general the longitudinal studies reveal that gang members display higher rates of juvenile delin- quency and drug use than non-gang members do, and that joining a gang increases the frequency of these behaviours. However, some discrepancy can be seen among data on delinquency in the period prior to gang membership; in some cases delin- quency rates have been reported to be similar to those seen among youths who do not belong to a gang, while in other cases higher rates have been reported. In the first (1993) study by Thornberry et al., for example, stable gang members displayed higher rates of violent crime than non-gang mem- bers, even during periods in which they did not belong to a gang, while no difference in the rates of such crimes was observed between transient gang members and non-gang members. Our aims were: 1) to establish whether gang members display a higher level of delinquency and drug use than youths who do not belong to a gang, using both self-reported and official data; 2) to test whether any greater involvement of gang members in delinquency could be explained by the selection, facilitation or enhancement models; 3) to ascertain whether the higher level of delin- quency was related with belonging to a gang, or whether it was due to the fact that these youths had personal, family and social problems and deficiencies before joining the gang. Several lon- gitudinal studies have shown that low S.E.S., family break-up, scant supervision, failure at school and early delinquent behaviour are risk factors for both delinquency and gang member- ship in adolescence (Lipsey & Derzon, 1999; Thornberry, Krohn, Lizotte, Smith, & Tobin, 2002). We therefore aimed to establish whether the association between delinquency and gang membership was spurious, or whether it persisted after the control variables had been introduced; 4) to test whether a specific effect was exerted by gang membership, regardless of whether the youths had delinquent friends; 5) finally, since all of the longitudinal studies on gang effects appear to have been conducted in the USA, often with African-American children, our further aim was to establish the extent to which these gang membership findings can be general- ised to another culture, in this case white French- speaking Canadian males. Methods Participants The sample initially consisted of 1,161 boys who had attended kindergarten classes in 1984, in disadvantaged areas of Montreal. To obtain a high base rate of boys at risk for delinquent behaviour, the 53 schools with the lowest socioeconomic indexes were chosen. To control for cultural effects, the boys were included in the study only if both their biological parents were born in Canada and if their mother tongue was French. After eliminating those who did not have Canadian-born French-speaking par- ents, those who refused to participate and those who could not be traced, the sample was reduced to 1,037 Caucasian males. After the first evaluation in kinder- garten, annual evaluations were made, starting at age 10. These evaluations were made by the parents, teach- ers, classmates, and the children themselves. Self- reported delinquency was also assessed for the first time at age 10. The following analyses included subjects with data available from 10 to 16 years of age (n = 756). Par- ticipants who were not included in the analyses because of missing data (n = 281) tended, at age 6, to be more disruptive than the rest of the sample; their parents tended to have a lower educational level, and their fathers tended to have lower occupational prestige; no differ- ences were found in the age of the parents when the first child was born, the occupational prestige of the mother, or the family status. Instruments and procedure Gang membership. At 14, 15, and 16 years of age the boys were asked: During the past 12 months, were you part of a group or gang that did reprehens- ible acts?' If the answer was affirmative in at least one of the three assessments, the youth was regarded as being a gang member; if the answer was always negative, he was considered to be a non-gang mem- ber. Several studies have used this type of question to assess gang membership (Howell, Egley, & Gleason, 2000; Lacourse, Nagin, Tremblay, Vitaro, & Claes, 2003; Thornberry et al., 1993). In accordance with the classification drawn up by Thornberry et al. (1993), the gang members were subdivided into two categor- ies: 1) transient gang members, if they had been in a gang in only one of the three periods considered; 2) stable gang members, if they had belonged to a gang in at least two consecutive years of the periods con- sidered. On the basis of the period in which they belonged to a gang, and in order to evaluate the level of delinquency before, during and after belonging to a gang, gang members were further subdivided into the following categories: 1.1) age 14 only gang members (n = 62); 1.2) age 15 only gang members (n=63); 1.3) age 16 only gang members (n = 45); 2.1) age 14-15 only gang members (n=34); 2.2) age 15-16 only gang members (n=29); 2.3) age 14-15-16 gang members (n = 43). Age 14 only gang members were those who had be- longed to a gang only in the 12 months prior to the interview conducted at the age of 14, in 1992; age 15 only had been members only during the 12 months prior to the interview conducted at the age of 15, in 1993; and age 16 only during the 12 months prior to the 1994 interview, at the age of 16. The other cat- egories of gang members were defined on the basis of more than one period of gang membership over the three years considered. Out of a total of 756 youths interviewed, 480 had never belonged to a gang, while 276 (37% of the sample) had belonged to a gang for at least a one-year period, and only 16% of the latter had been part of a gang during the three-year period. A high percentage of gang members had therefore left the gang during the study period. Delinquency and drug use. Delinquency and drug use data were collected through questionnaires administered when the youths were aged 14, 15 and 16 years. The boys were asked whether, in the 12 months prior to the interview, they had ever in- dulged in any of the behaviours listed in the question- naire. The deviant behaviours made up five indexes: Person offences (5 items), Property offences (10 items), Drug sale (1 item), Drug use (8 items), and General delinquency, which comprised 16 items drawn from the first 4 indexes; for a description of the various indexes and items, see Appendix 1. Cronbach's alphas for the general delinquency scale ranged from .87 to .89. The range for the person offences scale was from .66 to .71, for the property offences from .83 to .86, and for drug use from .63 to .70. The items making up the first three indexes were scored as: 0) never, 1) once or twice, 2) sometimes, and 3) often. For each index, the sum of the scores on the various items was calculated. With regard to the index of drug sales, the number of illegal acts reported was considered. For the index of drug use, the variable concerning the use of marijuana was dichotomised (use vs. non-use) - because the original scores were highly skewed - and summed to the variables concerning the use of other substances, all of which were dichotomous. In addition to self-reported delinquency, we also asked the participants whether they had been arrested by the police in the 12 months prior to each assess- ment. We chose to record self-reported arrests rather than arrests registered by the police since, in many cases involving minors, the arrest procedure is informal and leaves no trace in the official statistics. Further- more, we also recorded whether the youths appeared before the juvenile court (scored 1) or not (scored 0); unlike the previous data, which were collected by means of self-reporting, these were official statistics taken from judicial archives. Control variables Sociodemographic information. Mothers reported on family structure (intact, non-intact) and occupation of both parents (or the parent with whom the child was living) when the boys were 11, 12 and 13 years of age. Parental occupation was scored on a continuous scale using the Blishen, Carroll, and Moore (1987) Occupa- tional Prestige scale. This score is based on the average income and average education level associated with occupations in Canada. For boys living with two work- ing parents, the highest parental occupation score was used. The minimum score on the scale was assigned to boys living in families on social welfare or on unem- ployment insurance. Disruptive behaviour. Teachers rated the boys' behaviour at ages 11, 12 and 13, using the Social Behavior Questionnaire (SBQ; Tremblay et al., 1991). Ratings took place near the end of the school year, in April or May. The SBQ is a 32-item behaviour-rating questionnaire used to assess disruptiveness (i.e., aggressiveness-opposition-hyperactivity, 13 items; inattention, 4 items; anxiety-withdrawal, 5 items; and prosocial behaviour, 10 items). Teachers indicated whether items did not apply (0), applied sometimes (1), or applied often (2). For the purpose of the present study, only the disruptiveness scale was used. Internal consistency was high, with alphas > .85. General delinquency. The same general delinquency items previously described were administered when participants were 11, 12 and 13 years of age. Parental supervision. At ages 11, 12 and 13, boys reported on parents' monitoring (2 items). Each item was rated 0, 1, 2 or 3, with higher scores indicating more monitoring. The two monitoring items were, 'Do your parents know where you are when you go out?' and 'Do your parents know who you hang out with?' (alphas for the total monitoring scale were .71, .73, and .74 for ages 11, 12 and 13 respectively). Friends' deviancy. At ages 14, 15 and 16, particip- ants reported on their exposure to deviant friends over the previous 12 months by answering the following question: 'How many of your best friends have been arrested by the police for an illegal act they have com- mitted?' Participants answered: none (1), a few (2), many (3), or all of them (4). However, because answers were highly skewed and rare, deviancy of a friend was computed as a dichotomous score of 0 or 1 (combining scores of 2, 3, and 4). School difficulties. The boys were regarded as having school difficulties if they were not in an age-appropriate regular classroom (AARC) when they were 11, 12 and 13 years old. Not being in an AARC environment meant that the boys had been held back for at least one year or placed in a special environment, special class, special school or residential institution. Results Frequency of deviant behaviour in gang and non-gang members Table 1 illustrates the mean frequency of self- reported delinquency and drug use among youths Table 1 Mean frequency of self-reported delinquency for gang members and non-gang members (N = 756) Person offences Gang members 14 years 15 years 16 years For current year 2.4 (2.69) 2.93 (2.72) 2.70 (2.54) Non-gang members For current year .77 (1.30) .70 (1.25) Property offences Gang members For current year 4.11 (4.57) Non-gang members For current year 1.19 (2.11) Drug sales .61 (1.17) 5.32 (5.13) 5.42 (5.06) 1.22 (2.15) 1.36 (2.47) Gang members For current year .17 (.58) Non-gang members For current year .04 (.27) .53 (.82) .06 (.31) .69 (1.01) .17 (.53) Drug use Gang members For current year Non-gang members .46 (.96) .13 (52) For current year "Mean "Standard deviation 1.14 (1.45) 1.41 (1.59) .27 (.71) .53 (.86) who belonged to a gang (gang members) with that of youths who did not belong to a gang (non-gang members), for each of the three years considered. A series of multivariate analyses of variance revealed that the mean level of every kind of delinquent behaviour was higher during the years of gang membership than during the years of non-member- ship, multivariate F (24, 2916) = 8.94, 13.84, and 13.00 for ages 14, 15, and 16 years, respectively. Furthermore, univariate Fs (6, 746) varied from a minimum of 8.31 to a maximum of 50.61 (all ps Table 3 Repeated measures analysis of variance for each group separately and t-tests for related samples with Bonferroni correction (p < .002) Never gang members Multivariate F Univariate F Person offences Property offences Drug sales Drug use Age 14 only gang members Multivariate F Univariate F Person offences Property offences Drug sales Drug use Age 15 only gang members Multivariate F Univariate F Person offences Property offences Drug sales Drug use Age 16 only gang members Multivariate F Univariate F Person offences Property offences Drug sales Drug use Age 14-15 only gang members Multivariate F Univariate F Person offences Property offences Drug sales Drug use Age 15-16 only gang members Multivariate F Univariate F Person offences Property offences Drug sales Drug use Age 14-15-16 gang members Multivariate F Univariate F Person offences Property offences Drug sales F(8,1812) 11.07*** F(2,908)=1.57 =.75 =14.34*** =.39.70*** F(8,232) 5.00*** F(2,118) = 7.65** =1.22 =6.85*** =14.77*** F(8,236) 7.47*** F(2,120) 15.83*** = =8.60*** =7.32*** =12.29*** F(8,156) 6.14*** F (2,80) = 3.10" =16.41*** =4.93** =14.54*** F(8,116) 3.98** = F(2,60) = 5.92** =6.50** =3.41 =5.17** F(8,108) 4.20*** F(2,56) 6.42** = =9.10*** =6.56** =11.29*** F(8,152) 6.00*** F(2,78) = 1.21 =3.09 =11.85*** =23.76*** t-test 14 ys 16 ys; 15 ys < 16 ys 14 ys 16 ys 14 ys < 16 ys 14 ys 15 ys; 14 ys < 16 ys t-test 14 ys < 15 ys > 16 ys 14 ys < 15 ys > 16 ys 14 ys < 15 ys 14 ys < 15 ys; 14 ys < 16 ys t-test 14 ys < 16 ys; 15 ys < 16 ys 16 ys 15 ys; 16 ys > 14 ys t-test 15 ys > 16 ys 15 ys > 16 ys 14 ys < 15 ys t-test 14 ys < 15 ys 14 ys < 15 ys; 14 ys < 16 ys 14 ys < 16 ys t-test 16 ys > 15 ys; 14 ys < 16 ys Drug use ***p < .001; **p < .01; p < .05. F (48, 5680) = 6.79, p < .001, which suggested that: a) the differences between the groups were not con- stant across ages, or b) the increase in delinquent behaviours from age 14 through age 16 was different across the groups. More importantly, this significant interaction applied to all four dependent measures, i.e., person offences, property offences, use of drug and sales of drugs, F (12, 1420) = 9.02, 11.55, 11.22, and 9.76, respectively, p < .001. In order to break down this interaction, a series of one-factor (i.e., age) repeated measures analysis of variance within each group were performed, followed by t-tests for related samples when necessary; following Bonferroni's procedure, the p value for the t-tests was set to p < .002 to adjust for the number of comparisons (7 groups x 3 comparisons/group) for each of the four delinquency scales (see Table 3). Likewise, a series of three one-factor (i.e., group) analyses of variance was performed at each age separately, followed by Seheff's tests for multiple group comparisons (Tables 4, 5, 6 and 7). Group comparisons at each of the three age periods were used to pick out the difference in the frequency of delinquency between gang members and non-gang members, both when the gang members were in the gang and when they were not (i.e., before joining or after leaving the gang). These results were examined with reference to each deviant behaviour separately. Table 4 Multivariate and univariate analysis of variance comparing the seven gang groups at each age separately and post hoc comparison using Scheff's test procedure when applicable (N = 755) Person offences Age 14 Mean SD A B Never gang members (A) .63 1.15 Age 14 only gang m. (B) 1.72 2.36 Age 15 only gang m. (C) 1.03 1.31 Age 16 only gang m. (D) 1.09 1.70 Age 14-15 only gang m. (E) 2.50 2.86 Age 14-15-16 gang m. (G) Age 15 Age 15-16 only gang m. (F) 1.90 2.11 3.56 2.69 = Univariate F (6,755) 31.13, p < .001 D E Univariate F (6,755) = 49.26, p < .001 Mean SD A B C Never gang members (A) .61 1.21 Age 14 only gang m. (B) 1.02 1.38 Age 15 only gang m. (C) 1.95 2.06 Age 16 only gang m. (D) 1.13 1.32 Age 14-15 only gang m. (E) 2.91 3.03 Age 15-16 only gang m. (F) 3.48 2.52 Age 14-15-16 gang m. (G) 4.02 3.00 Age 16 Mean SD A B Never gang members (A) .53 1.15 Age 14 only gang m. (B) .79 1.07 Age 15 only gang m. (C) .76 1.15 Age 16 only gang m. (D) 1.69 1.33 Age 14-15 only gang m. (E) 1.06 1.54 Age 15-16 only gang m. (F) 3.03 2.57 Age 14-15-16 gang m. (G) 3.53 3.12 D E Univariate F (6,754) = 42.97, p < .001 D E F G F G F G Person offences There were no significant variations over time in the frequency of person offences committed by youths who had never belonged to a gang from 14 through 16 years of age (never gang members). The mean frequency of this kind of delinquency among tran- sient gang members (who had belonged to a gang for only one of the three age periods considered) did, however, vary for two of the three cases (age 14 only gang members and age 15 only gang members) (see Table 3). The delinquency for the age 14 only gang members diminished but only by age 16. The age 15 only members showed an increase in delinquent behaviour after joining the gang and a decrease after leaving. Among the age 16 only gang members, the level of delinquency seemed stable from age 14 to age 15 and from age 15 to age 16, despite a global increase during this period. The data on stable gang members (youths who had belonged to a gang for at least two consecutive periods), which are reported in the lower panel of Table 3, reveal a significant decline in the level of person offences on leaving the gang (in the third period). This finding refers to youths classified as age 14-15 only gang members. Similarly, among youths who did not belong to a gang during the first period, the delinquency rate increased on joining the gang and subsequently remained higher (age 15-16 only gang members). Among the 14-15-16 gang mem- bers, however, no significant variations are seen from one year to the next. The delinquency scores for groups involving transient gang members are presented in Table 4. The differences between the youths who had never belonged to a gang from 14 through 16 years of age (never gang members) and the transient gang members are significant only in the year of gang membership for each of the three types of transient gang member. This means that before the tran- sient gang members joined a gang, or after they left the gang, their level of person offences was no higher than that of youths who had never belonged to a gang. The differences between stable gang members and the youths who had never belonged to a gang from 14 through 16 years of age (never gang members) are also presented in the table. Differences are noted for each period (except for one out of nine cases). If we consider the age 15-16 only gang members, we can observe that these boys display a higher level of delinquency than never gang members even in a period (age 14) in which they do not belong to a gang, thus lending support to the selection model. During the period of gang membership, stable gang mem- bers have a higher level of person offences than Table 5 Multivariate and univariate analysis of variance comparing the seven gang groups at each age separately and post hoc comparison using Scheff's test procedure when applicable (N = 755) Property offences Age 14 Mean SD A B Never members (A) .95 1.78 Age 14 only gang m. (B) 2.85 3.40 Age 15 only gang m. (C) 1.73 2.36 Age 16 only gang m. (D) 1.60 2.34 Age 14-15 only gang m. (E) 4.35 5.32 Age 15-16 only gang m. (F) 3.27 4.07 Age 14-15-16 gang m. (G) 5.72 4.96 Age 15 Mean SD A B Never gang members (A) 1.05 2.02 Age 14 only gang m. (B) 2.18 2.74 Age 15 only gang m. (C) 3.29 3.45 Age 16 only gang m. (D) 1.73 2.13 Age 14-15 only gang m. (E) 6.47 5.46 Age 15-16 only gang m. (F) 5.45 4.44 Age 14-15-16 gang m. (G) 7.30 6.31 Age 16 Univariate F (6,755) = 32.33, p < .001 C D E Univariate F (6,755) = 52.55, p < .001 C D E F G F G Univariate F (6,754) = 40.97, p < .001 Mean SD A B D E F Never gang members (A) 1.06 2.07 Age 14 only gang m. (B) 2.31 3.63 Age 15 only gang m. (C) 1.83 2.63 Age 16 only gang m. (D) 3.58 3.56 Age 14-15 only gang m. (E) 3.12 3.53 Age 15-16 only gang m. (F) 6.24 5.92 Age 14-15-16 gang m. (G) 6.81 5.30 transient gang members (except for one on seven cases seven cases), even in the first year of gang membership. The results obtained therefore partly support the selection model and partly the enhancement model. Data concerning transient gang members support only the facilitation model, while data on stable gang members support the enhancement model. Property offences The data on property offences (Table 5) are similar to those on person offences. The mean frequency of property offences remains relatively constant for youths who had never belonged to a gang from 14 through 16 years of age (never gang members) and for the age 14-15-16 gang members, but (except in the case of age 14 only gang members) crimes against property increase when the youths join a gang and diminish when they leave the gang. In the years of gang membership, the level of property offences of any group of gang members (transient or stable) is higher than that of the youths who had never belonged to a gang from 14 through 16 years of age (never gang members); furthermore, stable gang members display a higher level of this kind of offence than transient gang members (except for two out of seven cases). In general, similar patterns observed with regard to person offences are also seen for property of- fences, again confirming the facilitation model for transient gang members and the enhancement model for stable gang members. Drug sales Data on drug sales are reported in Tables 3 and 6. The overall picture for drug sales differs from that of the other delinquent behaviours considered so far. In particular, it should be noted that the frequency of drug sales increases over time for all youth categor- ies with few exceptions, including never gang mem- bers; this pattern contrasts with those of other types of crime, which remain constant over time, among never gang members. There is therefore an age or period effect underlying a certain increment in drug sales, regardless of the group effect. With regard to the transient gang members, the temporal comparisons reveal that the mean fre- quency of drug sales increases on joining the gang (indeed, this could be the result of the age effect), but does not diminish significantly on leaving. For stable gang members, drug sales do not vary significantly over time. Concerning group comparisons between the transient gang members and never gang mem- bers, if we look at the periods of gang membership, Table 6 Multivariate and univariate analysis of variance comparing the seven gang groups at each age separately and post hoc comparison using Scheff's test procedure when applicable (N = 755) Drug sales Age 14 Mean SD A B Never gang members (A) .03 .23 Age 14 only gang m. (B) .02 .13 Age 15 only gang m. (C) .05 .21 Age 16 only gang m. (D) .02 .15 Age 14-15 only gang m. (E) .27 .72 Age 15-16 only gang m. (F) .24 .69 Age 14-15-16 gang m. (G) Age 15 .31 .78 Mean SD A B Never gang members (A) .05 .30 Age 14 only gang m. (B) .18 .43 Age 15 only gang m. (C) .33 .65 Age 16 only gang m. (D) .02 .15 Age 14-15 only gang m. (E) .62 .92 Age 15-16 only gang m. (F) .52 .74 Age 14-15-16 gang m. (G) Age 16 .76 .96 Univariate F (6,752) = 8.31, p < .001 D E Univariate F (6,740) 26.22, p < .001 D E F F G Univariate F (6,738) = 25.21, p < .001 Mean SD A B C D E F G Never gang members (A) .12 .47 Age 14 only gang m. (B) .25 .54 Age 15 only gang m. (C) .31 .67 Age 16 only gang m. (D) .16 .43 Age 14-15 only gang m. (E) .45 .75 Age 15-16 only gang m. (F) .90 1.08 Age 14-15-16 gang m. (G) 1.12 1.14 we observe that the frequency of drug sales among gang members is significantly higher in only one case out of three (age 15 only and never gang members in the second period), while in the periods in which the transient members do not belong to a gang, the dif- ferences from the never gang members are not sig- nificant. During the years of gang membership, the stable gang members always have a higher level of drug sales than never gang members, and in some cases than transient gang members. In conclusion, the highest level of involvement in drug sales is chiefly seen among stable gang mem- bers; leaving the gang does not lead to a reduction in this type of offence, though this might depend on the age effect. Moreover, the results do not enable us to establish that any of the three models is at work. Drug use Tables 3 and 7 report the data on drug use. As in the case of drug sales, an age or period effect can be observed, in that the frequency of drug use increases over time for all of the categories considered with only few exceptions. Transient gang members show an increase in the frequency of drug use when they join the gang, but no significant decrease when they leave. As for drug sales, group comparisons between transient gang members and never gang members show that, in the periods of gang membership, the frequency of drug use among gang members is higher in only one case out of three (age 15 only and never gang members in the second period), and that in the periods in which the transient members do not belong to a gang, the differences from the never gang members are not significant. Group comparisons involving stable gang members also yield similar results to those obtained for drug sales; during the years of gang membership, the stable gang members always dis- play a higher level of drug sales than never gang members, and in some cases than transient gang members. Gang membership and risk factors Having analysed the rates of deviant behaviours among gang members and non-members and the trends in these behaviours in relation to joining and leaving the gang, we examined the relationship be- tween gang membership and delinquency by means of two new analyses. First, we attempted to establish whether the higher rate of delinquency reported during the period of gang membership depended on the influence of the gang, or whether it was instead determined by the personal and social deficiencies that might characterise individuals who become Table 7 Multivariate and univariate analysis of variance comparing the seven gang groups at each age separately and post hoc comparison using Scheff's test procedure when applicable (N = 755) Drug use Age 14 Mean SD A B Never gang members (A) .12 .51 Age 14 only gang m. (B) .24 .72 Age 15 only gang m. (C) .17 .55 Age 16 only gang m. (D) .02 .15 Age 14-15 only gang m. (E) .67 .99 Age 15-16 only gang m. (F) Age 14-15-16 gang m. (G) Age 15 .41 .82 .63 1.18 Univariate F (6,754)=9.32, p < .001 D E F G Univariate F (6,753) = 24.20, p < .001 Mean SD A B D E F G Never gang members (A) .23 .70 Age 14 only gang m. (B) .63 .85 Age 15 only gang m. (C) Age 16 only gang m. (D) Age 14-15 only gang m. (E) Age 15-16 only gang m. (F) Age 14-15-16 gang m. (G) Age 16 .89 1.36 .18 .44 1.32 1.30 .93 1.28 1.51 1.72 Univariate F (6,752) 29.55, p < .001 Mean SD A B Never gang members (A) .41 .77 Age 14 only gang m. (B) .85 .88 Age 15 only gang m. (C) .81 1.06 Age 16 only gang m. (D) .60 .96 Age 14-15 only gang m. (E) 1.03 1.10 Age 15-16 only gang m. (F) 1.45 1.30 Age 14-15-16 gang m. (G) 2.24 1.87 D E F G gang members (see Thornberry 1998; Thornberry et al., 2002). Second, we tried to ascertain whether the effect of the gang depended on the mere contact with delinquent friends, or whether it was also con- nected with specific effects exerted by the group. Indeed, it has been reported that a youth's individual delinquency correlates closely with peer delinquency and weakly with the gang structure (Morash, 1983), while others have observed that gang membership was strongly associated with delinquency above and beyond having delinquent peers (Battin et al., 1998). We carried out an analysis that took into account variables that could constitute risk factors with regard to both affiliation with a gang and delinquent behaviours. The risk factors considered concerned family breakdown, parents' occupational prestige, school performance, disruptive behaviour, parental supervision, friends' deviance, and previous delin- quency. These risk factors were assessed at ages 11, 12 and 13 prior to the analysis of the effect of belonging to a gang. An average score was computed across the three years to increase reliability. These previous risk factors were then correlated with three variables recorded over the following three years, at ages 14, 15 and 16 years: gang membership, the level of general delinquency and the condition of having at least one friend arrested by the police (delinquent friends). All of the factors that were systematically correl- ated with gang membership (previous delinquency, disruptiveness, lack of parental supervision, delin- quent friends) were also correlated with delinquency. The higher level of delinquency among gang mem- bers might therefore be due to social and family deficiencies that are present before the youth joins a gang. We then performed a multiple regression analysis in which the dependent variable was the mean fre- quency of deviant behaviour at ages 14, 15 and 16 years and the independent variables were gang membership in each of the three years considered, having friends arrested by the police in the same years, and the risk factors averaged across the pre- vious three years (i.e., family configuration, parental occupational prestige, teacher-rated disruptiveness, school difficulties, and parental supervision). This analysis allowed us to test whether gang member- ship made a unique contribution to subsequent delinquent behaviour, above and beyond all those control variables. The results of the multiple regression analysis (Table 8) revealed that gang membership signific- antly predicted the frequency of delinquency, above and beyond all other factors considered here. More- over, the results showed that gang membership and having delinquent friends exerted an independent