DEP

ARTMENT OF JUSTICE

OF

FIC

E

OF JUST I CE PRO

GR

AM

S

BJA

N

I JOJJ DP BJS

OV

C

U.S. Department of Justice

Office of Justice Programs

Office of Juvenile Justice and Delinquency Prevention

Shay Bilchik, Administrator

From the Administrator

Youth gangs are on the rise. Todaythey threaten virtually every majorcity, many small communities, andeven rural areas. The Office ofJuvenile Justice and DelinquencyPrevention (OJJDP) is committed tohelping communities overcome thisproblem, and an essential first stepis understanding the factors thatcontribute to the growth of youthgangs and the relationship of gangmembership to delinquency.

Gang Membership, DelinquentPeers, and Delinquent Behaviordescribes the findings of OJJDP-funded longitudinal research involv-ing juveniles in Seattle, WA, andRochester, NY. This research ad-dressed a fundamental question,“Does gang membership contributeto delinquency above and beyondthe influence of associating withdelinquent peers?” The answer wasyes in both cities, despite significantdifferences in demographics.

After describing study methodsand results, the authors summarizethe implications of their findings.One crucial implication is that com-munities developing comprehensiveapproaches to reducing juvenileviolence and victimization must con-sider the role of youth gangs andthe necessity of including youthgang prevention, intervention, andsuppression components.

Shay BilchikAdministrator

October 1998

Gang Membership,Delinquent Peers,and DelinquentBehaviorSara R. Battin-Pearson, Terence P. Thornberry,J. David Hawkins, and Marvin D. Krohn

The proliferation of youth gangs since1980 has fueled the public’s fear and mag-nified possible misconceptions about youthgangs. To address the mounting concernabout youth gangs, the Office of JuvenileJustice and Delinquency Prevention’s(OJJDP’s) Youth Gang Series delves intomany of the key issues related to youthgangs. The series considers issues such asgang migration, gang growth, female in-volvement with gangs, homicide, drugs andviolence, and the needs of communitiesand youth who live in the presence ofyouth gangs.

Gang membership intensifies delinquentbehavior. From the earliest to the mostrecent investigations, criminologists haveconsistently found that, when comparedwith youth who do not belong to gangs,gang members are far more involved indelinquency, especially serious and violentdelinquency. Associating with delinquentpeers also contributes to delinquency.Indeed, peer delinquency is one of thestrongest predictors of delinquency that

researchers have identified. However, theeffect of belonging to a gang has not beenseparated from the effect of simply asso-ciating with delinquent peers.

Some gang researchers have suggestedthat gang membership constitutes a quali-tatively different experience than merelyassociating with delinquent peer groups.For example, Moore states that “…gangsare no longer just at the rowdy end of thecontinuum of local adolescent groups—they are now really outside that con-tinuum” (1991:132). Klein makes a similarpoint: “…street gangs are something spe-cial, something qualitatively different fromother groups and from other categories oflaw breakers” (1995:197). Although theseand other researchers view gangs as“qualitatively different,” until recently nostudy had attempted to disentangle theinfluence of gang membership from theeffects of delinquent peers on involvementin delinquency.

In 1997, studies conducted by theSeattle Social Development Project andthe Rochester Youth Development Studywith funding from OJJDP both answeredthe question, “Does gang membershipcontribute to delinquency above and be-yond the influence of associating with

The results from the Seattle Social Development Projectreported in this Bulletin were originally published inCriminology 36(1): 93–115, 1998, American Society ofCriminology.

2

delinquent peers?” Findings from the twostudies are presented in this Bulletin.

Seattle SocialDevelopment Project

Project OverviewThe Seattle Social Development Project

(SSDP) is a longitudinal study guided by thesocial development model (Catalano andHawkins, 1996), which incorporates infor-mation on how protective and risk factorswork together to enhance both positive andantisocial development. The model buildson differential association theory (Cressey,1953; Matsueda, 1988), social learningtheory (Bandura, 1977), and social controltheory (Hirschi, 1969). The model hypoth-esizes that socialization follows the sameprocesses whether it produces prosocial orproblem behavior and suggests that devel-opment of prosocial or antisocial behavioris influenced by the degree of involvementand interaction with prosocial or delin-quent peers (differential association), theskills required and the costs and rewardsfor that interaction (social learning), andthe extent to which the youth subsequently

become bonded to prosocial or antisocialindividuals (social control).

The study has followed a multiethnicurban sample of 808 children since theyentered the fifth grade in 1985. The sampleincludes nearly equal numbers of males(n=412) and females (n=396). Slightlyfewer than half (46 percent) identifiedthemselves as European-Americans.African-Americans (24 percent) and Asian-Americans (21 percent) also made upsubstantial portions of the sample. Theremaining youth were Native-American(6 percent) or of other ethnic groups(3 percent). Forty-six percent of respon-dents’ parents reported a maximum fam-ily income under $20,000 per year in 1985,and more than half of the sample (52 per-cent) participated in the National SchoolLunch/School Breakfast Program at somepoint in the fifth through seventh grades,indicating that they came from families inpoverty. The analyses presented in thisBulletin are based on surveys conductedwhen the youth were age 13 (n=654), 14(n=778), and 15 (n=781). Sample sizes varyfor each assessment year based on thenumber of respondents who completedthe interview in that year. Nonparticipa-

tion was not related to gender, lifetimeuse of tobacco or alcohol, or participationin delinquency by age 10, nor was it con-sistently related to ethnicity. Data wereobtained from the youth and from KingCounty court records.

MethodsTo determine whether gang member-

ship contributes to delinquency aboveand beyond associating with delinquentpeers, the SSDP sample was divided intothe following three groups:

◆ Gang members: Respondents who self-reported membership in a gang in thepast year and who identified the gangby name.

◆ Youth with delinquent peers: Respon-dents who were not members of a gangin the survey year but who reportedthat at least two of their three bestfriends had been arrested or donethings that could get them in troublewith the police.

◆ Youth with nondelinquent peers:Respondents who were not membersof a gang in the survey year and who

Table 1: Classification of Individual Offense Rates (Seattle Social Development Project)

Category Offense or Frequency

Self-reported IOR’s*Violent Hitting teacher, hitting to hurt, picking a fight, using force to get things,

throwing objects

Nonviolent Taking something worth more than $50, taking something worth between $5 and$50, breaking into a house, destroying property, writing graffiti, selling illegal drugs

General Combined self-reported violent and nonviolent offenses and frequency of beingarrested and in trouble with the police

Court-recorded IOR’sViolent Simple assault, aggravated assault, hit and run, murder, threat, robbery, sex offense,

disorderly conduct, using a weapon

Nonviolent Arson, reckless arson, burglary, larceny, motor vehicle theft, trespassing,prostitution, stolen property, selling illegal drugs

General Combined court-recorded violent and nonviolent offenses

Self-reported rates of drug sellingand substance use

Drug selling Past-year frequency

Alcohol use Past-month frequency

Binge drinking Past-month incidence of drinking five or more drinks in a row

Marijuana use Past-year frequency

Illicit drug use Past-year frequency of using crack, other forms of cocaine, amphetamines,tranquilizers, sedatives, narcotics, psychedelics

*Individual offense rates.

3

reported that only one or none of theirthree best friends had been arrested ordone things that could get them introuble with the police.

These three groups were comparedaccording to various measures of delin-quency and substance use to determinewhether there were significant differencesin their rates of offending. The respondent’sindividual offense rate (IOR), which is theactual frequency of committing the offenseslisted in table 1, was used as the measureof delinquency and substance use.

ResultsThe analysis was done cross-sectionally

(comparing age 15 group status with age15 behaviors) and longitudinally (com-paring age 14 group status with age 15behaviors). The cross-sectional results atage 15 are presented in figures 1, 2, and 3.Results from the longitudinal comparisonare similar to the cross-sectional resultsand therefore are not presented. Figure 1presents the mean, or average, IOR’s forself-reported delinquency during the pastyear; figure 2 presents the mean IOR’s forcourt-recorded delinquency. Figure 3 pre-sents annual rates for measures of self-reported drug selling and substance use.An asterisk has been placed next to thevariables for which mean delinquency rateswere significantly higher for gang membersthan for youth with delinquent peers.

A consistent pattern of offending wasfound across the 3 status groups for all 11measures of delinquency and substanceuse. On all measures of delinquency andsubstance use, rates of offending werelowest for youth with nondelinquentpeers, higher for youth with delinquentpeers, and highest for gang members.For example, as shown in figure 1, youthwith nondelinquent peers committed anaverage of 1.6 self-reported acts of violentdelinquency in the past year, while youthwith delinquent peers committed an aver-age of 5.1 violent acts and gang memberscommitted more than 11 violent acts.

For this analysis, t-tests were conductedto determine whether observed differencesin offending between gang members andnongang youth with delinquent peerswere statistically significant. Gang mem-bers had significantly higher offense rateson 9 of the 11 measures of delinquencyand substance use—that is, at age 15,gang members committed significantlymore of the following acts than nongangyouth with delinquent peers (as indicatedby an asterisk in figures 1, 2, and 3):

Figure 1: Self-Reported Individual Offense Rates at Age 15(Seattle Social Development Project)

*An asterisk indicates that the rates for gang members are significantly higher than those foryouth with delinquent peers (t-test, p<0.05).

Note: IOR, individual offense rate.

0

5

10

15

20

25

30

Violent IOR Nonviolent IOR General IOR

Gang members (n=51)

1.6

5.1

11.2*

0.8

5.1

8.8*

3.1

13.0

25.6*

Nondelinquent peers (n=643) Delinquent peers (n=87)

Pas

t-Ye

ar F

req

uen

cy

Type of Self-Reported Offense

Figure 2: Court-Recorded Individual Offense Rates at Age 15(Seattle Social Development Project)

*An asterisk indicates that the rates for gang members are significantly higher than those foryouth with delinquent peers (t-test, p<0.05).

Note: IOR, individual offense rate.

0.0

0.5

1.0

1.5

2.0

Violent IOR Nonviolent IOR General IOR

0.1

0.3

0.8*

0.1

0.5

1.0*

0.2

0.7

1.8*

Pas

t-Ye

ar F

req

uen

cy

Type of Court-Recorded Offense

Gang members (n=51)Nondelinquent peers (n=643) Delinquent peers (n=87)

4

Figure 3: Self-Reported Rates of Drug Selling and SubstanceUse at Age 15 (Seattle Social Development Project)

*An asterisk indicates that the rates for gang members are significantly higher than those foryouth with delinquent peers (t-test, p<0.05).

◆ Self-reported acts of violent, nonviolent,and general delinquency.

◆ Court-recorded acts of violent, nonvio-lent, and general delinquency.

◆ Self-reported drug selling, marijuanause, and alcohol use.

In summary, gang membership wasassociated with increased participationin various acts of delinquency and sub-stance use, even in comparison with youthwho associate with delinquent peers. Itwould thus appear that gang membershipdoes contribute to delinquency over andabove associating with delinquent peers.However, it is also possible that delin-quency rates are higher among gang mem-bers because they also associate with de-linquent peers. Therefore, the observedeffect of gang membership may actuallyderive from the simple fact that gangmembers have a lot of delinquent friends.

To rule out this possibility, a statisticaltechnique called structural equationmodeling was used; this technique testscausal relationships among a variety ofvariables at the same time. It was used toexamine the impact of gang membershipon delinquency after controlling for asso-ciation with delinquent peers. Structuralequation modeling provides four kinds ofinformation:

◆ The path coefficient, an estimate of thestrength of the causal relationship,that can range from –1 to +1.

◆ R2, the amount of a given behavior thatis explained by prior variables in themodel. R2 can range from 0 to 1.

◆ An acknowledgment that factors otherthan those included in the model cancontribute to the behavior (called the“error” and not usually quantified).

◆ A measure of the overall fit of themodel that can range from 0 to 1.

Specifically, the effect of gang mem-bership on delinquency at age 15 wasexamined, controlling for associationwith delinquent friends at ages 14 and15 and for delinquency at age 13. If gangmembership provides a unique andstrong contribution to delinquencyabove and beyond that made by associ-ating with delinquent peers and previousdelinquency, then the path coefficientsfrom gang membership to delinquencyshould be significant in the causal mod-els presented in figures 4 and 5.

The results revealed that gang mem-bership contributed to delinquency aboveand beyond associating with delinquentpeers and previous delinquent behavior.As shown in figure 4, the paths from gang

membership at age 14 and at age 15 toself-reported general delinquency at age15 were significant, even when associat-ing with delinquent friends and previousdelinquency were included in the model(path coefficients of 0.18 and 0.22, respec-tively, p<0.01). Similar patterns were foundfor court-recorded delinquency, as shownin figure 5.

Overall, SSDP respondents who weregang members always had the highestrates of delinquency and substance use.For 9 of the 11 delinquency and substanceuse measures, rates for gang memberswere significantly higher than those foryouth with delinquent peers. In addition,structural equation modeling revealedthat gang membership contributed to de-linquency even after the effects of delin-quent peers and previous delinquencyhad been accounted for.1

Rochester YouthDevelopment Study

Project OverviewThe Rochester Youth Development

Study (RYDS) is a longitudinal studyof the development of delinquencyand drug use, guided by interactionaltheory (Thornberry, 1987) and social net-work theory (Krohn, 1986). According tointeractional theory, delinquency comesabout because of the pattern of interac-tions between the individual and his or herenvironment. As bonds to conventionalsociety (e.g., parents and teachers) weaken,social control is reduced and delinquencybecomes more likely. For prolonged seriousdelinquency to emerge, however, associa-tion with other delinquent youth and theformation of delinquent beliefs are re-quired. Once these delinquent patternsemerge, they have feedback effects, furthereroding the person’s bond to conventionalsociety. These mutually reinforcing effectscreate trajectories toward increasing levelsof involvement in delinquency. Socialnetwork theory is a complementary per-spective that focuses on the impact of thesocial groups, or networks, in which theperson is involved. All networks controlthe behavior of their members and channelthat behavior toward consistency withgroup norms. Prosocial networks (e.g., BoyScouts) increase the likelihood of conform-ing behavior; antisocial networks (e.g.,

0

5

10

15

20

25

30

35

40

Drug Selling

Marijuana Use

Alcohol Use

0 0.41.3* 1.7

10.2

6.4

16.4

1.7

31.1*

5.5

10.0

0.2 0.8 1.1

Binge Drinking

Other Illicit Drug Use

Gang members (n=51)Nondelinquent peers (n=643) Delinquent peers (n=87)

Pas

t-Ye

ar F

req

uen

cy

Type of Self-Reported Offense

38.6*

1 A complete description of these analyses can befound in Battin et al. (1998).

5

gangs) increase the likelihood of antisocialbehavior. The more pervasive the networkis in a person’s life, the more powerful theeffect it has on his or her behavior.

The Rochester study has followed asample of 1,000 urban adolescents initiallyselected in 1988, when they were in eitherthe seventh or eighth grade in the Roches-ter, NY, public schools. They have beenfollowed until the present and are now 22years of age on average. The sample is 75percent male and 25 percent female andis composed primarily of minority groupmembers—68 percent African-American,17 percent Hispanic (mostly Puerto Rican),and 15 percent white. Although the sampleoverselected youth at elevated risk forserious delinquency, the results presentedhere are statistically adjusted to representthe entire population of seventh andeighth grade students in the Rochesterpublic schools.

MethodsEach student was interviewed at

6-month intervals over the course of themiddle school and high school years. Thedata analyzed in this Bulletin were takenfrom interviews covering ages 14 and 15

for the subjects. Age 15 is near the peakage of involvement for both gang mem-bership and delinquency (Loeber andFarrington, 1998).

An analysis strategy similar to thatemployed with the SSDP data was used toexamine the RYDS data. First, the samplewas divided into respondents who indi-cated that they had been a member of ayouth gang during the 6 months since theprevious interview and those who werenot gang members. Second, respondentswho were not gang members during thisperiod were divided into quartile group-ings based on their responses to the delin-quent peer associations scale. Using a4-point response scale ranging from “noneof them” to “most of them,” each respon-dent reported how many of his or herpeers were involved in eight delinquentactivities. The lowest quartile representsthe respondents who had the fewest delin-quent peers; the highest quartile repre-sents those who had the most delinquentpeers. The division of nonmembers intoquartiles allows for a much finer compari-son of gang members with nonmemberssince the nonmembers in the highestquartile are very heavily involved with

delinquent peers. The groups were com-pared in terms of the frequency with whichthey self-reported general delinquency,violent delinquency, drug selling, and druguse (see table 2). Comparisons were madeseparately for males and for females (seefigures 6 and 7).

ResultsFigure 6 shows the comparison of male

gang members with nonmembers in termsof the frequency of general delinquency,violent delinquency, drug selling, anddrug use. Among those who were notgang members, offense rates for all fourtypes of offenses were higher for the re-spondents who scored higher on the de-linquent peer associations scale. Moreimportant, however, was the finding thatrespondents who were gang members al-ways had the highest rates of offending.

The results of the comparison of of-fense rates of gang members and non-members with delinquent peers in Roch-ester provide a strikingly similar pictureto those obtained with the Seattle data.Although associating with delinquentpeers is related to offense rates, being amember of a gang facilitates delinquency

0.41

0.17

0.12

0.27 R 2=0.08

0.23

R 2=0.080.22

0.18

0.35

0.09

0.21

R 2=0.22R 2=0.17

e2e5

e4 e1

e3

0.37

R 2=0.45

Figure 4: Structural Equation Model: Self-Reported General Delinquency as Outcome Measure(Seattle Social Development Project)

Note: Path coefficients are indicated above the directional arrows. All path coefficients are significant at p≤0.05, with the exception of those markedby an asterisk. Values for R 2 (explained variance) for each predicted variable are noted above the boxes, as are the errors (e). Goodness of fit (GFI)measures indicate acceptably fitting models (GFI=0.975).

DelinquentFriendsAge 14

DelinquentFriendsAge 15

Self-ReportedGeneral Delinquency

Age 15

Self-ReportedGeneral Delinquency

Age 13

GangMembership

Age 14

GangMembership

Age 15

6

Table 2: Self-Reported Delinquency Indices(Rochester Youth Development Study)

Index Definition

General delinquency 32-item index of past-year frequency of offenses,ranging from running away from home to assault witha weapon (violent delinquency items are also included ingeneral delinquency)

Violent delinquency Past-year frequency of assault with a weapon, assaultwithout a weapon, throwing objects at people, robbery, rape

Drug selling Past-year frequency of selling marijuana and hard drugs

Drug use Past-year frequency of use of marijuana, inhalants, LSD,cocaine, crack, heroin, phencyclidine (angel dust),tranquilizers, downers, uppers

Figure 5: Structural Equation Model: Court-Recorded General Delinquency as Outcome Measure(Seattle Social Development Project)

Note: Path coefficients are indicated above the directional arrows. All path coefficients are significant at p≤0.05, with the exception of thosemarked by an asterisk. Values for R2 (explained variance) for each predicted variable are noted above the boxes, as are the errors (e). Goodnessof fit (GFI) measures indicate acceptably fitting models (GFI=0.963).

over and above that effect. For violentdelinquency among male respondents, forexample, there is an increase in the levelof offending across the four categoriesof nonmembers—from 0.2 for those withfew delinquent peers to 2.2 for those whohave the highest level of association withdelinquent peers. However, the mean forgang members (4.9) is more than twiceas high. This rate is significantly differentfrom the rate for nonmembers in thehighest quartile of delinquent peers. Thisfinding is particularly important becausenonmembers in the highest quartile ofinvolvement with delinquent peers asso-ciate with delinquent peers as much asgang members do. This pattern is alsoobserved for general delinquency, drugselling, and drug use.

Figure 7 examines the same relation-ships for female respondents. Female in-volvement in delinquency and drugs waslower than male involvement; as a result,the patterns are somewhat less consistent,especially for nonmembers. The mostimportant comparison, however, is be-tween female gang members and nonmem-bers in the highest quartile. In all cases,gang members reported significantly

higher involvement as compared with non-members. There is a particularly strikingeffect for drug selling—among female re-spondents, only gang members sold drugs.

As with the earlier Seattle analysis, thisanalysis does not control for the impactof association with delinquent peers. Theearlier analysis of the Seattle data con-trolled for the effect of delinquent peers in

examining the impact of gang membershipon violent delinquency. To provide a morerigorous examination of whether gang mem-bership has an effect on offense rates, theRYDS controlled for five additional risk fac-tors that covered the domains of family(poverty level and parental supervision),school, stress, and prior delinquency. Theseadditional variables test the possibility that

0.26

0.08

0.14

0.17 R 2=0.03

0.24

R 2=0.09 0.16

0.16

0.33

0.06*

0.10

R 2=0.21R 2=0.07

e2e5

e4 e1

e3

0.43

R 2=0.26

DelinquentFriendsAge 15

Court-RecordedGeneral Delinquency

Age 15

GangMembership

Age 15

GangMembership

Age 14

Court-RecordedGeneral Delinquency

Age 13

DelinquentFriendsAge 14

7

rates of violence are high for gang membersnot because of a gang effect, but because ofthe accumulation of risk in their back-grounds. That is, it may not be gang mem-bership that brings about the higher ratesof violence; it may instead be other risk fac-tors that are related to gang membershipand to delinquency. The variables that wereheld constant here include family povertylevel, parental supervision, commitment toschool, negative life events, previous in-volvement in violence, and association withdelinquent peers. The risk factors weremeasured at the interview prior to the yearof gang membership. The analysis is limitedto males because of the relatively smallnumber of female gang members.

The results in table 3 indicate that evenwhen the variables listed above are heldconstant, gang membership still exerts astrong impact on the incidence of violentbehavior. The standardized coefficient forgang membership is 0.28, approximatelythe same magnitude of coefficients ob-served for previous violence (0.27). In-deed, gang membership has the greatestimpact on violent behavior of any of thevariables included in the equation.2

Figure 6: Self-Reported Delinquency Rates at Age 15 for MaleGang Members and Nonmembers With Delinquent Peers(Rochester Youth Development Study)

*An asterisk indicates that the rates for gang members are significantly higher than those fornonmembers in the highest quartile of association with delinquent peers (t-tests, p<0.05).†Nonmembers are divided into quartiles of association with delinquent peers.

I Low (n=176)

0

10

20

30

40

50

60

70

80

General Delinquency

Violent Delinquency

Drug Selling

1.4

6.3

17.0

40.9

0.2 0.82.2

4.9*

70.0*

0 1.2

10.1*

0 0.4 0.5

Drug Use

0.2

5.8

0

20.8

26.6*

II (n=128) III (n=130) Gang members (n=68)IV High (n=146)

Nonmembers with delinquent peers:†

Pas

t-Ye

ar F

req

uen

cy

Type of Self-Reported Offense

Figure 7: Self-Reported Delinquency Rates at Age 15 for FemaleGang Members and Nonmembers With Delinquent Peers(Rochester Youth Development Study)

*An asterisk indicates that the rates for gang members are significantly higher than those fornonmembers in the highest quartile of association with delinquent peers (t-tests, p<0.05).†Nonmembers are divided into quartiles of association with delinquent peers.

0

20

40

60

80

100

General Delinquency

Violent Delinquency

Drug Selling

03.8

26.9

22.3

0 0.7 1.25.2*

87.9*

5.0*0 2.2

Drug Use

0.1 0

6.5

11.7*

0 0 0 0

I Low (n=65) II (n=63) III (n=44) Gang members (n=19)IV High (n=55)

Pas

t-Ye

ar F

req

uen

cy

Type of Self-Reported OffenseNonmembers with delinquent peers:†

Table 3: Impact of GangMembership and Various RiskFactors on the Incidence ofSelf-Reported Violence,Males Only (Rochester YouthDevelopment Study)

Self-ReportedViolence at

Risk Factor Year 2 (Logged)*

Gang membership 0.28†

Family poverty level –0.06†

Parental supervision –0.04

Commitment to school –0.02

Negative life events 0.12†

Prior violence 0.27†

Delinquent peers 0.06

R2=0.34

n=484

*Standardized ordinary least squaresregression coefficients.†p<0.05

2 More detailed information on these results can befound in Thornberry (1998) and Krohn and Thornberry(in press). In the latter report, more refined measuresof highly delinquent peer groups (e.g., using decilesrather than quartiles) generate results a little moremuted than those reported here.

8

Program of Research on the Causes and Correlates of Delinquency

same juveniles over a substantial por-tion of their developmental years.

The research teams on the threeprojects collaborated in creating themost comprehensive, common mea-surement package ever used in delin-quency research. Thus, each of thethree sites uses core measures to col-lect data on a wide range of key vari-ables, including delinquent behavior,drug use, juvenile justice system in-volvement, community characteristics,family experiences, peer relationships,educational experiences, attitudes andvalues, and demographic characteris-tics. This allows for comparison acrosssites on common measures and theopportunity to reach more valid con-clusions regarding cross-site similaritiesand differences on such factors as theage of onset of violent crime.

In each project, researchers conductface-to-face interviews with individual ju-veniles in a private setting to collect self-report information on the nature and fre-quency of serious violent behavior. Theadvantage of using self-report data,rather than juvenile justice records of ar-rests, is that researchers come muchcloser to measuring actual violent be-haviors and ascertaining when a violentcareer began. Multiple perspectives oneach child’s development and behaviorwere obtained through interviews withthe child’s primary caretaker and, when-ever possible, teachers. In addition to in-terview data, the studies have collectedextensive data from official records suchas school, police, and juvenile court.This provides comparison data on therelationship between self-reportedbehavior and that which is officiallydetected and recorded.

In an effort to learn more about the rootcauses of juvenile delinquency andother problem behaviors, the Office ofJuvenile Justice and Delinquency Pre-vention (OJJDP) is sponsoring the Pro-gram of Research on the Causes andCorrelates of Delinquency. Serious de-linquency and drug use are major prob-lems in American society. Past researchindicates that many variables correlatewith delinquency and that many factorstend to increase the risk of later delin-quent behavior. Among these risk fac-tors are birth trauma, child abuse andneglect, ineffective parental discipline,family disruptions, conduct disorder andhyperactivity in children, school failure,learning disabilities, negative peer influ-ences, limited employment opportuni-ties, inadequate housing, and residencein high-crime neighborhoods.

Overall, research findings supportthe conclusion that no single cause ac-counts for all delinquency and that nosingle pathway leads to a life of crime.To date, however, research has notclearly identified all the causal path-ways that lead to delinquency or thefactors that cause different individualsto take different paths. There is generalagreement among social scientists andpolicymakers that longitudinal studiesare the best way to gain information onthe causes of delinquency. This type ofinvestigation involves repeated contactswith the same individuals so that pat-terns of development can be studied.The strength of the longitudinal designis that it permits researchers to sort outwhich factors precede changes in of-fending, to predict such changes, andto do so independent of other factors.With the aid of repeated measures, itis possible to identify pathways todelinquency, each with unique causal

factors that, like delinquency itself, maychange with time. Successfully accom-plishing this will provide the informationneeded to develop truly effective inter-vention programs.

OJJDP has been in the forefront of sup-porting basic, long-term research thatprovides the hard empirical informationneeded to design effective action pro-grams. The Program of Research on theCauses and Correlates of Delinquencyis an example of OJJDP’s support forlong-term research. The Causes andCorrelates program, initiated in 1986,includes three coordinated longitudinalprojects: the Denver Youth Survey, di-rected by Dr. David Huizinga at the Uni-versity of Colorado; the Pittsburgh YouthStudy, directed by Dr. Rolf Loeber at theUniversity of Pittsburgh; and the Roch-ester Youth Development Study, directedby Dr. Terence P. Thornberry at the Uni-versity at Albany, State University ofNew York. This program represents amilestone in criminological research be-cause it constitutes the largest shared-measurement approach ever achieved indelinquency research. From the begin-ning, the three research teams workedtogether to ensure that they used similarmeasurement techniques, thus enhanc-ing generalizability by allowing for analy-ses that include all three sites.

The Causes and Correlates studies aredesigned to improve the understandingof serious delinquency, violence, anddrug use through the examination ofhow individual youth develop within thecontext of family, school, peers, and thecommunity. While each of the threeprojects has unique features, theyshare several key elements. All of theprojects are longitudinal investigationsthat involve repeated contacts with the

SummaryAlthough research has consistently

found that gang members are more in-volved in serious and violent delinquentoffenses than nonmembers, the effect ofbelonging to a gang has not been sepa-rated from the effect of simply associat-ing with delinquent peers. Longitudinaldata from both the SSDP and the RYDSprovide strong and consistent evidencethat being a member of a gang increasesthe rate of involvement in a variety of

deviant behaviors over and above theimpact of having delinquent peers. In-deed, gang membership significantlypredicts delinquency, even when con-trolling for other predictors of bothdelinquency and gang membership.

The consistency and strength of theresults of each study are convincing evi-dence concerning the impact of gangmembership on deviant behavior. Evenmore impressive, however, is the consis-tency of the results across the two stud-

ies. The SSDP and the RYDS have beenconducted in cities that differ in theirhistories and demographic characteris-tics. For example, the majority of RYDSrespondents were African-American(68 percent), while most SSDP respon-dents were European-American (46 per-cent). The studies also used somewhatdifferent measures and included some-what different variables in the multivari-ate equations. Yet both studies came tothe same fundamental conclusion. The

Continued on next page

9

The three longitudinal studies are pro-spective in nature; that is, subjects arerepeatedly contacted to report on theircurrent and recent violent activities.Deterioration of recall is minimized byavoiding lengthy gaps between inter-views. Reporting periods were either6 or 12 months, and all self-report vio-lence data have been calculated forannual periods. Sample retention hasbeen excellent; as of 1997, at least84 percent of the subjects had beenretained at each of the sites, and theaverage rate of retention across allinterview periods was 90 percent.

Samples were carefully drawn to cap-ture inner-city youth considered at highrisk for involvement in delinquencyand drug abuse. The samples can bedescribed as probability samples, inwhich youth at greater risk areoversampled.

◆ Denver’s sample includes 1,527youth (806 boys and 721 girls) whowere 7, 9, 11, 13, and 15 years oldwhen data collection commenced in1988. This sample represents thegeneral population of youth residingin 20,000 households in high-riskneighborhoods in Denver.

◆ Pittsburgh’s sample consists of1,517 boys who ranged in age from7 to 13 years and attended grades 1,4, and 7 when data collection beganin 1987. This sample represents thegeneral population of boys attendingPittsburgh’s public schools.

◆ Rochester’s sample of 1,000 youth(729 boys and 271 girls) was drawnfrom students attending grades 7 and8. This sample represents the entirerange of seventh and eighth gradestudents attending Rochester’s publicschools.

The Causes and Correlates program hascontributed to an understanding of a vari-ety of topics related to juvenile violenceand delinquency, including developingand testing causal models for chronicviolent offending; examining interrelation-ships among gang involvement, drugselling, and gun ownership/use; changesover time in delinquency and drug use;and neighborhood, individual, and socialrisk factors for serious juvenile offenders.Major findings from the three projects todate include the following:

◆ Delinquency, drug use, and otherproblem behaviors begin at earlierages than previously thought. Formany children, these behaviors areevident before the teenage years.The co-occurrence of problem be-haviors is also quite common. Seri-ous delinquents are likely to beinvolved in drug use, precocioussexual activity, school failure, juve-nile gangs, gun ownership, andother related behaviors.

◆ There has been a shift in the demo-graphic characteristics of adolescentviolent offenders. Older males, chil-dren (as young as 10 years old),and females reported greater in-volvement in serious violence thanwould have been expected fromprevious research.

◆ The development of disruptive anddelinquent behavior in boys gener-ally takes place in an orderly, pro-gressive fashion, with less seriousproblem behaviors preceding moreserious problems. Three distinctdevelopmental pathways were iden-tified: authority conflict (e.g., defi-ance and running away), covertactions (e.g., lying and stealing),and overt actions (e.g., aggression

and violent behavior). Individualsmay proceed along single or mul-tiple developmental pathwaystoward serious antisocial behavior.

◆ Childhood maltreatment is associ-ated with an increased risk of atleast 25 percent for engaging in ahost of adolescent problem behav-iors: serious and violent delinquency,drug use, poor performance inschool, mental illness, and teenagepregnancy. Furthermore, a historyof maltreatment nearly doublesthe risk that teenagers will experi-ence multiple problems duringadolescence.

Each project has disseminated theresults of its research through a broadrange of publications, reports, andpresentations.

In 1997, OJJDP initiated the YouthDevelopment Series, a series of Bulle-tins created to present findings from theProgram of Research on the Causesand Correlates of Delinquency. To date,four Bulletins have been released:Epidemiology of Serious Violence, Inthe Wake of Childhood Maltreatment,Developmental Pathways in Boys’Disruptive and Delinquent Behavior,and Gang Members and DelinquentBehavior.

For more information on OJJDP’sCauses and Correlates studies or toobtain copies of the Youth DevelopmentSeries Bulletins or other Youth GangSeries Bulletins, contact the JuvenileJustice Clearinghouse by telephoneat 800–638–8736; by mail at P.O. Box6000, Rockville, MD 20849–6000; by e-mail at askncjrs@ncjrs.org; or byviewing OJJDP’s home page.

fact that both studies generated resultsthat led to the same interpretation rein-forces the conclusion that the observedeffect of gang membership on involve-ment in delinquency is not unique to onecity or to one ethnic group.

Implications for Theoryand Practice◆ There are national implications

from the two studies. The consis-

tency of results and conclusionsobtained in the two studies, whichwere conducted in two diverse com-munities, suggests that similar dy-namics are likely to be operating inother areas. Given the recent spreadof gangs to more and more citiesacross America (Thornberry, 1998),these findings underscore the im-portance of developing effectivegang prevention and suppressionprograms.

◆ Gang membership has an independentcontributing role in the etiology of de-linquency over and above other riskand protective factors. These findingspoint to the tremendous importance ofstreet gangs to understanding the dy-namics of delinquency, especially seri-ous and violent delinquency. They alsoindicate that it may not be enough tointervene only with regard to risk fac-tors in the family, school, and similarareas. Specific attention must be given

10

to understanding the dynamics of gangsthat produce these effects and then indeveloping appropriate interventionprograms.

◆ Preventing youth from joining gangsholds promise for preventing andreducing crime and substance use.Because gangs have such a major effecton delinquent behavior, prevention ef-forts aimed at reducing delinquencyand substance use should seek to pre-vent and reduce gang involvement.

◆ Determining why youth join andleave gangs may provide informationfor prevention programs. Becausegang members are so much more in-volved in delinquency and substanceuse than nonmembers, understandingwhy they join and leave gangs mayhave great practical value. Such anunderstanding may lead to programsto keep some youth out of gangs inthe first place or to shorten periods ofactive membership for those who dojoin. If successful, these programsshould have an impact on reducingthe level of juvenile delinquency anddrug involvement.

ReferencesBandura, A. 1977. Self-efficacy: Toward a

unifying theory of behavioral change. Psycho-logical Review 84(2):191–215.

Battin, S.R., Hill, K.G., Abbott, R.D.,Catalano, R.F., and Hawkins, J.D. 1998. Thecontribution of gang membership to delin-quency beyond delinquent peers. Criminol-ogy 36(1):93–115.

Catalano, R.F., and Hawkins, J.D. 1996. TheSocial Development Model: A theory ofantisocial behavior. In Delinquency and Crime:Current Theories, edited by J. David Hawkins.Cambridge, MA: Cambridge University Press.

Cressey, D.R. 1953. Other People’s Money.New York, NY: The Free Press.

Hirschi, T. 1969. Causes of Delinquency.Berkeley, CA: University of California Press.

Klein, M.W. 1995. The American StreetGang: Its Nature, Prevalence, and Control.New York, NY: Oxford University Press.

Krohn, M.D. 1986. The web of conformity:A network approach to the explanation ofdelinquent behavior. Social Problems33(6):581–593.

Krohn, M.D., and Thornberry, T.P., eds. Inpress. Gangs and other law violating groups.In Gang Membership and Delinquency: StreetGangs in Developmental Perspective. Report.

Washington, DC: U.S. Department of Justice,Office of Justice Programs, Office of JuvenileJustice and Delinquency Prevention.

Loeber, R., and Farrington, D.P., eds. 1998.Serious and Violent Juvenile Offenders: RiskFactors and Successful Interventions. Thou-sand Oaks, CA: Sage Publications, Inc.

Matsueda, R.L. 1988. The current state ofdifferential association theory. Crime & Delin-quency 34(3):277–306.

Moore, J. 1991. Going Down to the Barrio:Homeboys and Homegirls in Change. Philadel-phia, PA: Temple University Press.

Thornberry, T.P. 1987. Toward an interac-tional theory of delinquency. Criminology25(4):863–891.

Thornberry, T.P. 1998. Membership inyouth gangs and involvement in serious

Acknowledgments

The authors would like to express their appreciation to colleagues on the SeattleSocial Development Project and the Rochester Youth Development Study. We arevery grateful to the dedicated staff who collected and processed the data and tothe participants for their willingness to be interviewed repeatedly for this study.Without their assistance, the research could not have been conducted.

Sara R. Battin-Pearson, M.Ed., is a research analyst for the Seattle Social Devel-opment Project at the University of Washington. She is currently the primary ana-lyst for the OJJDP-funded study titled The Dynamics of Gang Membership andDelinquency. Her expertise is in measurement, statistics, and research design. Herresearch interests include the etiology of adolescent delinquency, substance use,and mental health problems.

Terence P. Thornberry, Ph.D., is a professor and former dean at the School ofCriminal Justice, University at Albany, State University of New York. He is the au-thor of The Criminally Insane, From Boy to Man—From Delinquency to Crime, andnumerous articles and book chapters. His research interests focus on the longitu-dinal examination of the development of delinquency and crime and the construc-tion of an interactional theory to explain these behaviors.

J. David Hawkins, Ph.D., is a professor of social work and the Director of the So-cial Development Research Group at the University of Washington. His researchfocuses on understanding and preventing child and adolescent health and behav-ior problems. He is also committed to translating research into effective practiceand policy to improve adolescent health and development. Since 1981, he hasbeen conducting the Seattle Social Development Project, a longitudinal preventionstudy based on his theoretical work.

Marvin D. Krohn, Ph.D., is a professor in the Department of Sociology, Universityat Albany, State University of New York. His research interests include the investi-gation of social psychological theories of adolescent substance abuse and delin-quent behavior. He is currently involved in a panel study of inner-city youth de-signed to examine hypotheses derived from those perspectives.

Research for the Seattle Social Development Project and the Rochester Youth De-velopment Study was supported by OJJDP under grants 95–JD–FX–0017 and96–MU–FX–0014, respectively. The Seattle Social Development Project was alsosupported by grants from the National Institute on Drug Abuse and the RobertWood Johnson Foundation.

The Office of Juvenile Justice and Delin-quency Prevention is a component of the Of-fice of Justice Programs, which also includesthe Bureau of Justice Assistance, the Bureauof Justice Statistics, the National Institute ofJustice, and the Office for Victims of Crime.

Points of view or opinions expressed in thisdocument are those of the authors and do notnecessarily represent the official position orpolicies of OJJDP or the U.S. Department ofJustice.

and violent offending. In Serious & ViolentJuvenile Offenders: Risk Factors and Success-ful Interventions, edited by R. Loeber andD.P. Farrington. Thousand Oaks, CA: SagePublications, Inc.

11

Share With Your Colleagues

Unless otherwise noted, OJJDP publications are not copyright protected. Weencourage you to reproduce this document, share it with your colleagues, andreprint it in your newsletter or journal. However, if you reprint, please cite OJJDPand the authors of this Bulletin. We are also interested in your feedback, such ashow you received a copy, how you intend to use the information, and how OJJDPmaterials meet your individual or agency needs. Please direct your comments andquestions to:

Juvenile Justice ClearinghousePublication Reprint/FeedbackP.O. Box 6000Rockville, MD 20849–6000800–638–8736301–519–5212 (Fax)E-Mail: askncjrs@ncjrs.org

Related Readings

In addition to the Youth Gang Bulletin series, other gang-related publications,sponsored by OJJDP and other Office of Justice Programs agencies, are availablefrom the Juvenile Justice Clearinghouse (JJC). These publications include:

Youth Gangs: An Overview. NCJ 167249.

1995 National Youth Gang Survey (Program Summary). NCJ 164728.

Addressing Community Gang Problems: A Model for Problem Solving(Monograph). NCJ 156059.

A Comprehensive Response to America’s Youth Gang Problem (Fact Sheet).FS 009640.

Gang Members and Delinquent Behavior (Bulletin). NCJ 165154.

Gang Suppression and Intervention: Community Models (Research Summary).NCJ 148202.

Gang Suppression and Intervention: Problem and Response (ResearchSummary). NCJ 149629.

Highlights of the 1995 National Youth Gang Survey (Fact Sheet). FS 009763.

Prosecuting Gangs: A National Assessment (Research in Brief). NCJ 151785.

Street Gangs and Drug Sales in Two Suburban Cities (Research in Brief).NCJ 155185.

Urban Street Gang Enforcement (Monograph). NCJ 161845.

Youth Gangs (Fact Sheet). FS 009772.

For copies of these publications, contact JJC at 800–638–8736 or send yourrequest via e-mail to puborder@ncjrs.org. These documents are also availableonline. Visit the Publications section of OJJDP’s Web site, www.ncjrs.org/ojjhome.htm.

OJJDP’s National YouthGang Center

As part of its comprehensive, coordi-nated response to America’s gangproblem, OJJDP funds the NationalYouth Gang Center (NYGC). NYGCassists State and local jurisdictionsin the collection, analysis, and ex-change of information on gang-related demographics, legislation,literature, research, and promisingprogram strategies. It also coordi-nates activities of the OJJDP GangConsortium—a group of Federalagencies, gang program representa-tives, and service providers thatworks to coordinate gang informa-tion and programs. For more infor-mation contact:

National Youth Gang CenterP.O. Box 12729Tallahassee, FL 32317850–385–0600Fax: 850–385–5356E-Mail: nygc@iir.comInternet: www.iir.com/nygc

Information newly available on theWeb site includes gang-related legis-lation by subject and by State and theYouth Gang Consortium Survey ofGang Problems.

NCJ 171119

U.S. Department of Justice

Office of Justice Programs

Office of Juvenile Justice and Delinquency Prevention

Washington, DC 20531

Official Business

Penalty for Private Use $300

Bulletin

BULK RATEU.S. POSTAGE PAID

DOJ/OJJDPPermit No. G–91