ebook img

Trajectories of desistance and continuity in antisocial behavior following court adjudication among PDF

24 Pages·2010·0.41 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Trajectories of desistance and continuity in antisocial behavior following court adjudication among

DevelopmentandPsychopathology22(2010),453–475 #CambridgeUniversityPress,2010 doi:10.1017/S0954579410000179 REGULAR ARTICLE Trajectories of desistance and continuity in antisocial behavior following court adjudication among serious adolescent offenders EDWARDP.MULVEY,aLAURENCESTEINBERG,bALEXR.PIQUERO,cMICHELLEBESANA,d JEFFREYFAGAN,eCAROLSCHUBERT,fANDELIZABETHCAUFFMANg aUniversityofPittsburgh;bTempleUniversity;cFloridaStateUniversity;dUniversityofPhilippinesVasayas;eColumbiaUniversity; fUniversityofPittsburghMedicalCenter;andgUniversityofCalifornia–Irvine Abstract Becausemanyseriousadolescentoffendersreducetheirantisocialbehavioraftercourtinvolvement,understandingthepatternsandmechanismsoftheprocessof desistancefromcriminalactivityisessentialfordevelopingeffectiveinterventionsandlegalpolicy.Thisstudyexaminedpatternsofself-reportedantisocialbehavior overa3-yearperiodaftercourtinvolvementinasampleof1,119seriousmaleadolescentoffenders.Usinggrowthmixturemodels,andincorporatingtimeat riskforoffendinginthecommunity,weidentifiedfivetrajectorygroups,includinga“persister”group(8.7%ofthesample)anda“desister”group(14.6%of thesample).Casecharacteristics(age,ethnicity,antisocialhistory,deviantpeers,acriminalfather,substanceuse,psychosocialmaturity)differentiatedthe fivetrajectorygroupswell,butdidnoteffectivelydifferentiatethepersistingfromdesistinggroup.Weshowthateventhemostseriousadolescentoffenders reportrelativelylowlevelsofantisocialactivityaftercourtinvolvement,butthatdistinguishingeffectivelybetweenhigh-frequencyoffenderswhodesistandthose whopersistrequiresfurtherconsiderationofpotentiallyimportantdynamicfactorsrelatedtothisprocess. There is broad recognition of the potential of longitudinal of aggressive behavior; Coie & Dodge, 1998; Piquero, Far- datato inform the studyof juvenile crime and delinquency. rington,&Blumstein,2003),andtheimportanceofparticular Overthelastfewdecades,researchersconcernedwiththede- eventsatdifferentagesforpromotingonsetormaintenanceof velopment of antisocial behavior have producedmanylarge antisocialactivity (e.g.,Kokko,Tremblay,Lacourse,Nagin, prospective studies worldwide (see Thornberry & Krohn, &Vitaro,2006;Patterson&Stouthamer-Loeber,1984). 2003)andnumeroussecondaryanalysisprojects(e.g.,Broidy Panelstudieshaveconsiderablepotentialforhelpingjuvenile et al., 2003; Sampson& Laub, 1993). The introduction and justiceandchildwelfareprofessionalsformulatemoreinformed refinementofnewmethodologicalandstatisticaltechniques, identificationofat-riskgroupsandmorefocusedpreventivein- particularly trajectory modeling (Muthe´n & Muthe´n, 2000; terventions (Mulvey & Woolard, 1997). Existing longitudinal Nagin,1999,2005;Piquero,2008),havefueledtheseefforts, researchisminimallyuseful,however,inprovidingaclearpic- allowingresearcherstodirectlyexaminegroup-basedpatterns tureoftheoffendingpatternsofadolescentswhoareintheju- ofantisocialbehaviorovertime.Theseeffortshaveclarified venilejusticesystem,especiallyseriousadolescentoffenders,an ourunderstandingofthecourseofparticularbehavioralpat- importantgroupfor thedevelopmentofcriminologicaltheory ternsoverdifferentperiodsofdevelopment(e.g.,thestability andjuvenilejusticepolicy(Laub&Sampson,2001).Ingeneral, extant studies have been directed mainly toward mapping out developmentalregularitiesconnectedwiththeonsetandmain- WearegratefulforthesupportofthisprojectbyfundsfromtheOfficeofJu- tenanceofantisocialbehavior,providingapictureofwhenand venileJusticeandDelinquencyPrevention(2007-MU-FX-0002),NationalIn- stitute of Justice (2008-IJ-CX-0023),JohnD.and CatherineT. MacArthur howparticularchildrenveeroffthepathofnormaldevelopment Foundation,WilliamT.GrantFoundation,RobertWoodJohnsonFoundation, to ones of high rates of antisocial activity (at least for an ex- WilliamPennFoundation,CenterforDiseaseControl,NationalInstituteon tended time period).More specifically,longitudinal studies of DrugAbuse(R01DA019697),PennsylvaniaCommissiononCrimeandDelin- antisocial behavior usually have followed cohorts of children quency,andtheArizonaGovernor’sJusticeCommission.Thecontentofthis andadolescentssampledfromschoolsorcommunities,some- articleissolelytheresponsibilityoftheauthorsanddoesnotnecessarilyrepre- senttheofficialviewsoftheseagencies.WealsothankDanielNaginand timesoversamplingthosewithhigh-riskstatustoprovideade- BobbyJonesfortheirconsultationandThomasLoughranforassistancewith quate numbers of both subjects whowill and will not display dataanalysis. problem behaviors. Even with oversampling, though, this ap- Address correspondence and reprint requests to: Edward P. Mulvey, proachusuallyyieldsonlyasmallnumberofadolescentswho WesternPsychiatricInstituteandClinic,UniversityofPittsburghSchoolof enduppenetratingdeeplyintothejuvenilejusticesystem(El- Medicine,3811O’HaraStreet,Pittsburgh,PA15213;E-mail:mulveyep@ upmc.edu. liott&Huizinga,1987).Accordingly,thereislimitedinforma- 453 454 E.P.Mulveyetal. tion about the specific developmental contexts andbehavioral Ourfocusin thisarticle isoncharacteristics ofoffenders characteristics that differentiate among adolescents whose of- thatpredictpatternsofoffendinganddesistance,andnoton fendingismostserious. the impact of various types of interventions on these same Wewereonlyabletoidentifysixprevioustrajectorystudies outcomes. In the present analyses, we estimate the extent, undertakenwithsamplesofadjudicatedoffenders.Threediffer- types,andmagnitudeofheterogeneityofsubsequentpatterns ent data sets have been used, and all subjects were followed ofantisocialbehavioramongseriousoffenders,andwhether throughportionsofadulthood.Threestudiesemploythesample thisheterogeneityismeaningfullyrelatedtoindividuals’de- ofBostonareadelinquentsstudiedbytheGluecks(Eggleston, velopmental histories and life circumstances at the time of Laub,&Sampson,2004;Laub,Nagin&Sampson,1998;Samp- court adjudication. Taking on the ambitious task of thor- son & Laub, 2003), two involve cohorts of California Youth oughly mapping out the desistance process also requires an Authorityparolees(Piqueroetal.,2001;Piquero,MacDonald, examination of the sanctioning and treatment experiences & Parker, 2002), and one uses an offenders’ index from the ofoffenderswhileinthesystemaswellastheirexperiences British Home Office (Francis, Soothill, & Fligelstone, 2004). inthecommunityincloselyspacedintervalswhileonproba- Althoughimportant,thesestudiesarelimitedinfourways:(a) tionorafterreleasefromsecureconfinement.Staticcharacter- theytendtoemployonlyofficialrecordsofoffending,(b)they isticsalonewillbeinsufficienttodifferentiatesubgroupswho are based on a single site, (c) they contain primarily White follow different patterns of offending behavior (Nagin & subjects(withtheexceptionoftheCaliforniaYouthAuthority Tremblay; 2005; Piquero, 2008; Sampson & Laub, 2005). parolees), and (d) they contain a limited numberof important Nonetheless, examining relations between offender charac- and relevant theoretical predictors that have been found to be teristicsandpatternsofoffendingarevaluableleadsforfram- associated with antisocial and criminal activity over the life ingfutureresearchonthedesistanceprocessandthedevelop- course. ment of possible screening tools for disposition decision Information about patterns of change over time in serious making and risk management. The present analysis takes offenders,ifavailable,wouldbeofenormousvalue,because thefirststepinaddressingthisbroadergoal. itisthestartingpointformappingouttheprocessofdesistance Inthepresentstudy,trajectoryanalysisapproachesareused frominvolvementinantisocialactivity(Ezell&Cohen,2005; todetermineifthereareidentifiablesubgroupsofseriousadoles- Laub&Sampson,2001).Longitudinalresearchhasrepeatedly centoffenderswithdifferentpatternsofself-reportedoffending documentedthatlessthanhalfofseriousadolescentoffenders during a 3-year follow-up period following adjudication. Al- likely will continue their adult criminal career into their 20s thoughtheuseofgroup-basedtrajectorymodelingamongcrim- (see Elliott, 1994; Piquero et al., 2001; Redding, 1997). De- inologistshasbeenbothwidelyemployedandcontroversial(Pi- scribingthepathwaysoutofinvolvementinantisocialactivity quero,2008),webelievethattheenterpriseofidentifyinggroups (andthejusticesystem)andidentifyingthekeyfactorsrelated ofindividualswithdifferentovertimepatternsofoffendingisa todesistanceconstitutemajorquestionsthathavereceivedonly usefulendeavor,bothheuristicallyandpractically.Clearly,solu- limitedattentionfromresearchers. tions obtained in any individual analysis (including this one) are dependent on the characteristics of the sample examined, the outcome measures used, and the length of the follow-up PurposeofthePresentStudy period(seeEgglestonetal.,2004;Piqueroetal.,2001;Roeder, Thepresentarticledrawsondatafromanongoing,large-scale, Lynch,&Nagin,1999).However,trajectorysolutionscaniden- prospectivestudyofacohortofseriousjuvenileoffenders,the tify individuals within meaningful populations, such as the Pathways to Desistance study (see Mulvey et al., 2004). The sample for this study, who share common behavior patterns primaryaimoftheanalysespresentedhereistoprovideapor- overtime. traitoftheoffendingpatternsofagroupofseriousadolescent offenders in the period following court adjudication. It ad- FactorsPredictingSubgroupMembership dressesthebasicquestionofhowmuchandwhattypeofvaria- bilitymightexistinadolescentoffendersatthedeepestendof In addition to estimating distinct offending trajectories, our the juvenile justice system. Knowing the trajectories of these interestinthisarticleisintherelationbetweenpatternsofof- offenders in the critical period after their court involvement fendingandnumeroustheoreticallyrelevantfactorsthathave providesvaluableinformationforfocusingassessmentandin- beenassociatedwithcontinuedoffendinginpreviousinves- tervention strategies withthese individuals.We also examine tigations,butrarelystudiedwithinlargesamplesofjuveniles thepowerofbackgroundcharacteristicstodifferentiateserious whohavebeenadjudicatedofseriousoffenses.Theuniverse adolescentoffenderswhofollowdifferentbehavioralpathways ofpossiblevariablestoconsideraspredictorsofcontinuedof- throughthistimeperiod.Inshort,thisstudyprovidesaunique fendingislong,andnosinglestudycouldhopetobeexhaus- opportunitytoaddressthelimitationsofpreviouslongitudinal tiveinitscoverage.Followingacomprehensivereviewofthe studiesofseriousjuvenileoffenders:itcontainsacomprehen- mostcommonandimportantriskorprotectivefactorsassoci- sive, soundly measured, and generally complete set of both atedwithadolescentoffending(Loeber&Farrington,1998), predictorandoutcomemeasuresonasampleofseriousadoles- we examined a set of variables indicative of an individual’s centoffendersfromtwomajormetropolitanareas. riskorpropensityforfutureoffending(“casecharacteristics”) Trajectoriesofdesistanceandcontinuityinantisocialbehavior 455 andasetofvariablesreflectingpotentiallycriminogenicso- (D’Amico,Edelen,Miles,&Morral,2008;Dembo,Wareham, cialcontexts(“socialcontextcharacteristics”).Thisapproach &Schmeidler,2007)andthatcrimeandsubstanceusefluctuate allowedustoexaminetherelativeinfluenceofbothpersonal togetherovertime(Sullivan&Hamilton,2007),suggestinga andenvironmentalcharacteristicssimultaneously,andtocon- reciprocalrelationshipbetweenthetwobehaviors. sidercontemporary theoretical modelsthat have been tested Affectivedisordersalsohaveademonstratedconnectionto underawiderangeofsamplingandmeasurementconditions. involvementindelinquentbehavior.Affectivestatessuchas Acentralquestionofthisinquiryiswhetherthefactorsasso- depressionoranxietycanoftenbemanifestedinadolescents ciated with patterns of adolescent offending in prior studies asanger,leadingtoincreasedinvolvementinviolentactsto- using population samples of adolescents are also predictive wardothers(Mattila,Parkkari,&Rimpela,2006).Inaddition, ofdesistancewithinasampleofseriousadolescentoffenders. disproportionately higher rates of affective disorders have beenobservedinsamplesofjuvenileoffenders(Abram,Tep- lin, McClelland, & Dulcan, 2003). Although less is known Casecharacteristics abouttherelationbetweendelinquencyandaffectivepsycho- Foursetsofindividualcharacteristicswereexamined:(a)crim- pathologyovertime,itseemsreasonabletoassumethatthere inalhistory,(b)substanceuseandmooddisorders,(c)attitudes isasubstantialrelationbetweenthetwo(McReynoldsetal., toward the law, and (d) psychosocial maturity. Much prior 2008). research on antisocial behavior in community and high-risk Thethirdsetofindividualpredictorsincludesindividuals’ samplesindicatesthatindividualswithahistoryofoffending, attitudes and values about antisocial behavior and beliefs substanceuse,andothermentalhealthproblems,cynicalatti- about the law, law authorities, and legal institutions; some- tudestowardthelegitimacyofthelaw,andpsychosocialimma- times referred to as “legal socialization” (Fagan & Tyler, turity,especiallyproblemsinself-regulation,areatgreatriskfor 2005;Sampson&Bartusch,1998;Tapp&Kohlberg,1971; continuedinvolvementinantisocialactivity.However,toour Tyler&Huo,2002).Attitudesabouttheacceptabilityofanti- knowledge,nostudieshaveexaminedthesefactorssimultane- socialactivitytoachievedesiredendsortheelevationofone’s ously,inconjunctionwiththecontextualvariableswedescribe owninterestsabovetheobligationsofthesocialcontracthave below,orinasampleofseriousoffendersoveralengthystudy longbeenconsideredacentralcomponentofthecharacterofa intervalthatspansdevelopmentaltransitionsfromadolescence repeat offender (Samenow, 1996). More specific attitudes intoadulthood. aboutthelegitimacyofthelegalsystemhavealsobeenlinked Thefirstsetoffactors,whichisthemostinfluentialfactorin to individuals’ inclination to follow a law-abiding lifestyle judges’dispositiondecisions,concernstheoffender’spriorhis- (Tyler,1990,1997)andtocooperatewithpoliceandotherle- tory of offending and arrest. A large literature (Campbell & galactors(Tyler&Fagan,2008).Ourownanalysesoflegal Schmidt,2000;Hoge,Andrews,&Leschied,1995;Horwitz& socializationwithinthecurrentsampleindicatesquiteclearly Wasserman, 1980; Matarazzo, Carrington, & Hiscott, 2001; thatoffendersdiffersignificantlyintheirviewsofthelawand Thomas & Cage, 1977), including our own analyses with legalsystempriortotheiradjudication,thatthesedifferences thissample(Cauffmanetal.,2007),indicatesthatdispositional arestableforasignificantperiodafteradjudication(Piquero, decisionsincontemporaryjuvenilecourtsarebasedmainlyon Fagan, Mulvey, Steinberg, & Odgers, 2005), and that the thetypeandseverityofthecurrentoffenseandtheindividual’s changesthatdooccurinthesebeliefsarerelatedlogicallyto prior record, even when the adolescents considered are re- perceptionsofdeterrenceandthecostsandbenefitsofcrime strictedtoseriousoffenders(thuslimitingtheheterogeneityof aswellastheriskofdetectionandpunishment (Fagan&Pi- thesetwovariables).Accordingly,inthepresentsetofanalyses quero,2007). we ask whether individuals with a history of relatively more Thefinalsetofpersonalfactorsincludesvariablesthatin- seriousantisocialbehavior,asindexedbothbyjuveniles’self- dexindividualdifferencesinaspectsofpsychosocialmaturity reportsofoffendingpriortotheircurrentadjudicationandtheir thathavebeenhypothesizedtoaffectantisocialbehavior.The officialpriorarrestrecord,aremorelikelytofollowatrajectory mostinfluentialtheoryoftheunderlyingindividualcausesof ofcontinuedoffendingthantheirpeers. individualdifferencesinantisocialbehaviorwithincriminol- Thesecondsetofindividualcasecharacteristicscharacter- ogy is Gottfredson and Hirschi’s (1990) General Theory of izes individuals’ use and abuse of alcohol and illicit drugs Crime.Accordingtothisview,theonlysignificantindividual (seeFagan,1990)andtheirmoodoranxietyproblems(Grisso, predictor of criminal activity is low self-control, defined as 2004).Agrowingbodyofevidenceindicatesahighdegreeof the “tendency to pursue short-term, immediate pleasure” to comorbidityofsubstanceuseproblems(i.e.,highorproblem- theneglectoflong-termconsequences(p.93).ForGottfred- atic levels of substance use or the presence of a diagnosable son and Hirschi (1990), self-control is, “for all intents and substance use disorder of dependence or abuse) and delin- purposes, the individual-level cause of crime” (emphasis in quency(i.e.,highratesofself-reportedcriminalbehaviororof- original,p.232)andisbelievedtostablypredictcriminalacts ficialarrest)inadolescence,withabouthalfofallseriousjuve- throughoutthelifecourse(Hirschi&Gottfredson,1995),apo- nile offenders having substance use problems(Grisso, 2004). sitionthathasreceivedextensivesupportinpriorresearch(Pratt Thereisalsoevidencethatsubstanceuseatoneageisahighly &Cullen,2000).Deficienciesinself-controlwithinasampleof consistentindicatorofcontinuedseriousoffendingatalaterage seriousoffendersshoulddistinguishindividualswhocontinue 456 E.P.Mulveyetal. offendingfromthosewhodesistfromantisocialbehaviorover ExtendingPriorResearch time(Cauffman,Steinberg,&Piquero,2005;Monahan,Stein- Theanalysespresentedheredifferfrommuchoftheprevious berg,Cauffman,&Mulvey,inpress).1Wealsoincludeotherin- workexaminingpatternsofself-reportedoffending.First,we dicatorsofpsychosocialmaturitythathavebeenlinkedtoado- studyasampleofseriousadolescentoffenders,ratherthana lescentantisocialbehavior,suchasresponsibility,perspective general community or high-risk sample. Because trajectory taking, and susceptibility to peer pressure (see Cauffman & solutions are groupings of individuals in a sample based on Steinberg, 2002). These constructs, indicating key perceptual relative patterns of observations over time, any trajectory abilitiesassociatedwithsuccessfultransitionfromadolescence groupsobtainedcanonlybeinterpretedfortheirapplicability to early adulthood roles, have been investigated in numerous against the backdrop of the sample used for the particular normativestudiesandoneslinkingtheseskillstoincreasedlike- analyses. The trajectories derived here reflect possible sub- lihood of involvement in riskyand antisocial behaviors (e.g., groupsofseriousadolescentoffenders,notalladolescentof- Steinbergetal.,2008,2009). fenders or adolescents in general. Second, the time period covered starts with a baseline interview conducted shortly Socialcontexts afteradjudicationandconsidersobservationsoverthesubse- quent3years;mostpreviousanalysesconsiderchangesover Theoriesofcontextualinfluencesonoffendinghavefocusedon differentages,ratherthanovertime.Thus,unliketheseprior family, peers, and community as interrelated factors that in- analyses,thetrajectorygroupspresentedhereindicatesocial crease the likelihood of involvement in antisocial behavior adaptationinrelationtotimesinceadjudication,ratherthanas (Chung&Steinberg,2006).Ithasbeenshownthatadolescents afunctionofchronologicalage. aremorelikelytoengageinantisocialbehaviorwhentheyare Bothoftheseaspectsofthedesignandanalysisaredelib- exposedtoharshorlaxparenting,whentheyaffiliatewithde- erate. The Pathways to Desistance Study is primarily con- linquentpeers,andwhentheygrowupinneighborhoodschar- cerned with providing data relevant for improving practice acterizedbypoverty,disorganization,andlow“collectiveeffi- and policy in the justice system. As a result, our interest is cacy” (Farrington, 2004; Leventhal & Brooks-Gunn, 2004; intheidentificationofgroupsofseriousadolescentoffenders Loeber & Farrington, 1998). Cross-sectional analyses of data who present the most difficult individual and policy conun- on histories of offending within the current sample indicate, drums,whodrivepolicyandlegislationwellinexcessoftheir infact,thatoffendersfromcontextscharacterizedbypoorpar- numbersinthesystem,andaboutwhomleastisknown:those enting, deviant peer groups, and neighborhood disadvantage who do not desist from offending, despite having been ar- aredisproportionatelyoverrepresentedamongthemostserious restedandconvictedofaseriousoffense(e.g.,Laub&Samp- of the serious offenders in the cohort (Chung & Steinberg, son,2001).Inaddition,dataareexaminedaboutthetimepe- 2006; Steinberg, Blatt-Eisengart, & Cauffman, 2006). What riod directly after court involvement. The pressing question wedonotknow,however,iswhetherthesecontextualfactors for professionals in the juvenile and adult justice systems is predictdistinctpatternsofoffendingovertimeandcontinued to determine what distinguishes those who get out of the involvement in offending behavior at a high rate over time deependofthesystemoncetheyarethere(Jones,Harris,Fa- andacrossdevelopmentalstages. der,&Grubstein,2001).Lookingaheadfromapointofcourt Knowledgeofwhichindividualcasecharacteristicsandcon- involvementtoseewhatfactorsmightberelatedtodesistance textualfactorsprospectivelypredictpatternsofdesistanceinse- canpromotereasonedandinformedinterventions.Afirsttask riousadolescentoffendersandthestrengthoftheseeffectsare inthislargereffortistodetermineiftherearedistinctgroups valuableforboththeoreticalandpracticalreasons.Information ofseriousadolescentoffenderswhofollowdifferentpathsof about the factors most related to different patterns of involve- offendingaftertheyaredetectedbythecourt,andifthereare ment in antisocial activities provides leads about the possible distinctcharacteristicsoftheadolescentswhoprogressalong mechanismsbehindthesepatternsofbehaviors,withinthesam- eachofthesepathways. pleasawholeandinidentifiedsubgroups(cf.Haviland&Na- gin,2005).Onamorepracticalnote,theseresultsprovidethe empiricalbaseneededtoconstructstructuredheuristicsorjudg- Methods mentsystemsforclassifyingoffenders.Inshort,giventhelack oflong-termdataonseriousadolescentoffendersgenerally,and Participants thefactorsassociatedwithpersistence/desistanceamongsucha Thepathwaysstudyenrolled1,354adjudicatedadolescents2 groupspecifically(Laub&Sampson,2001),thisstudycanpro- who were at least 14 and below 18 years of age at the time videfertilegroundforspurringtheoreticalandpolicy-relevant of the offense precipitating a court petition to the juvenile discussionsaboutthecourse,nature,anddecisionsregardingse- riousadolescentoffendingandoffenders. 2. Previousarticlesfromthisstudyreportedasamplesizeof1,355.Sincethe publicationofthesearticles,1enrolledadolescentrevealedinalaterinter- 1. Itisimportanttonotethat,evenwithinapopulationofoffenders,there viewthatheliedabouthisageduringtheinterviewsandintheofficial willbesufficientvariabilityinself-controltoassessitsrelationtoantiso- records.Givenhiscorrectage,hedidnotmeeteligibilityrequirements cialandcriminalactivity(Hirschi&Gottfredson,2000). andthiscasewasremovedfromthesample. Trajectoriesofdesistanceandcontinuityinantisocialbehavior 457 or adult court systems in Philadelphia, Pennsylvania, and proached (either the adolescent or the parent) declining to Phoenix, Arizona, and found guilty (adjudicated) of one of participate. Some small, but statistically significant, differ- alistofseriousoffenses.Eligiblecrimesincludedallfelony enceswerefoundbetweentheadolescentswhowereadjudi- offenses with the exception of less serious property crimes, cated of eligible crimes, but not enrolled, and those adoles- aswellasmisdemeanorweaponsoffensesandmisdemeanor centsenrolledinthestudy(thesearereportedinmoredetail sexualassault.3Becausedruglawviolationsrepresentsucha inSchubertetal.,2004).Theenrolledgroupwasyoungerat significantproportionoftheoffensescommittedbythisage theiradjudicationhearing(15.9vs.16.1yearsold;t¼–4.42, group, and because males account for the vast majority of p,.001),hadmorepriorpetitionstocourt (meanof2.1vs. thosecases(Stahl,2003),wewereconcernedaboutcompro- 1.5;t¼8.78,p,.001),andappearedinthecourtforthefirst misingtheheterogeneityofthesampleifwedidnotlimitthe timeatanearlierage(13.9vs.14.2yearsold;t¼–3.29,p¼, number of study participants who were drug offenders. .001). We did enroll proportionately more White offenders Therefore,theproportionofjuvenilemaleswithdrugoffenses (test of proportions z ¼ 3.27, p , .005) and fewer African wascappedatapproximately15%ofthesampleateachsite. Americans (test of proportions z ¼ 3.09, p , .005), most All females meeting the age and adjudicated crime require- likelybecauseoftheimpositionofacapontheproportionof mentsandallyouthswhosecaseswerebeingconsideredfor thesampleadjudicatedondrugcharges.Overall,thesedatain- trialintheadultsystemwereeligibleforenrollment,evenifthe dicatethattheenrolledadolescentswerenoless,andpossibly chargedcrimewasadrugoffense.Formostoftheseadoles- more,seriousoffendersthanthosenotenrolled. cents,themostseriousadjudicatedchargethatqualifiedthem Abaselineinterviewwasconductedwithin75daysofad- forenrollmentinthestudywasaseriouscrimeagainstperson judicationforenrolledyouthsinthejuvenilesystemand,for (e.g., armed robbery, felony assault), and this was not their thosereferredtotheadultsystem,within90daysoftheirlegal firstappearanceincourt.Furtherdetailsabouttheenrollment certificationasadults(astheresultofadecertificationhearing process and sample characteristics are available in Schubert inPennsylvaniaoranadultarraignmenthearinginMaricopa etal.(2004). County).Inmostcases(62%),thebaselineinterviewoccurred We restricted ourcurrent analysesto male adolescent of- afterthedispositionhearing.Wethenconductedafollow-up fenders who had at least three completed follow-up inter- interview(“time-point”interview)every6monthsfor3years views out of the possible six (n ¼ 1,119). The ethnicity of andthenannuallythereafter.Allstudyparticipantshavecom- thisselectedsampleis19.6%White,41.1%AfricanAmeri- pletedatleast3yearsoffollow-upinterviews,butinterviews can,34.7%Hispanic,and4.6%other.Theparticipants’aver- beyondthatpointarestillongoing.Theanalysesreportedhere ageagewas16.0years(SD¼1.2years)atthetimeofthein- usedatafromthebaselineinterviewandthesixfollow-upin- itial interview. Table 1 presents the characteristics of the terviews completed over the first 3 years after the baseline entire enrolled group, each gender group, and the selected interview. subsample. Thecomputer-assistedinterviewassessesstatusandchange There is limited research on longitudinal patterns of fe- across multiple domains such as individual functioning, psy- maleoffending(Piquero,Brame,&Moffitt,2005),butexist- chosocial development and attitudes, family and community ingstudiesindicatethatpatternsandtypesoffemaleoffend- context,andrelationships.Acombinationofstructuredandinter- ing over time are likely to differ substantially from male viewer-ratedinstrumentswereused.4Onaverage,follow-upin- offending(Moretti,Odgers&Jackson,2004).Unfortunately, terviewstook2hrtocomplete.Participantswerepaidfortheir wehadonlyamarginallysufficientnumberoffemalesinthe participation. sample(N¼184)toobtainastabletrajectorymodelforthis Interviewswerecompletedattheparticipants’homes,institu- groupalone(cf.Nagin,2005).Ratherthanimposingstatisti- tionalplacement,orinapublicplacesuchasalibrary.Attempts calcontrolsforthisvariableinanoverallmodel,weoptedto weremadetoprovideaprivatesettingortoconducttheinterview consideronlymalesinthisanalysis,astrategythatallowsfor outofthehearingrangeofotherswithineachoftheselocations. directcomparisonswithexistingstudies.Analysesofthepat- Trainedinterviewersreadeachitemaloudandrespondentsgen- terns of female offending in this sample and in others are erallyansweredaloud.However,insituationsorinsectionsof importantdirectionsforfutureresearch. theinterviewwhereprivacywasaconcern,aportablekeypad wasprovidedasanoptiontoobtainanonverbalresponse. Retentionofparticipantsinthestudywasveryhighduring Procedures the interval for this analysis. Overall, 3% of participants Potential participants were identified from a daily review of dropped out of the studyand 3% died during the 36-month court record information in each site. Adolescents and their follow-upperiod.Onaverage,wecompletedover90%ofthe parents(oraparticipantadvocateinsituationswhereparental expectedinterviewsateachtimepoint.Asaresult,atthe3- or guardian contact was unobtainable) provided informed yearpoint,77%oftheparticipantshadnomissedinterviews consent to participate in the study, with 20% of those ap- 4. Acompletelistoftheconstructsthatwereassessedandtheinstruments 3. Thelistofchargesusedforenrollmentisavailablefromthecorresponding thatwereusedcanbeobtainedfromthecorrespondingauthoruponre- author. quest. 458 E.P.Mulveyetal. Table 1.Sample characteristics Characteristic EntireSample Males Females Sample N 1,354 1,170 184 1,119 Meanageatstudyindexpetition 16.24(1.10) 16.25(1.11) 16.16(1.02) 16.23(1.11) Meannumberofpriorpetitionsa 1.92(2.14) 2.06(2.21) 1.04(1.41) 2.03(2.20) Meanageatfirstpriorpetition 14.93(1.64) 14.85(1.67) 15.40(1.33) 14.86(1.66) Race/ethnicity Caucasian/White 0.20 0.19 0.27 0.20 AfricanAmerican/Black 0.41 0.42 0.37 0.41 Hispanic 0.34 0.34 0.30 0.34 Other 0.05 0.05 0.06 0.05 Mostseriousoffenseb Crimeagainstperson 0.44 0.45 0.41 0.45 Propertycrime 0.25 0.27 0.14 0.27 Drugoffense 0.16 0.13 0.31 0.13 Weaponsoffense 0.10 0.10 0.09 0.10 Other 0.04 0.04 0.04 0.04 Missingdata 0.01 0.01 0.01 0.01 Note:Thevaluesinparenthesesarestandarddeviations. aAveragecountofallpriorpetitionsavailableinthesubject’scourtrecordsexcludingprobationviolations. bMostseriouschargeonstudyindexpetition. (theyhaveabaselineandsixtime-pointinterviews)and17% manner as the general variety score, except using a limited havefourorfiveofthesixpossibletime-pointinterviews. numberofbehaviorallyhomogeneousactivities.Thatis,each ofthesescoresisthesumoftheendorsedactsthatareeitherag- gressiveoffenses(foraggressiveoffendingvarietyscore)orin- Measures come offenses (for income offending variety score). These Self-reported offending. We used a modified version of the scoresarenotusedasoutcomemeasures,butarecomponents Self-ReportofOffending(SRO;Elliott,1990;Huizinga,Es- ofacompositescoreregardingprioroffendingthatisusedas bensen, & Weihar, 1991) at each interview to measure the apredictorvariableinlateranalyses. adolescent’s account of his/her involvement in antisocial and illegal activities. The scale used here is composed of Casecharacteristics. 22itemslistingdifferentseriousillegalactivities(thespecific items arepresentedlater).The subject indicateswhether he/ Demographics. Research participants provided their date she has done any of these activities “ever” or over the“last of birth, ethnicity, and parental involvement in crime (i.e., 6months”(bothtimeframeswereusedatthebaselineinterview, whether father was ever arrested or jailed, whether mother butonlythe“last6months”timeframewasusedduringeach waseverarrestedorjailed)duringtheinitialinterview.Re- follow-upinterview).Asumofthenumberofitemsendorsed search participants completed the Wechsler Abbreviated (a “general variety” score ranging from 0 to 22) is calculated ScaleofIntelligence(Wechsler,1999),whichproducesanes- foreachsubjectateachtimepoint. timateofgeneralintellectualability(IQ)basedontwosub- The SRO measure used in the Pathways to Desistance tests: one for vocabulary and one for matrix reasoning. The Studyisaversion ofthe mostcommonly usedself-reported Wechsler Abbreviated Scale of Intelligence performance is delinquencymeasureacrosslongitudinalstudiesofantisocial correlatedwithboththeWechslerIntelligenceScaleforChil- behavior. Repeated analyses have demonstrated the scale’s dren and the Wechsler Adult Intelligence Scale, and it has validity,reliability,associationwith predictor variables, and beennormedforindividualsaged6to89years. correlation with official measures of offending among both general- and offender-based samples in childhood, adoles- Prior history of offending and arrest. Juvenile court rec- cence,andadulthood(seeThornberry&Krohn,2000).Pre- ords were coded regarding prior involvement with the legal viousresearchhasshownthatavarietyscoreprovidesacon- systemforcriminaloffenses.Thetotalnumberofpriorpetitions sistent and valid estimate of overall involvement in illegal tocourtandtheageatfirstpetitionweretakenfromtheseoffi- activityoveragivenrecallperiod(Osgood,McMorris,&Po- cialsources.Toreducepotentialproblemsfrommulticollinear- tenza,2002;Thornberry&Krohn,2000). ity,weconstructedasinglemeasurefortheconstructofantiso- Twosubcategoriesoftheoffendingvarietyscorearealso cialhistoryfrommultipleindicators(ageatfirstarrest,number computed: aggressive offending variety (e.g., “Been in a ofpriorpetitionsinpastyear,levelofself-reportedincomegen- fight?”) and income offending variety (e.g., “Used checks eratingoffensesever,levelofself-reportedaggressiveoffenses orcredit cards illegally?”). These are calculated inthesame ever; see Mulvey, Schubert, & Chung, 2007). To derive the Trajectoriesofdesistanceandcontinuityinantisocialbehavior 459 compositemeasure,weperformedaconfirmatoryfactoranaly- holdependence,drugabuse,ordrugdependence;(b)significant sesusingthefullsampleofPathwaysstudysubjects,including lifetime social consequences from alcohol use; (c) significant the variables of age at first petition to court, number of prior lifetimesocialconsequencesfromdruguse;and(d)significant courtpetitionsinthepastyear,SROaggressiveoffendingvari- dependencesymptomsfrombothalcoholanddruguse.Thefirst etyscore(ever),andtheSROincomeoffendingvarietyscore variable was based on the CIDI interview and the remaining (ever).Thiscomposite measure fit the datawell (comparative three from the Substance Use/Abuse Inventory. Using items fitindex[CFI]¼0.99;rootmeansquareerrorofapproximation fromtheSubstanceUse/AbuseInventory,wealsoconstructed [RMSEA]¼0.04). measuresoftheamountofsubstanceuseoverthe6monthsprior Mood/anxietyandsubstanceuseproblemswereassessed totheinitialinterview.Variablesreflectingtheoveralllevelof usingtheCompositeInternationalDiagnosticInterview(CIDI), alcoholuseandtheoverallsubstanceusefrequencywerecalcu- a highly structured clinical interview based on DSM-IV and latedfromtheself-reportsofsubstanceuse. ICD-10diagnosticcriteria(Kessler&Ustu¨n,2004).TheCIDI isacomputerizedassessmenttooladministeredbynonclinical Attitudestowardthelegalsystem.Forthisstudy,wedevel- interviewers(Kessleret al., 2004) that has good concordance oped measures of two central constructs regarding percep- withotherclinician-baseddiagnosticinstruments(First,Spitzer, tions of the legal process, or legal socialization. Following Gibbon, & Williams, 2002).The present studyonlyobtained Sampson and Bartusch (1998), we modified Srole’s (1956) diagnosticinformationonmajordepressivedisorder,dysthymia, legal anomie scale to create a measure of legal cynicism manic episode, posttraumatic stress disorder, alcohol abuse, that assesses general values about the normative basis of alcohol dependency, drug abuse, and drug dependency in lawandsocialnorms.Theitemsassesswhetherlawsorrules the previousyear. All items were coded as 0 (no diagnosis) arenotconsideredbindingintheexistential,presentlivesof or1(diagnosis). respondents (Sampson & Bartusch, 1998). Respondents are ResearchparticipantsalsocompletedtheRevisedChildren’s askedtoreporttheirlevelofagreementwithfivestatements, ManifestAnxietyScale(Reynolds&Richmond,1985),a37- suchas“lawsaremadetobebroken”and“therearenoright item,self-reportinstrumentdesignedtoassessthelevelandna- orwrongwaystomakemoney.”Themeasureiscomputedas ture of anxiety. The subject is asked to endorse or deny each the mean of the five items, and it fit the baseline data ade- statement as descriptive of his/her feelings or actions. A total quately(a¼0.60;CFI¼0.99;RMSEA¼0.03). anxietyscoreiscomputedbasedonthenumberofpositiveen- WealsoadaptedTyler’s(1990,1997)measureoflegitimacy dorsements from among 28 items, exclusive of 9 items that oflawandlegalactors.Itemsmeasuredrespondent’spercep- comprisealiesubscale(a¼0.87forthebaselinedata). tion of fairness and equity of legal actors in their contacts Inaddition,weassessedthequantityandfrequencyofal- withcitizens,includingbothpolicecontactsandcourtprocess- coholanddruguseattheinitialinterviewusingitemsadapted ing(Tyler,1997;Tyler&Huo,2002;Tyler&Lind,1992).Re- fromtheAlcoholandHealthStudyattheUniversityofMis- spondentsindicatetheiragreementwith11statementssuchas souri(Chassin,Rogosch,&Barrera,1991;Sher,1981).This “overall,thepolicearehonest,”and“thebasicrightsofcitizens measure considers the adolescent’s use of illegal drugs and are protectedbythe courts.” The measureiscomputedasthe alcohol over the course of his/her lifetime and in the past 6 meanforthe11itemsandthisscorefitthebaselinedataade- months.Theself-reportmeasureiscomprisedofthefollowing quately(a¼0.80;CFI¼0.92;RMSEA¼0.07). subscales:substanceuse(e.g.,“Howoftenhaveyouhadalcohol Inaddition,weusedtheMechanismsofMoralDisengage- todrink?”)andsocialconsequences,dependency,andtreatment ment(Bandura,Barbaranelli,Caprara,&Pastorelli,1996)to (e.g.,“Haveyoueverhadproblemsorargumentswithfamilyor assessattitudesconcerningthetreatmentofothers,akeycompo- friendsbeforebecauseofyouralcoholordruguse?”,“Haveyou nentofassessingthelegitimacyoflegalsanctions.Theself- everwantedadrinkordrugssobadlythatyoucouldnotthinkof reportmeasurecontains32itemsthattapavarietyofjustifica- aboutanythingelse?”). tionsformistreatingothers(e.g.,“Itisalrighttobeatsomeone Inpriorwork(Mulveyetal.,2007),wedevelopedandre- whobadmouthsyourfamily.”).Followingtherecommenda- portedthepsychometricpropertiesofcompositescoresderived tionsoftheauthorsofthescale,anoverallmeanofthe32items from these measures to indicate the presence or absence of wasusedasageneralmoraldisengagementscore.Thisoverall mood/anxiety problems and substance use problems, and we scoreshowedgoodinternalconsistencyatbaseline(a¼0.88). usedthesesameindicatorshere.Mood/anxietyproblemswere ratedusingfourdichotomousyes/novariables:(a)past-yeardi- Psychosocialmaturity.SteinbergandCauffman’s(1996) agnosisofmajordepressivedisorder,dysthymia,oramanicepi- model of psychosocial maturity consists of three elements: sode;(b)past(ever)impairmentfromdepressivesymptomatol- temperance, perspective, and responsibility; each of these ogy;(c)past (ever)diagnosisofposttraumaticstressdisorder; hastwocomponents. Inthepresentstudy,weexamine each and (4) significant anxietyproblems. The first three variables ofthesesixcomponentsindependently.Specifically,fortem- werebasedonitemsfromtheCIDI,andthelastwasbasedon perance,weexamineimpulsecontrolandsuppressionofag- theRevisedChildren’sManifestAnxietyScale.Thecomposite gression;forperspective,weexamineconsiderationofothers forsubstanceuseproblemswascomposedoffourdichotomous andfutureorientation;andforresponsibility,weexamineper- yes/noindicators:(a)past-yeardiagnosisofalcoholabuse,alco- sonal responsibility and resistance to peer influence. Four 460 E.P.Mulveyetal. measureswereusedtocreatethesesixindices:theWeinber- mostcloselyreflectstheirbehavior.Next,theparticipantisasked ger Adjustment Inventory (WAI; Weinberger & Schwartz, toratethedegreetowhichthestatementisaccurate(i.e.,“sortof 1990), which includes subscales assessing index impulse true”or“reallytrue”).Eachitemisthenscoredona4-pointscale, control,suppressionofaggression,andconsiderationofoth- rangingfromreallytrueforthecharacterizationindicatingless ers;thePsychosocialMaturityInventory(FormD;Greenber- resistancetoinfluence(1)toreallytrueforthecharacterization ger,Josselson,Knerr,&Knerr,1974),whichincludesascale indicatingmoreresistancetoinfluence(4),withanswersofsort that assesses personal responsibility; the Resistance to Peer oftrueassignedascoreof2(ifassociatedwiththelessresistant Influencemeasure(Steinberg&Monahan,2007);andtheFu- option)or3(ifassociatedwiththemoreresistantoption).Ten tureOutlookInventory(Cauffman&Woolard,1999),which suchitemsarepresentedtotheparticipant,eachexploringadif- wasusedtoderiveameasureoffutureorientation. ferent dimension of peer influence (e.g., going along with ThreesubscalesoftheWAIwereused:impulsecontrol(e.g., friends,sayingthingsonedoesnotreallybelieve),andoneresis- “Isaythefirstthingthatcomesintomymindwithoutthinking tancetopeerinfluencescoreiscomputedforthismeasureby enough about it”), suppression of aggression (e.g., “People averagingscoresonthe10items.Themeasureshowedexcellent whogetmeangrybetterwatchout”),andconsiderationofoth- reliabilityandadequatefittothebaselinedata(a¼0.73;NFI¼ ers(e.g.,“Doingthingstohelpotherpeopleismoreimportant 0.92;CFI¼0.94;RMSEA¼0.04). tomethan almost anything else”). Themeasure asks partici- Threecomposite measures(usedinprevious research,see pantstoassesshowaccuratelyaseriesofstatementsmatchtheir Cauffman&Steinberg,2000)wereconstructedfromthesein- ownbehaviorintheprevious6months(5-pointscale,falseto struments.Atemperancescaleiscomputedasthestandardized true).Eachsubscalewasfoundtohaveadequatereliability(as meanofthe15itemscomprisingtheimpulsecontrolandsup- indexedbyCronbacha)andgoodfittothebaselinedata(asin- pression of aggression subscales of the WAI. A perspective dicatedbyconfirmatoryfactoranalysis):impulsecontrol(eight scaleiscomputedasthemeanofthestandardizedFutureOut- items,a¼0.76;normativefitindex[NFI]¼0.95;CFI¼0.95; lookInventoryscoreandthe“considerationofothers”subscale RMSEA¼0.07),suppressionofaggression(sevenitems,a¼ fromtheWAI.Aresponsibilityscaleiscomputedasthemean 0.78;NFI¼0.96;CFI¼0.97;RMSEA¼0.06),andconsid- ofthestandardizedPsychosocialMaturityInventoryscoreand erationofothers(sevenitems,a¼0.73;NFI¼0.98;CFI¼ thestandardizedresistancetopeerinfluencescore. 0.99;RMSEA¼0.04). Table2summarizesthecompositemeasuresofindividual The Future Outlook Inventory is a 15-item measure that characteristics that were used in these analyses. Composite utilizesitemsfromtheLifeOrientationTask(Scheier&Car- measures were used to represent prior criminal behavior, ver,1985),theZimbardoTimePerspectiveScale(Zimbardo, mood/anxietyproblems,substanceuseproblems,andaspects 1990), andtheConsideration of Future Consequences Scale of psychosocial maturity (temperance, perspective, and re- (Strathman, Gleicher,Boninger, & Edwards, 1994). The in- sponsibility).Thesecompositesallfitthedatawellandrepre- ventory asks participants to rank the degree to which each sentconstructsofinterestinaparsimoniousmanner. statementreflectshowtheyusuallyact,onascaleof1(never true) to 4 (alwaystrue). A future orientation score is calcu- Socialcontextualcharacteristics. lated based on the mean of eight items from the scale (e.g., Neighborhooddisadvantage.Wecreatedasummaryscore “Iwillkeepworkingatdifficult,boringtasksifIknowthey oftheadolescent’sneighborhoodresourcesbasedoncensusdata willhelpmegetaheadlater”).Thescaleshowedgoodreliabil- foreachofthelocales(c.f.,Chung&Steinberg,2006).“Neigh- ityandanexcellentfittothebaselinedata(a¼0.68;NFI¼ borhood” was defined by 2000 census tract boundaries and 0.96;CFI¼0.97;RMSEA¼0.03). basedontheaddressyouthidentifiedastheirprimaryresidence The Psychosocial Maturity Inventory includes a 30-item atthetimeofthebaselineinterview.Themeasureofneighbor- subscale that assesses personal responsibility (e.g., “If some- hood disadvantage was derived using four indicators from thing more interesting comes along, I will usually stop any 2000Censusdata:percentageofhouseholdsbelowthepoverty workI’mdoing,”reversescored).Individualsrespondona4- line;percentageofhouseholdsreceivingpublicassistance;per- pointscale,fromstronglydisagreetostronglyagree;anoverall centage of unemployed residents; and percentage of residents personalresponsibilityscoreiscalculatedasthemeanacrossall withlessthanahighschooleducation(USBureauoftheCensus, 30items.Themeasureshowedexcellentreliabilityandanade- 2000).Thesefactorsarerobustpredictorsofneighborhoodrisk quatefittothebaselinedata(a¼0.89;NFI¼0.82,CFI¼0.87, thatareassociatedwithelevatedratesofbothjuvenileandadult andRMSEA¼0.04). crime(forareview,seeFagan,2008).Aprincipalcomponents The measure of resistance to peer influence (Steinberg & analysis,runseparatelyforPhiladelphiaandMaricopaCounty, Monahan,2007)assessesthedegreetowhichadolescentsactau- revealed one factor that accounted for 79% and 77%, respec- tonomouslyininteractionswiththeirpeergroup.Participantsare tively,of total explained variance. Factor scores for neighbor- firstpresentedwithtwoconflictingstatements(e.g.,“Somepeo- hooddisadvantagewereusedinthemainanalyses. plegoalongwiththeirfriendsjusttokeeptheirfriendshappy” and “Other people refuse to go along with what their friends Parenting.TheParentalMonitoringInventory(Steinberg, wanttodo,eventhoughtheyknowitwillmaketheirfriendsun- Lamborn,Dornbusch,&Darling,1992)wasadaptedforthis happy”)andarethenaskedtochoosethecharacterizationthat studytoassessparentingpracticesrelatedtosupervisionofthe Trajectoriesofdesistanceandcontinuityinantisocialbehavior 461 Table 2.Composite measures of constructs Construct Indicators Instrument Priorcriminalbehavior Ageatfirstarrest Courtrecord Numberofpriorcourtpetitions(pastyear) Courtrecord Aggressiveoffenses Self-reportoffending Incomeoffenses Self-reportoffending Mood/anxietyproblems Diagnosisofselectmooddisorder(pastyear) CIDI Impairmentfromdepressivesymptoms(ever) CIDI Diagnosisofposttraumaticstressdisorder(ever) CIDI Significantanxietyproblems RevisedChildren’sManifestAnxiety Scale Substanceuseproblems Diagnosisofsubstanceusedisorder CIDI Significantsocialconsequencesfromalcoholuse Substanceuse/abuseinventory Significantsocialconsequencesfromdruguse Substanceuse/abuseinventory Dependencesymptomsfromalcoholordruguse Substanceuse/abuseinventory Temperance Impulsecontrol WAI Suppressionofaggression WAI Perspective Futureoutlookscore FutureOutlookInventory Considerationofothers WAI Responsibility Psychosocialmaturity PsychosocialMaturityIndex Resistancetopeers Resistancetopeerinfluence Note:CIDI,CompositeInternationalDiagnosticInterview;WAI,WeinbergerAdjustmentInventory. adolescent.Preliminaryquestionsestablishthepresenceofa Lizotte,Krohn,Farnworth,&Jang,1994)toassessthedegree singleindividualwhoisprimarilyresponsiblefortheyouth. ofantisocialactivityamongtheadolescent’speers.Research Therespondent’sanswerstoseveralitemsabouttheircurrent participantsanswerquestionsaboutthelevelofinvolvement livingsituation,specificallywhethertheylivewiththeiden- oftheirfriendsinillegalactivities(e.g.,“Howmanyofyour tifiedcaretaker,establishestheskippatternfollowedinthepa- friends have sold drugs?”) and the amount of pressure that rentalmonitoringitems.Thescaleiscomposedofnineitems. theirfriendsexertonthemtobeinvolvedinillegalactivities Fiveitemsassessparentalknowledge,andareaskedevenifa (e.g., “How many of your friends have suggested that you youthdoesnotlivewiththepersonidentifiedastheirprimary shouldselldrugs?”).Twoscoresarederivedfromthismea- caretaker. If theyouth lives with the primary caretaker, four sure. The peer delinquency–behavior score is the mean of additional items are asked to assess parental monitoring of the12itemsregardingtheinvolvementofpeersinillegalac- the youth’s behavior (e.g., “How often do you have a set tivity. The peer delinquency–influence score isthe mean of timetobehomeonweekendnights?”).Aparentalknowledge the7itemsregardingwhetherpeerspressuretheadolescentto scoreiscalculatedasthemeanofthefiveitemsregardingpa- engage in illegal activities. Confirmatory factor analyses for rental knowledge and a parental monitoring score is calcu- each of these scales showed good fits to the baseline data latedasthemeanofthefouritemsaddressingthisaspectof (peer delinquency–behavior: a ¼ 0.92; NFI ¼ 0.93; CFI ¼ therelationship.Confirmatoryfactoranalyseswereconducted 0.94; RMSEA ¼ 0.09; peer delinquency–influence: a ¼ fittingatwo-factorsolutionwiththeabovesubscalescoresto 0.89;NFI¼0.95;CFI¼0.96;RMSEA¼0.07). thebaselinedata.Thissolution,allowingforonecorrelateder- rortermfitthedatawell(CFI¼0.92;RMSEA¼0.08). Results The Quality of Parental Relationships Inventory (Conger, Ge,Elder,Lorenz,&Simons,1994)wasadaptedforthisstudy Latentclassgrowthanalysiswasusedtoidentifygroupsthat toassesstheaffectivetoneoftheparental–adolescentrelation- follow distinctive patterns or pathways of self-reported anti- ship,askedseparatelywithregardtomotherandfather.Forty- social behaviorover time. Latent class growth analysis uses two items tap parental warmth (e.g., “How often does your a single outcome variable measured at multiple time points mother let you know she really cares about you?”). For this todefinealatentclassmodelinwhichthelatentclassescorre- study,weusedthescalefortheparentalwarmthofthemother, spondtodifferentgrowthcurveshapesfortheoutcomevari- whichshowedreasonablyacceptablefittothebaselinedata(a able.Theanalysisestimatesthedifferentgrowthcurveshapes ¼0.92;NFI¼0.95;CFI¼0.95;RMSEA¼0.08).Therewere and class probabilities for each group (Muthe´n & Muthe´n, toomanymissingvaluesfortheratingsoftheparentalwarmth 2000). We used the group-based trajectory modeling proce- ofthefathertoincludethisvariable. dure developed by Nagin and colleagues (Land, McCall, & Nagin, 1996; Land & Nagin, 1996; Nagin, 2005; Nagin & Peers.ThePeerDelinquentBehavioritemsareasubsetof Land,1993;Nagin&Tremblay,1999;Roeder,Lynch,&Na- 19questionsusedbytheRochesterYouthStudy(Thornberry, gin, 1999) to identifysubgroups of individuals who display 462 E.P.Mulveyetal. similar patterns of behavior over time. Analyses were done Table3. BIC and 2log (B ) of the models considered e 10 usingthePROCTRAJprogram,anSASprocedureforesti- matinggroup-basedtrajectorymodels.Becausewewereana- No.ofGroups BIC NullModel 2log (B ) e 10 lyzingcountdata(numberofactsendorsed)andmorezeros 1 214586.20 were present than would be expected in the purely Poisson 2 213008.40 1 3155.6 model, we used the zero-inflated Poisson model (Lambert, 3 212641.75 2 733.3 1992;Zorn,1998).Becausewewerenottestingnestedmodels, 4 212522.83 3 237.8 we followed the lead of D’Unger, Land, McCall, and Nagin 5 212461.63 4 122.4 (1998)andusedthechangeintheBayesianinformationcriter- 6 212417.87 5 87.52 7 212390.26 6 55.22 ion(BIC)tocomparethefitofdifferentmodels,accordingto theguidelinesprovidedbyJones,Nagin,andRoeder(2001). Note:BIC,Bayesianinformationcriterion;2log(B ),twicethelogarithmof e 10 Thelevelofmissingdatafortheself-reportmeasurewas theBayesfactor.Thenullmodelcolumnrepresentsthenumberofgroups low.Thetotalnumberofinterviewscompletedonthesample testedinthenullhypothesis.Here,2loge(B10)(cid:2)2(DBIC),whereDBICis (N¼6,365)represents95%ofthenumberofinterviewsthat theBICofthealternative(morecomplex)modelminustheBICofthenull (simpler)model.ThelogformoftheBayesfactorisinterpretedasthedegree couldhavepossiblybeencollectedifallsubjects(N¼1,119) ofevidencefavoringthealternativemodel. were successfully interviewed at each time point. Only 10 completedinterviewshadmissingdataonself-reportedanti- ofthesample,isalsoalowoffendinggroupbutwithaslightly social activity. Missing data were assumed to be missing at moremarkeddeclinethangroup1inthefirsttwofollow-uppe- randominthePROCTRAJprogram. riods. Group 3, which constitutes 17.6% of the sample, has The analysis examining the sample for trajectory groups moderate levels of offending across the 36-month period. took into account the effect of institutional confinement on Group 4, which represents about 14.65% of the sample, is a thesubject’slevelofoffending,becausethisfactorcansub- high declining group whose level of offending is relatively stantiallyaffectthederivedsolutioninsamplesofactiveof- highatthestartandsteadilydecreasesacrossthe36-monthpe- fenders (see Piquero et al., 2001). Exposure time or the riod(i.e.,“desisters”).Group5,whichformsabout8.7%ofthe amount of time the subject was free to engage in criminal sample,isahighoffendinggroupwhoselevelofoffendingis acts in the community was used as atime-varyingcovariate highatthestartandremainsrelativelyhighercomparedtoother intheanalysis.Thisvaluewasaproportionindicatingtheto- groupsacrossthe36-monthperiod(i.e.,“persisters”).Compar- taldaysduring the6-month recall period thatthe individual isonsofthesetwogroupsmaybeespeciallyinformativetoour was reported to be in the community (i.e., not in a detox/ understandingofthedesistanceprocess. drug treatment facility, psychiatric hospital, secure facility, We checkedthe appropriateness of thisfive-groupmodel, or residential treatment facility). This information was not usingfourdiagnosticstandardsrecommendedbyNagin(2005): available for the baseline observation, so these values were (a) the average posterior probability of group membership setuniformlyto1forthisstartingperiod(Nagin,2005). shouldbeatleast0.7forallthegroups,(b)theoddsofcorrect classificationisgreaterthan5forallgroups,(c)thereisarea- sonably close correspondence between a group’s estimated Patternsofself-reportedoffending probability of membership and the proportion of individuals andidentificationofsubgroups classifiedtothegrouponthebasisofmaximumposteriorprob- Mixturesofuptosevenlatentclasseswereconsidered.Atthein- ability assignment rule, and (d) the confidence interval for itialstagewherewedecidedonthenumberofclasses,theform groupmembershipprobabilityissufficientlynarrow.Table4 ofthepolynomialusedtocapturetheshapeofeachtrajectory presentsthefourdiagnosticsofmodelperformanceinthesam- group was cubic in time. Table 3 presents the values of BIC pleandsuggeststhatthecapacityofthemodeltoestimategroup and2log (B )forthesolutionswithdifferentnumberofgroups. membershipprobabilitiesandtosortcasesamongthegroupsis e 10 For these data the BIC continued to improve as more verygood.Foreachgroup,thereisaclosecorrespondencebe- groups were added, which is typical. The five-group trajec- tweentheestimatedprobabilityofmembershipandthepropor- tory solution was chosen as the overall best fitting model, tionassignedtothegrouponthebasisofamaximumposterior however, because the six- and seven-group trajectory solu- assignmentprobabilityrule.The95%confidenceintervalsare tionsdidnotaddsubstantiallytotheunderstandingofdiffer- also relatively narrow for each group, less than 0.06 plus or ent group patterns. The additional subgroups identified in minuspˆ .Theaverageposteriorprobabilityofgroupmember- j thesesolutionsweresmall(,5%ofthesample)anddidnotin- shipisconsiderablyabovethe0.7cutoffforeachgroup,andthe dicatetrajectoriesthatweredistinctinshapefromtheonesap- oddsofcorrectclassificationisalsoconsiderablyabove5forall pearinginthefive-groupsolution. thegroups. Figure1showsthefinal,five-grouptrajectorysolutionusing Anillustrativefigureshowsthedifferencesinthelevelsof exposure time as a time-varying covariate. Group 1, which self-reported antisocial activity among the groups. Figure 2 comprisesabout24.8%ofthesample,isalowoffendinggroup, showstheproportionofeachofthreeofthegroupsatdifferent withalowlevelofoffendingatbaselinethatapproacheszeroin levelsofself-reportedoffending(Groups1,3,and5)thaten- thefollow-upperiods.Group2,whichmakesupabout34.4% dorseseachoftheitemsincludedintheSROvarietyscoreat

Description:
refinement of new methodological and statistical techniques, particularly trajectory modeling (Muthén & Muthén, 2000;. Nagin, 1999, 2005; Piquero,
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.