nded in 1999 inWord下载.docx

nded in 1999 inWord下载.docx

《nded in 1999 inWord下载.docx》由会员分享，可在线阅读，更多相关《nded in 1999 inWord下载.docx（8页珍藏版）》请在冰豆网上搜索。

AdvancingRandomizedTrialsinCrimeandJustice

DavidWeisburd

HebrewUniversityandUniversityofMaryland

LorraineMazerolle

GriffithUniversity

and

AnthonyPetrosino

LearningInnovationsatWestEd

TheAcademyofExperimentalCriminology（AEC）wasfoundedin1999toadvancetheuseofrandomizedtrialsincrimeandjusticeresearchandrecognizecriminologistswhohavesuccessfullyledrandomizedfieldexperimentsincriminology.Thefounders,primarilycriminologistsactiveintheAmericanSocietyofCriminology,wantedtocreateavehicletofocusattentiononexperimentersandcreateaplatformtosupportandencourageexperimentalapproachesincriminologicalinquiry.Experimentalcriminologyhad,forsomeyears,occupiedasmall,yetsignificantnichewithincriminology.TheAECfoundersbelieved,however,thatadedicatedforumspecificallydevotedtoexperimentsandtheircontributionscouldsupportnewcomerstothefield,createsynergiesamongstthosealreadyconductingexperiments,helpincreasetheuptakeofexperimentalfindingsbypolicymakers,andgenerallyfacilitatediscoursearoundthedesign,implementation,managementandoutcomesfromexperimentalresearch.

InthefirstfiveyearsofthedevelopmentoftheAEC,weprimarilyfocusedonidentifyingandhonoringindividualswhohadmadesubstantivecontributionstocriminologyandcriminaljusticepracticethroughexperimentalresearchmethods.Thesescholars,whoareidentifiedbelow,areessentiallythe“founders”ofmodernexperimentalcriminology,andtheirworkcollectivelyrepresentstheimportantrolethatexperimentationplaysinourprofession.Since2003,theAECalsobegantorecognizeotherpersonswhohavemadesubstantialcontributionstotheadvancementofexperimentalcriminology,buthavenot,themselvesconductedrandomizedfieldtrials.ThesepeopleareelectedasHonoraryFellows.In2005,theAECFellowsvotedtocreateamembershipgroupopentoanypersoninterestedinexperimentalresearchincriminology.

Withthecreationofamembershipgroup,theAcademybegantodevelopasaformalorganization.Wenowpublishanewslettertwiceayear,whichcanbeviewedonourwebsiteathttp:

//www.crim.upenn.edu/aec;

theAECsponsorstheJournalofExperimentalCriminology,publishedbySpringerVerlagsince2005;

anditawardstwoprizeseachyeartorecognizeoutstandingscholarshipinexperimentalcriminology:

theJoanMcCordAwardandtheEarlyCareerExperimentalScholarAward（seebelowfordetails）.

WethinkthatadiscussionoftheAECmustbeginbyplacingexperimentalcriminologyincontext.Accordingly,webeginourarticlebydescribingwhyrandomizedexperimentsshouldbeimportantincrimeandjusticeresearchandprovideabriefhistoryofrandomizedexperimentsincriminology.WethendescribetheworkoftheAcademyanditsmembershipgroup,anditseffortstoadvanceexperimentalresearchincrimeandjustice.Wehopethatouressaywillencouragediscussionofexperimentalmethodsincriminology,andencourageinterestintheAcademyofExperimentalCriminology.

Whyarerandomizedtrialsimportant?

Arandomizedexperimentisatypeofresearchdesignthatusesrandomallocationtogainequivalencebetweenthesubjectsorunitsstudied.Randomizedexperimentsmaybeusedtotestboththeoreticalandsubstantivequestions.Incriminologyandcriminaljustice,however,randomizedexperimentsaremostcommoninevaluationsoftreatmentsandinterventionprograms.Whenusedoutsidethelaboratoryandinanactualfieldsettingsuchasaprisonorcommunity,theyarereferredtoasrandomizedfieldexperimentsorrandomizedfieldtrials.Thereisbroadagreementamongsocialandbehavioralscientiststhatrandomizedexperimentsprovidethebestmethodfordrawingcausalinferencesbetweentreatmentsandprogramsandtheiroutcomes.

Thekeytounderstandingthestrengthofexperimentalresearchdesignsisthe“internalvalidity”ofastudy.Aresearchdesigninwhichtheeffectsoftreatmentorinterventioncanbeclearlydistinguishedfromothereffectshashigh“internalvalidity.”Inrandomizedexperimentalstudies,internalvalidityisgainedthroughtheprocessofrandomallocationoftheunitsoftreatmentorinterventiontoexperimentalandcontrolorcomparisongroups.Theresultofrandomizationisthattheresearcherhasrandomizedallotherfactorsbesidestreatmentitself,andremovesanypotentialsystematicbiasthatcomesfromprovidingoneparticulartypeofsubjecttreatmentandanothertypeofsubjectthecontrolorcomparisoncondition.Althoughthegroupsarenotnecessarilythesameoneverycharacteristic—indeedsimplybychance,therearelikelytobedifferences—suchdifferencesareassumedtoberandomlydistributed,andthereforepartandparcelofthestochasticprocessestakenintoaccountinstatisticaltests.

Randomallocationthusallowstheresearchertoassumethattheonlylogicalexplanationforanysystematicdifferencesbetweenthetreatmentandcomparisongroupsareduetothetreatmentsorinterventionsapplied.Whenthestudyiscomplete,theresearchercanarguewithconfidencethatifadifferencehasbeenobservedbetweentreatmentandcomparisongroups,itislikelytheresultofthetreatmentitself（sincerandomizationhasisolatedthetreatmenteffectfromotherpossiblecauses）.Innon-randomizedstudies,itismuchmoredifficulttomakethisclaimbecauseofthedifficultyofcontrollingforbothmeasuredandunmeasuredfactorsorinfluences.Forthisreason,randomizedexperimentshaveoftenbeendescribedasthe“goldstandard”forevaluationresearch.

Thereisconsiderabledebatebothamongcriminologistsandacrossotherfields（e.g.,educationormedicine）astothetypeofbiasthatresultsfromrelyingonevidencefromnon-randomizedevaluationsasopposedtorandomizedexperimentstodeterminetreatmentorprogrameffects.Somescholarshavearguedthatthedifferencesbetweenwell-designednon-randomizedstudiesandrandomizedexperimentsarenotlarge（LipseyandWilson,1993）.However,areviewofalargenumberofcriminaljusticeevaluationssuggeststhatnon-randomizedstudiesarelikelytosystematicallyover-estimateprogramsuccess（Weisburd,LumandPetrosino,2001）.Mostexperimentsincriminologyinvolvetherandomassignmentofindividuals.Butwhenlargerentitiessuchasschools,policebeats,orprisonlivingunitsarerandomlyassigned,theyaresometimesreferredtoasclusterrandomizedorplace-basedtrials.

RandomizedExperimentsinCriminology

RonaldFisherisoftencreditedwiththefirstdiscussionofusingrandomizationtoensureequivalentgroupsinagriculturalresearch.Hisworkappearedinhis1935treatiseentitled“DesignofExperiments.”Thefirstreportinaprofessionaljournaloftheresultsofarandomizedexperimentinanyfieldwasprobablythetestofstreptomycinforpulmonarytuberculosisin1948.ThefirstrandomizedexperimentconductedincriminologyiscommonlybelievedtobetheCambridge-SomervilleYouthStudy（PowersandWitmer1951）.Inthatexperiment,investigatorsfirstmatchedindividualparticipants（youthsnominatedbyteachersorpoliceas“troubledkids”）oncertaincharacteristicsandthenrandomlyassignedonetotheinnovationgroupreceivingcounselingandtheothertoacontrolgroupreceivingnocounseling.Investigatorshavecontinuouslyreportedthatthecounselingprogram,despitethebestintentions,actuallyhurttheprogramparticipantsovertimewhencomparedtodoingnothingtothematall.AlthoughthefirstparticipantintheCambridge-Somervillestudywasrandomlyassignedin1937,thefirstreportofresultswasnotcompleteduntil1951.

Experimentswereusedselectivelyuntilthemid-1960s.AsOakley（1998）notes,randomizedexperimentsbecameanimportantresearchdesignduringthattimetoevaluatetheJohnsonAdministration’s“GreatSociety”programsintheU.S.Shearguesthatrandomizedexperimentssoonfelloutoffavor,notbecauseofanynewacceptanceofthemethodologicalcriticismsofexperiments,butbecausetheresultscontinuouslyreportednopositiveeffectfortheAdministration’ssocialpoliciesandprograms.

DavidFarrington,LloydOhlinandJamesQ.Wilson（1986）deservecreditforhelpingtoputrandomizedexperimentsontheradarscreenofcriminologistsinthemid-1980s.Theirinfluentialbookentitled“UnderstandingandControllingCrime,”recommendedtheuseofrandomizedexperimentswheneverpossibletotestjusticeinnovations.ThisbookhadramificationsfortheU.S.NationalInstituteofJustice,whichunderthedirectionofJamesChipsStewartsoonthereaftersupportedovertwodozensuchexperiments（Sherman1992）.In2003Petrosinoandhiscolleaguesreportedthattherewerefully267distinctcriminologicalexperimentspublishedoravailableinEnglish.

Althoughrandomizedexperimentsincriminaljusticearemorecommonnowcomparedtothe1980s,theycontinuetorepresentasmallpercentageofthetotalnumberofimpactoroutcomeevaluationsconductedinareasrelevanttocrimeandjusticeeachyear.IntheMarylandReport（Shermanetal.2003）,forexample,only16%ofthe657evaluationstudiesreviewedusedexperimentalmethods.Thereareanumberofpossibleexplanationsforwhyexperimentshavelaggedbehindotherresearchmethodsinthefieldofcriminology.Onereasonmaybesimplythattheoriginsofcriminologycanbetracedmostdirectlytodisciplinesinwhichexperimentalmethodsarenotcommon,likesociology.Butanumberofscholarshavearguedthatethicalorpracticalconcernsmakefieldexperimentationparticularlydifficultincrimeandjusticeresearch.Forexample,thereisconcernthattherandomallocationofcriminaljusticesanctio