Circle sentencing restorative justice incarceration recidivism Aboriginal overrepresentation Reoffending Indigenous
AIM To examine the relationship between Circle Sentencing (CS) and the likelihood of incarceration and recidivism.
METHOD Weusetwodatasets.ThefirstisanextractfromtheNewSouthWales(NSW)BureauofCrimeStatisticsandResearch’sReoffendingDatabase(ROD).ThesecondisanextractfromtheAboriginalServicesUnit’s(DepartmentofCommunitiesandJustice)internaldatabase.Thesedataallowustoidentify656courtappearancesfinalisedthroughCS,andover90,000appearancesfinalisedthroughTraditionalSentencing(TS)between1March2005and31August2018.Usingthesedata,wecompareoutcomesforoffendersparticipatinginCSandTS,aftercontrollingfordefendant-casecharacteristicsandtimefixedeffects.Wethendiscusstherole of selection bias in our estimates.
RESULTS Netofcontrolsandfixedeffects,offendersparticipatinginCSare9.3percentagepointslesslikelytoreceiveaprisonsentence.WhencomparedtotherateatwhichoffendersundergoingTSareincarcerated,thisequatestoarelativedecreaseof51.7percent.Amongoffendersnotsenttoprison,offendersundergoingCSare3.9percentagepointslesslikelytoreoffendwithin12months.Whencomparedtothe12monthreoffendingrateofoffendersundergoingTS,thisequatestoarelativedecreaseof9.6percent.Finally,amongoffendersthatdoreoffend,thoseundergoingCStakeanadditional55dayslongertoreoffendthantheirtraditionallysentencedcounterparts.Weare,unfortunately,unabletoaddressthepossibilitythatselectionbiasisdrivingour(associative,non-causal)estimates.Assuch,ourestimatesmustbeinterpretedwithcaution.
CONCLUSION CS is associated with lower levels of incarceration and recidivism.
CRIME AND JUSTICE BULLETIN
NUMBER 226 | APRIL 2020
1
This bulletin has been independently peer reviewed.
KEYWORDS
Circle Sentencing, incarceration and recidivism Steve Yeong and Elizabeth Moore
Suggestedcitation:Yeong,S.andMoore,E.,(2020).CircleSentencing,incarcerationandrecidivism(CrimeandJusticeBulletinNo.226).Sydney:NSWBureauofCrimeStatisticsandResearch.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 2
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
INTRODUCTIONOverthe2016-17financialyear,Aboriginaland/orTorresStraitIslandersconstituted2.8percentoftheAustralianpopulation(ABS,2016)and27.6percentoftheprisonpopulation(ProductivityCommission,2018).Overthissameperiod,governmentsaroundAustraliacollectivelyspentinexcessof$4billionontheprisonsystem,representingarealyear-on-yearincreaseof7.2percent(ProductivityCommission,2018).Giventheeconomic,financialandsocialcostsofAboriginaloverrepresentationincustody,evaluating programs aimed at reducing Aboriginal incarceration rates is crucial for decision makers.
InAustralia,RestorativeJustice(RJ)programsbecameanincreasinglypopularalternativetothetraditionalcriminaljusticeprocessinthelate1990s.RJprogramstypicallyinvolvebringingtheoffenderface-to-facewiththeirvictim(s)inordertorepairharm,restorerelationshipsandstrengthensocialbondswithinacommunity(Larsen,2014).Tothisend,thegeneralconsensusinbothAustralia(Larsen,2014)andinternationally(Latimer,Dowden,&Muise,2005)isthatRJprogramsarebeneficialforoffendersandvictims.VictimsinvolvedinRJprogramstypicallyreporthighlevelsofsatisfactionwiththeprocess,astheybelievetheyaretreatedinafairandrespectfulway(Latimeretal.,2005).Severalstudiesalsosuggestthatoffendersaremorelikelytotakeresponsibilityfortheiractionsandthusaremorelikelytocomplywiththeirsentencingconditions(Larsen,2014;Latimeretal.,2005;Shaplandetal.,2007;Strangetal.,2006).
Unfortunately,thereislittleevidencetosuggestthatRJprogramshaveanyimpactonreoffendingrateswhencomparedwiththebusiness-as-usualCriminalJusticeSystem(CJS)response.1ForinstanceinNSW,priorresearchindicatesthatyouthjusticeconferencing(usedtodivertyoungoffendersfromcourt)isnobetterthantheChildren’sCourtinreducingrecidivism(Smith&Weatherburn,2012),andForumSentencing(anRJinformedapproachtosentencingadultoffenders)isnobetterthantheLocalCourtinreducingrecidivism(Jones,2009;Poynton,2013).
Thereis,however,almostnoresearchinvestigatingtheimpactofsuchprogramsonanimportantsubset of the general population: Aboriginal Australians.2FollowingtherecommendationsoftheRoyalCommissionintoAboriginalDeathsinCustody,RJprogramshavebecomeincreasinglyavailableforAboriginaloffendersinAustralia(Marchetti&Daly,2004).RJprogramsdirectedtowardIndigenousAustralians generally aim to involve members of the local community in the sentencing process. This bulletinfocusesonthelargestRJinformedprogramforAboriginaloffendersinNSW,CircleSentencing.
Circle Sentencing in NSW
CircleSentencing(CS)hasbeeninoperationinNSWsince2002.3CSisanalternativesentencingoption,withthefullsentencingpowerofatraditionalcourt,forAboriginaloffendersthatmeetaspecificsetofconditions.TheideabehindCSistoincludethelocalAboriginalcommunityinthesentencingprocess.Inpractice,thistypicallyinvolvesthepresidingmagistrateworkingwithagroupofAboriginalelders,victims,respectedmembersofthecommunityandtheoffender’sfamilytodeterminetheappropriatesentence.
CSwasintroducedwitheightobjectives.Theseobjectives,outlinedintheCriminal Procedure Act 1986 (NSW),include:(1)toincludemembersofAboriginalcommunitiesinthesentencingprocess;(2)toincreasetheconfidenceofAboriginalcommunitiesinthesentencingprocess;(3)toreducebarriersbetweenAboriginalcommunitiesandthecourts;(4)toprovidemoreappropriatesentencingoptionsforAboriginaloffenders;(5)toprovideeffectivesupporttovictimsofoffencescommittedbyAboriginal
1 OnenotableexceptionisastudyconductedbyMcGrarrellandHipple(2007),whofindsomeevidenceofa(beneficial)relationshipbetweenaRJinterventionandreoffendinginIndiana.2 ArelatedbutdistinctbranchofresearchcomparestheeffectofRJprogramsforAboriginalandnon-Aboriginaloffenders.Forexample,Little,Stewart,andRyan(2018)comparetherecidivismratesofamatchedgroupofAboriginalandnon-Aboriginaloffenders.Littleetal.(2018)findthattheformerofthesegroupshadhigherratesofpost-conferencerecidivism,althoughthisgroupmayhavebeenatahigherriskofreoffendingirrespectiveoftheintervention.3 OtherAustralianstateandterritoriesrunsimilarprograms.Forexample,theKooricourtsinVictoria,theMurricourtsinQueenslandandtheNungacourts in South Australia.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 3
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
offenders;(6)toprovideforthegreaterparticipationofAboriginaloffendersandtheirvictimsinthesentencingprocess;(7)toincreasetheawarenessofAboriginaloffendersoftheconsequencesoftheiroffencesontheirvictimsandtheAboriginalcommunitiestowhichtheybelong;(8)toreducerecidivisminAboriginal communities.
TheNowraLocalCourtwasthefirstsitetointroduceCS.SincethenCShasexpandedtooperateinatotalof12LocalCourtsinNSW.4ThetimingandlocationofeachparticipatingcourtaredetailedinTable1andFigure1.
Table 1. Local Courts with Circle Sentencing by date of commencement
Local Court CS available from
Nowra February2002
Dubbo October2003
Brewarrina January2005
Lismore January2006
Bourke January2006
Kempsey January2006
Armidale April2006
Walgett July2006
Mt Druitt January2007
Nambucca April2009
Blacktown July2010
Moree October2010
Figure 1. Local Government Areas where Circle Sentencing is available
4 Forcontext,between2005and2019,149LocalCourtswereinoperationinNSW.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 4
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Selection criteria for Circle Sentencing in NSW
InordertoparticipateinCS,adefendantmustbe:
1. Aboriginal;
2. foundguiltyofanon-excludedoffence;5
3. appearingataparticipatingLocalCourt;
4. likelytoreceivearelevantsentence;6
5. agreetoparticipate;and
6. beassessedassuitablebythelocalAboriginalCommunityJusticeGroup(ACJG).7
Inordertobedeemedsuitable,anoffenderundergoesthefollowingprocess:
1. Thepresidingmagistratemustdecidetorefertheoffenderforasuitabilityassessment.Offendersnotreferred for assessment undergo Traditional Sentencing (TS).
2. Ifreferredforassessment,conditionalontheoffender’sconsent,thecourt’sProgramOfficer(PO)8 thenconvenesameetingoftheACJG.
3. TheACJGthenmeettoconducttheassessment.FactorsconsideredbytheACJGinclude:thedefendant’sconnectednesstothelocalcommunity;theimpactoftheoffenceonthecommunity;thenatureoftheoffence;andfinally,thebenefitsofthecircletotheoffender,victimandcommunity.
4. IftheACJGdeemtheoffendersuitable,thenthemagistratemakesaprogramparticipationorder.IftheACJGdeemsthedefendantunsuitable,thentheoffenderundergoesTS.
5. Afterbeingdeemedsuitable,thePOthenconvenestheCSgroupandthecircletakesplace.
A CS group is typically made up of: four Aboriginal elders (usually two men and two women) selected onthebasisoftheirexperiencewiththeoffender,victimand/ornatureoftheoffence;thepresidingmagistrate;thePO;apoliceprosecutor;theoffender;theoffender’slegalrepresentative;andfinally,thevictim and their support person. The presiding magistrate must approve all participants in the circle in orderforthecircletogoahead.Duringthecircle,participantssitinacircleanddiscuss:thebackgroundoftheoffender;theoffence;theimpactonthevictim;howsimilarcrimeshavebeenaffectingthecommunity;whatcanbedonetopreventfurtheroffending;andhowallofthiscanbeincorporatedintoasentencingplan.Whilethepresidingmagistrateretainsfinalsay,itisgenerallybymajorityrulethatmembers of the circle determine the penalty.
Prior research on Circle Sentencing in NSW
Priorqualitativeresearch(CulturalandIndigenousResearchCentre,2008;Daly&Proietti-Scifoni,2009;Potasetal.,2003)hasfoundCStobegenerallybeneficial.ThesestudiesreportthatCSreducesperceivedbarriersbetweenAboriginalpeopleandthecourts,increasestheoffender’sawarenessoftheconsequencesoftheiractions,increasesconfidenceinsentencingandresultsinmoreappropriatesentencingoutcomes.However,severallimitationsofCShavealsobeennotedbyresearchers,particularlyintheearlystagesofimplementation.Someoftheselimitationsinclude:inadequatedrugandalcoholsupportservicesinsomelocations;insufficientdatacollectedoninvolvementofvictims;andfinally,thecirclenotproceedingasplanned(e.g.,defendantsrefusingtolistentoorthefollowadviceoftheelders).TheonlyquantitativeevaluationofCSwasconductedbytheNSWBureauofCrimeStatisticsandResearch
5 Excludedoffencesinclude:assaultoccasioninggrievousbodilyharm;rapeandothersexualoffences;childpornographyoffences;offencesinvolvingtheuseofafirearm;andcertaindrugoffences.InterestedreadersaredirectedtoSection348oftheCriminalProcedureAct1986(NSW)forthecompletelistofexcludedoffences.6 Suchasasentenceofimprisonment,asuspendedsentence,anintensivecorrectionorder,homedetention,communityserviceorder,orgoodbehaviourbond.7 TheACJGisa(court-specific)groupofAboriginalpeopleappointedbytheresponsibleportfolioministerontherecommendationoftheProgramOfficer.8 TheProgramOfficerisaNSWDepartmentofCommunitiesandJusticeemployeeresponsiblefor,amongotherthings,coordinatingCSateachsite.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 5
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
(BOCSAR)in2008.Usingcourtoutcomedatafrom2002to2007,Fitzgerald(2008)employedamatchingstrategytoinvestigatetherelationshipbetweenCSandrecidivism.ShefoundnostatisticallysignificantdifferenceinrecidivismratesbetweenoffendersundergoingCSanda(matched)controlgroupwhoweresentenced in the usual way by the Local Court.
The current study
The objective of the current study is to provide a follow-up evaluation more than a decade later. Specifically,thecurrentstudyisconcernedwithansweringthefollowingthreeresearchquestions:
1. AreoffendersparticipatinginCSlesslikelytoreceiveasentenceofimprisonmentthanoffendersundergoing TS?
2. AreoffendersparticipatinginCSlesslikelytoreoffendthanoffendersundergoingTS?
3. DooffendersparticipatinginCStakelongertoreoffendthanoffendersundergoingTS?
METHOD
Data
Weusetwodatasetsinthepresentstudy.ThefirstdatasetisanextractfromtheNSWBOCSAR’sReoffendingDatabase(ROD).TheRODextractcontainsinformationrelatingtoallcriminalproceedingsfinalisedinaNSWLocalCourtbetween1March2005and31July2019.Foreachcourtappearance,weareabletoobserve:thecourthousewherethematterwasfinalised;thebailhearingdate;thedatethematterwasfinalised(i.e.,thedatethatthesentencewasformallyhandeddown);andtheprincipalpenaltyassociatedwiththefinalisation.Wearealsoabletoobserveeachdefendant’s:age;gender;SEIFApercentilerank;9numberofpriorcourtappearances(withaprovenoffence);priorsentencesofimprisonment;andfinally,whetherthedefendantwasgrantedbailattheirfirstbailhearingforthatmatter.Foreachchargewithinacourtappearance,weareabletoobserve:thedateofeachoffence;theANZSOCcodeassociatedwitheachoffence;10aswellasthepleato,andoutcomeof,eachcharge.
The second dataset is an extract from the Aboriginal Services Unit’s (ASU’s) internal database.11 The ASU databasecontainsinformationrelatingtoallparticipatingoffender’sfirstreferralandsubsequentcirclebetween1March2005and31August2018.12 The ASU database allows us to observe: the date that the offenderwasreferredforasuitabilityassessment;theoutcomeofthesuitabilityassessment;thedateofthecircle;thesentencingdate;andfinally,anindicatorforwhetherthecirclewascancelled(e.g.,becausetheoffenderreoffendedpriortothecircle).
TheASUdatabasecontainsinformationrelatingto976uniqueoffendersreferredforasuitabilityassessment.Wewereabletomatch972oftheseoffenderstoindividualsinRODusingtheirdateofbirth,firstandlastname.Ofthese972offenders,242(24.9%)wereeitherdeemedunsuitable,didn’tconsenttoCS,orhadtheircirclecancelled.13Hence,thisleftuswith730circles(and242referralsthatsubsequentlyresultedinTS)thatneededtobematchedtocourtappearancesinROD.Inordertomatchcircles/referrals(recordedintheASUdatabase)tocourtappearances(inROD),weemployedtheprocedureillustratedinFigure2.14
9 SEIFAscoresareameasureofsocioeconomicdisadvantagebasedonthedefendant’spostcodeofresidenceatthetimeoffinalisation.Higherscoresindicatelowerlevelsofdisadvantage.InterestedreadersaredirectedtoABS(2011a)formoreinformation.10 ANZSOCcodesareusedtogroupoffencesacrossAustralianandNewZealandjurisdictions.InterestedreadersaredirectedtoABS(2011b)formoreinformation.11 TheASUisabusinessunitwithintheNSWDepartmentofCommunitiesandJustice.12 Thatis,ifanoffenderhasmorethanonereferral/circle,onlyinformationrelatingtothefirstreferral/circleisrecorded.13 These three categories cannot be disaggregated using the ASU database.14 Thatis,becauseoffendersparticipatinginCSonlyshowuponceintheASUdatabasebut(typically)multipletimesinROD,weemploythefollowingfivestepprocedure.InthefirststepwedesignateacasetobefinalisedthroughCSifthesentencingdateisidenticalinRODandtheASUdatabase.Inthesecond
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 6
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Intotal,wewereabletomatch656ASUcirclestocourtappearancesinROD.15InordertoavoidinadvertentlydesignatingaCSappearanceasaTSappearance,weexcludeunmatchedindividuals(andalloftheirappearancesinROD)fromthesample.AsimilarissuerelatestothefactthattheASUdatabaseonlyallowsustoidentifyanoffender’sfirstreferral/circle.Onceagain,toavoidinadvertentlydesignatingaCSappearanceasaTSappearance,weexcludefromthesamplenon-CSappearancesforoffendersthathave participated in at least one circle.
Figure 2. Data matching process
ASU database: 976 offenders referred to a suitability assessment
Yes
No
No Yes
n=4
n=242 n=730
ASU = Aboriginal Services UnitCS = Circle SentencingROD = Reoffending Data Collection
Match to ROD
523 cases with an identical sentencing date in both the ASU database and ROD
OR52 cases where the sentencing date recorded in the ASU database was within 31 days of the sentencing date in ROD
OR32 cases where the bail hearing date in the ASU database was within 31 days of the bail hearing date in ROD
OR22 cases in the ASU database where the offence date was within 31 days of the principal offence date in ROD
OR27 cases where the CS date was within 31 days of the sentencing date in ROD
Match to ROD
31 cases with an identical sentencing date in both the ASU database and ROD
OR10 cases where the sentencing date recorded in the ASU database was within 31 days of the sentencing date in ROD
OR67 cases where the bail hearing date in the ASU database was within 31 days of the bail hearing date in ROD
OR41 cases in the ASU database where the offence date was within 31 days of the principal offence date in ROD
Undergo CS
n=972
Linked to ROD based on individual
identifier
No YesNo Yes n=74 n=656n=149n=93
InordertomakedefendantsundergoingTraditionalSentencing(TS)ascomparableaspossibletooffendersundergoingCS,weemployfivesamplerestrictionsbasedontheeligibilitycriteriadescribedunderthelegislation.First,welimitoursampletodefendantsfoundguiltyofanon-excludedoffence.Second,welimitoursampletooffendersthatidentifiedasAboriginaltothepolicewhencharged.16Third,
stepwedesignateacasetobefinalisedthroughCSifthesentencingdateintheASUdatabaseiswithinaplusorminus31dayintervalofthesentencingdateinROD.Third,wedesignateacasetobefinalisedthroughCSifthedatethecirclewasheldboth:differedfromthesentencingdate,andwaswithinaplusorminus31dayintervalofthesentencingdateinROD.Fourth,wedesignateacasetobefinalisedthroughCSifthebailhearingdateintheASUdatabasewaswithinaplusorminus31dayintervalofthebailhearingdateinROD.Andfinally,wedesignateacasetobefinalisedthroughCSiftheoffencedateintheASUdatabaseiswithinaplusorminus31dayintervalofthe(principal)offencedateinROD.Ifacasecannotbematchedinanyofthesesteps,wedesignatethecase as “unmatched” and exclude all court appearances related to the individual from our estimation sample.15 Asarobustnesscheck,reportedinTableA2oftheAppendix,welimittheestimationsampletothe523perfectlymatchedcasesandre-estimateourpreferredanalyticalspecification.Wefindnomeaningfuldeviationfromthemainresults.16 WhetherapersonidentifiesasAboriginalcanchangeovertime.InterestedreadersaredirectedtoBiddleandMarkham(2018)forfurtherinformationregardingthedynamicsofAboriginalself-identification.Changingself-identificationovertimedoesnot,however,poseanissueforouranalysisas,inourpreferredanalyticalspecification,welimitourcomparisontooffenderssentencedwithinthesamemonth-year.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 7
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
welimitoursampletooffendersreceivingapenaltythathasbeenimposedbyaCSgroup.17Fourth,welimitthesampletomonth-yearswithatleastonecircle(e.g.,iftherewerezerocirclesheldinJuly2005,thenweremoveallappearancesoccurringinJuly2005).18Andfinally,forappearancesfinalisedthroughTS,welimitthesampletoappearancesfinalisedincourtswithoutCSavailableatthetimeoffinalisation.This is to ensure that defendants (implicitly) deemed unsuitable for CS by the presiding magistrate are not used as a control for CS participants.19
Descriptive statistics
Webeginourinvestigationbyexaminingtheprincipal(i.e.,themostserious)offencecommittedbyoffendersinoursample.Theseoffencesinclude:violentcrime(i.e.,homicide;assault;sexualassault;dangerousornegligentacts;abduction,harassment;androbbery);propertycrime(i.e.,breakandenter;theft;fraudanddeceptionoffences;propertydamageandenvironmentalpollution);drugoffences(i.e.,import;deal;manufacture;useorpossessillicitdrugs);trafficoffences(i.e.,offencesinvolvingtheuseofamotorvehicle);publicorderandmiscellaneousoffences(e.g.,offensiveconduct,publichealthandsafetyoffences);andfinally,offencesagainstjusticeprocedures(i.e.,breachingacourtorder).
Table2reportsthenumberandproportionofoffendersinoursampleundergoingTSandCSthathavecommittedparticularoffences.FromTable2wecanseethatoffendersparticipatinginCSarefarmorelikelytohavecommittedaviolentoffencethanthoseparticipatinginTS(47.7%vs.28.5%),lesslikelytohavecommittedapropertyoffence(12.4%vs.20.8%),drugoffence(5.9%vs.0.5%)oranoffenceagainstjusticeprocedures(17.5%vs.20.9%).
Table 2. Index offences for traditional and circle sentenced groups
Traditional Sentencing Circle Sentencing
N % N % Difference Std Err
(1) (2) (3) (4) (5) (6)
Baseline offence
Violent crime 26,272 28.47 313 47.71 0.192*** (0.020)
Property crime 19,224 20.83 81 12.35 -0.085*** (0.013)
Drug offences 5,449 5.90 3.00 0.46 -0.054*** (0.003)
Traffic offences 13,238 14.35 89 13.57 -0.008 (0.013)
Public order & miscellaneous offence 8,845 9.58 55 8.38 -0.012 (0.011)
Offences against justice procedures 19,256 20.87 115 17.53 -0.033* (0.015)
Total 92,284 100.00 656 100.00 Note.N=observations,StdErr=StandardError,robuststandarderrorsinparentheses,p<.001***,p<.01*,p<.05*.
Table3reportsdescriptivestatisticsforallvariablesusedinourstudy.20Table3containsthreepanels.PanelAprovidesinformationfortheoutcomevariablesofinterest.Theseoutcomemeasuresinclude:
1. Prison: Abinaryvariableequaltooneifthedefendantreceivesasentenceofimprisonment,zerootherwise.
2. Reoffend within 12 months: Among(the82.1%of)offendersthatdonotreceiveaprisonsentence,21thisvariableisequaltooneiftheoffenderhasatleastone(proven)offencewithin12monthsofsentencing,zerootherwise.Weexcludeoffendersreceivingaprisonsentencefromourmeasureofrecidivisminordertoavoidtheeffectofbeingincarceratedfromcontaminatingtheestimates.22
17 Thatis,weremoveappearancesresultinginapenaltythathasneverbeenimposedthroughCS.Forexample,CShasneverresultedinjuveniledetentionsinceonlyadultoffendersareeligibletoparticipate.18 InTableA2oftheAppendixwerelaxthissamplerestriction.19 InTableA2oftheAppendixwerelaxthissamplerestriction(withandwithoutcourtfixedeffects).20 WedonotincludethesetofoffencefixedeffectsinTable2inanyofourregressionsbecauseoftherelativelysmallnumberofoffenderswithineachcategory.Insteadweuseacontinuousmeasureofoffenceseverity,theMedianSentencingRanking,whichisdescribedshortly.21 Itisalsoworthmentioningthatthis,ifanything,shouldproduceaconservativeestimateoftheprogram’sbenefitonrecidivismifCSlowerstheprobabilityof a prison sentence.22 Thatis,priorresearchhasconsistentlyfoundacausallinkbetweenincarcerationandincreasedratesofpost-releaserecidivism(seeforexampleRahman,2019).
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 8
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
3. Days until next re-offence:Among(the58.8%of)offendersthatdonotreceiveaprisonsentenceandre-offendatleastoncepriorto31July2019,thisvariableisequaltothenumberofdaysbetweentheoffender’ssentencingdateandtheirfirst(proven)re-offence.23
Thet-testfromPanelAindicatesthatdefendantsparticipatinginCSare11.7percentagepointslesslikelytoreceiveaprisonsentence.Inrelativeterms,expressedasafractionoftheincarcerationrateforoffendersundergoingTS,thisequatestoapproximatelyatwo-thirdsreduction.PanelAalsoindicatesthatoffendersundergoingCSare5.5percentagepointslesslikelytoreoffendwithin12months(13.6%lesslikelyinrelativeterms).Finally,wecanalsoseethatwhentheydoreoffend,offendersundergoingCStakeaboutfourmonthslongertoreoffend(23.5%longerinrelativeterms).
Itis,however,importanttobearinmindthatoffendersundergoingCSarelikelytosystematicallydifferfromtheirtraditionallysentencedcounterparts.NotonlyareoffendersundergoingCSmorelikelytohavecommittedaviolentoffence,theyhavealsoconsentedtoCSandpassedthesuitabilityassessment.Thedefendant-casecharacteristicsreportedinPanelBallowustoexaminethispropositioninmoredetail.PanelBprovidesdescriptivestatisticsforallcontrolvariablesusedinourstudy.
These control variables include:
• Age: Offender’sageatthetimeofsentencing;
• Age at first contact: Offender’sageatfirstknowncontactwiththeCJS;
• Sex: Abinaryvariableequaltooneiftheoffenderismale,zeroiftheoffenderisfemale;
• SEIFA: TheSEIFApercentilerankfortheoffender’splaceofresidence,whichwehaverecodedintofiveindicatorvariables,oneforeachquartileofthedistributionandoneforthosewithamissingSEIFArank;
• Remoteness: A set of binary variables indicating whether the defendant’s place of residence is in a MajorCity,Innerregional,OuterregionalorRemote/Veryremotearea.Wealsohaveabinaryvariableindicatingwhetherthisinformationismissing;
• Concurrent charges: Numberof(proven)concurrentchargesatthecourtappearance;
• Prior court appearances: Numberofpriorcourtappearances(withatleastoneprovenoffence);
• Prior prison sentences:Numberofpriorprisonsentences;
• Median Sentencing Ranking (MSR):MSRofthedefendant’sprincipaloffence;24
• Plea:Asetofbinaryvariablesindicatingwhetherthedefendantenteredintoapleaof:notguilty;guilty,ornopleaentered.
Table2andPanelBofTable3telltwocompetingstoriesregardingoffendersundergoingCS.Ononehand,offendersparticipatinginCShave:fewerpriorcourtappearances;fewerprisonsentences;aremorelikelytohaveenteredintoapleaofguilty;andhavebeengrantedbail.Thissuggeststhattheyare,onaverage,ofalowerriskprofilethanoffendersundergoingTS.Ontheotherhand,however,offendersparticipatinginCSare:younger;havemoreconcurrentcharges;havecommittedmoreserious(violent)offences;andfinally,madefirstcontactwiththeCJSatanearlierstageinlife.Weare,therefore,unabletosignthebiasassociatedwithCS(i.e.,toknowwhetheroffendersparticipatinginCSareofahigherorlowerriskprofilethanoffendersparticipatinginTS).
23 InTableA2oftheAppendixwelimittheestimationsampletoappearancesfinalisedonorbefore31July2018inordertoleavea12monthfollow-upwindow for all observations.24 TheMSRisameasureofoffenceseverityconstructedfromthepenaltyassociatedwithagivenoffence.MacKinnelletal.(2010)providefurtherinformationregardinghowtheMSRisconstructed.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 9
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Tabl
e 3.
Des
crip
tive
sta
tist
ics
Full
sam
ple
Trad
ition
ally
sen
tenc
edCi
rcle
sen
tenc
edD
iffer
ence
Obs
Mea
nSt
d. D
ev.
Obs
Mea
nSt
d. D
ev.
Obs
Mea
nSt
d. D
ev.
Estim
ate
Std.
Err
.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Pane
l A. O
utco
me
vari
able
s
Pris
on92
,940
0.17
90.38
392
,284
0.18
00.38
465
60.06
30.24
2-0.117
***
(0.010
)
Reoff
end
with
in 1
2 m
onth
s gi
ven
no
pris
on76
,303
0.40
40.49
175
,688
0.40
50.49
161
50.35
00.47
7-0.055
**(0.019
)
V
iole
nt o
ffen
ce76
,303
0.11
30.31
675
,688
0.11
20.31
661
50.15
40.36
20.04
2**
(0.015
)
P
rope
rty
offen
ce76
,303
0.10
00.30
075
,688
0.10
00.30
061
50.07
00.25
5-0.030
**(0.010
)
D
rug
offen
ce76
,303
0.04
30.20
275
,688
0.04
30.20
261
50.02
10.14
4-0.022
***
(0.006
)
W
eapo
n off
ence
76,303
0.00
70.08
475
,688
0.00
70.08
561
50.00
50.07
0-0.002
(0.003
)
T
raffi
c off
ence
76,303
0.07
90.27
075
,688
0.07
90.27
061
50.04
90.21
6-0.031
***
(0.009
)
P
ublic
ord
er &
mis
cella
neou
s
off
ence
s76
,303
0.03
80.19
175
,688
0.03
80.19
161
50.03
90.19
40.00
1(0.008
)
O
ffen
ces
agai
nst j
ustic
e pr
oced
ures
76,303
0.02
50.15
675
,688
0.02
50.15
661
50.01
10.10
6-0.014
**(0.004
)
Reoff
end
for s
elec
ted
crim
e92
,940
0.07
80.26
992
,284
0.07
90.26
965
60.04
00.19
5-0.039
***
(0.008
)
S
elec
ted
viol
ent c
rime
92,940
0.00
40.06
392
,284
0.00
40.06
365
60.00
30.05
5-0.001
(0.002
)
S
elec
ted
prop
erty
crim
e92
,940
0.07
40.26
292
,284
0.07
50.26
365
60.03
70.18
8-0.038
***
(0.007
)
Day
s un
til fi
rst r
eoff
ence
54
,674
534.88
765
8.51
654
,226
533.86
365
7.75
244
865
8.85
373
5.54
712
4.99
0***
(34.82
8)
Pane
l B. C
ontr
ol v
aria
bles
Age
92,933
31.355
9.93
392
,277
31.364
9.93
665
630
.140
9.48
8-1.223
***
(0.372
)
Age
at fi
rst c
onta
ct w
ith ju
stic
e sy
stem
92,790
20.894
8.64
992
,134
20.902
8.65
965
619
.739
7.06
8-1.163
***
(0.277
)
Mal
e92
,940
0.71
80.45
092
,284
0.71
80.45
065
60.69
70.46
0-0.021
(0.018
)
SEIF
A Q
1 (L
owes
t SES
)92
,940
0.35
90.48
092
,284
0.35
70.47
965
60.72
30.44
80.36
6***
(0.018
)
SEIF
A Q
292
,940
0.28
50.45
192
,284
0.28
50.45
265
60.16
80.37
4-0.118
***
(0.015
)
SEIF
A Q
392
,940
0.19
50.39
692
,284
0.19
60.39
765
60.08
80.28
4-0.107
***
(0.011
)
SEIF
A Q
4 (H
ighe
st S
ES)
92,940
0.05
60.23
092
,284
0.05
60.23
165
60.00
60.07
8-0.050
***
(0.003
)
Mis
sing
SEI
FA92
,940
0.10
50.30
792
,284
0.10
60.30
765
60.01
50.12
3-0.090
***
(0.005
)
Inne
r reg
iona
l92
,940
0.28
30.45
192
,284
0.28
20.45
065
60.49
40.50
00.21
2***
(0.020
)
Maj
or c
ities
92,940
0.39
00.48
892
,284
0.39
20.48
865
60.16
90.37
5-0.223
***
(0.015
)
Out
er re
gion
al92
,940
0.18
10.38
592
,284
0.18
10.38
565
60.13
40.34
1-0.047
***
(0.013
)
Rem
ote
92,940
0.04
10.19
992
,284
0.04
00.19
765
60.18
80.39
10.14
7***
(0.015
)
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 10
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Tabl
e 3.
Des
crip
tive
sta
tist
ics
(con
tinue
d)
Full
sam
ple
Trad
ition
ally
sen
tenc
edCi
rcle
sen
tenc
edD
iffer
ence
Obs
Mea
nSt
d. D
ev.
Obs
Mea
nSt
d. D
ev.
Obs
Mea
nSt
d. D
ev.
Estim
ate
Std.
Err
.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Mis
sing
Are
a92
,940
0.10
40.30
592
,284
0.10
50.30
665
60.01
50.12
3-0.090
***
(0.005
)
No.
con
curr
ent c
harg
es92
,940
1.64
71.01
992
,284
1.64
51.01
965
61.97
40.87
80.32
9***
(0.034
)
MSR
of p
rinci
pal o
ffen
ce92
,940
74.994
29.080
92,284
75.037
29.117
656
68.980
22.564
-6.057
***
(0.886
)
Bail
at fi
rst c
ourt
app
eara
nce
92,896
0.91
90.27
392
,240
0.91
90.27
365
60.95
30.21
20.03
4***
(0.008
)
Plea
d gu
ilty
92,940
0.61
30.48
792
,284
0.61
20.48
765
60.86
30.34
40.25
1***
(0.014
)
No
plea
ent
ered
92,940
0.24
00.42
792
,284
0.24
10.42
865
60.09
90.29
9-0.142
***
(0.012
)
Plea
d no
t gui
lty92
,940
0.14
70.35
492
,284
0.14
80.35
565
60.03
80.19
2-0.109
***
(0.008
)
No.
prio
r cou
rt a
ppea
ranc
es
(with
pro
ven
offen
ces)
92,940
8.05
07.53
592
,284
8.05
97.54
265
66.84
56.30
5-1.214
***
(0.247
)
No.
prio
r pris
on s
ente
nces
92,940
1.90
23.45
292
,284
1.90
73.45
765
61.20
12.52
1-0.706
***
(0.099
)
Pane
l C. S
ente
ncin
g ou
tcom
es
Bond
with
sup
ervi
sion
92,940
0.10
00.30
092
,284
0.09
90.29
865
60.31
30.46
40.21
4***
(0.018
)
Bond
with
out c
onvi
ctio
n w
ith
supe
rvis
ion
92,940
0.00
20.03
992
,284
0.00
10.03
665
60.03
00.17
20.02
9***
(0.007
)
Bond
with
out c
onvi
ctio
n w
ithou
t su
perv
isio
n92
,940
0.04
50.20
792
,284
0.04
40.20
665
60.07
90.27
00.03
5***
(0.011
)
Bond
with
out s
uper
visi
on92
,940
0.13
20.33
892
,284
0.13
20.33
865
60.08
20.27
5-0.050
***
(0.011
)
Conv
ictio
n on
ly92
,940
0.02
50.15
692
,284
0.02
50.15
665
60.00
30.05
5-0.022
***
(0.002
)
Fine
92,940
0.24
80.43
292
,284
0.25
00.43
365
60.02
70.16
3-0.222
***
(0.007
)
Hom
e de
tent
ion
92,940
0.00
10.02
892
,284
0.00
10.02
865
60.00
20.03
90.00
1(0.002
)
Impr
ison
men
t92
,940
0.17
90.38
392
,284
0.18
00.38
465
60.06
30.24
2-0.117
***
(0.010
)
No
conv
ictio
n re
cord
ed92
,940
0.01
60.12
592
,284
0.01
60.12
665
60.00
80.08
7-0.008
*(0.003
)
No
pena
lty92
,940
0.11
80.32
392
,284
0.11
90.32
465
60.01
40.11
6-0.105
***
(0.005
)
Oth
er p
enal
ties
92,940
0.02
10.14
492
,284
0.02
10.14
565
60.00
90.09
5-0.012
**(0.004
)
Inte
nsiv
e Co
rrec
tion
Ord
er92
,940
0.00
70.08
692
,284
0.00
70.08
665
60.00
90.09
50.00
2(0.004
)
Com
mun
ity S
ervi
ce O
rder
92,940
0.03
60.18
592
,284
0.03
50.18
465
60.11
40.31
80.07
9***
(0.012
)
Susp
ende
d se
nten
ce w
ith s
uper
visi
on92
,940
0.04
50.20
792
,284
0.04
40.20
465
60.20
70.40
60.16
4***
(0.016
)
Susp
ende
d se
nten
ce w
ithou
t su
perv
isio
n92
,940
0.02
60.15
992
,284
0.02
60.15
965
60.04
00.19
50.01
4(0.008
)
Note.N=o
bservatio
ns,rob
uststand
arderrorsinparentheses,p<.00
1***,p<.01
**,p<.05
*.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 11
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
PanelCprovidesthedistributionofpenaltiesimposedonoffendersinoursample.FromPanelCwecanseethatthemostcommonlyimposedpenaltiesare:abond(10.0%withsupervisionandconviction,13.2%withoutsupervisionwithconviction);afine(24.8%);imprisonment(17.9%);andfinally,nopenalty(11.8%).Interestingly,whencomparedwiththeirtraditionallysentencedcounterparts,offendersundergoingCSare:21.4,7.9,and16.4percentagepointsmorelikelytoreceiveabond(withsupervision),aCommunityServiceOrderorasuspendedsentence(withsupervision).ThisappearstobedrivenmostlybylargereductionsintheprobabilityofCSparticipantsreceivingafine,nopenaltyorimprisonment.
Empirical approach
FromTables2and3weknowthatoffendersparticipatinginCSsystematicallydifferfromoffendersparticipatinginTS.Assuch,weshouldapproachasimplecomparisonofoutcomeswithcaution.Forconcreteness,butwithoutlossofgeneralitytootheroutcomemeasures,supposethatwe’reinterestedinidentifyingthecausaleffectofCSontheprobabilitythatanoffenderre-offendswithin12monthsofsentencing.25
This relationship is given by the LogisticregressioninEquation1below.
yit = Λ(β0 + β1CSit + γX’it + λt + ϵit) (1)
Where i indexes a case and t indexes a month-year.26 yit is a binary variable taking value one if the defendantinagivencasereoffendswithin12monthsoftheirsentencingdate,zerootherwise.CS
it is
abinaryvariableequaltoonefordefendantssentencedthroughCS,zerofordefendantssentencedthrough TS. X’it representsthesetofcontrolvariablesdescribedinPanelBofTable3.λt represents a set ofmonth-by-yearFixedEffects(FEs).TheseFEslimitourcomparisontooffenderssentencedwithinthesamemonth-year.ThisrendersourestimatesrobusttotimevaryingfactorsthatinfluencecrimeacrossNSW(e.g.,changestotheunemploymentrate,seasonalityandadvancesinsecuritytechnology).ϵit is the errortermandallothertermsarecoefficientstobeestimated.Thecoefficientofinterestisβ1,whichcharacterisestherelationshipbetweenCSandtheprobabilityofatleastonereoffence.
Inorderforβ1tohaveacausalinterpretation,participationinCSwouldhavetobe,netofcontrolsandFEs,unrelatedtoallotherfactorsthatinfluencerecidivism.Thereis,however,goodreasontoexpectthatthisisnotthecase.Forexample,remorseandconnectednesstothelocalcommunityareexplicitlyconsideredbytheACJGwhenassessinganoffender’ssuitability.Thesefactorsarealsolikelytobeassociatedwithlowerlevelsofrecidivism.Assuch,anyregressionofEquation1maycauseustooverestimatethebenefitofCSonrecidivism.Tothebestofourknowledge,thereisnowaytocompletelyaddress this issue given available data.27We,therefore,recommendcautionwheninterpretingthe(associative,non-causal)estimatesreportedinthisbulletin.
25 WeuseaZero-TruncatedNegativeBinomialregressiontoestimatetherelationshipbetweenCSandthenumberofdaysbetweensentencingandtheoffender’sfirstreoffence.26 With Λ(z)=1/(1+e-z). 27 Wedid,however,considertwoalternativeidentificationstrategies.ThefirstwasanInstrumentalVariables(IV)strategydesignedtoexploitvariationineachmagistrate’spropensitytoreferanoffenderforCSasaninstrumentforparticipation.Unfortunately,thefirststagerelationshipwastooweaktosupporttheuseofthisinstrument.ThesecondwasanIVstrategydesignedtoexploitvariationinthetimingoftherolloutofCS.Specifically,thisstrategyinvolveslimitingtheestimationsampletomattersfinalisedincourtsthatwill(atsomepoint)haveCSavailable,creatinganindicatorvariableforwhetherthedefendant’smatterwasfinalisedinacourtwithCSavailable(inthecorrespondingmonth-year),andthenusingthisindicatorasaninstrumentforparticipation.Weelectednottoemploythisstrategyforthreereasons.First,thefactthatwehavesofewtreatedunitsmeanswewouldbeunlikelytodetectastatisticallysignificanteffect(evenifonewaspresent)underTwo-StageLeastSquares.Second,inspectionofeachcourt’saggregatepre-policytrendsinincarcerationsandrecidivismrevealeddivergingtrendsinmanysites.Andfinally,inourview,theexclusionrestrictionisparticularlyhardtojustify.Ifforexample,somesiteswereprioritisedforCSbecauseofanincreasingrateofAboriginalrecidivism,thentherolloutcannotbeusedasaninstrumentforparticipation.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 12
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
RESULTS
Incarceration and recidivism
Table4reportsthemainresultsandconsistsofthreepanels,oneforeachoutcomemeasure.PanelA examines the relationship between CS and the probability of a prison sentence.28Column1reportsestimatesfromasimple(unconditional)comparisonanalogoustothet-testinTable3.Columns2and3includecontrolvariablesandmonth-by-yearFEs,respectively.Columns2and3indicatethatoffendersundergoingCSare9.3percentagepointslesslikelytoreceiveaprisonsentence.Inrelativeterms,expressedasafractionoftherateatwhichoffendersundergoingTSaresenttoprison,thisequatestoadecreaseof51.7percent.Thesereductions,bothabsoluteandrelative,arestriking.Whileatleastsomeofthisreductionislikelyduetoselectionbias,ourviewisthatthepracticalsignificanceoftheseresultscannot be taken lightly.
Table 4. Relationship between Circle Sentencing, incarceration and recidivism
(1) (2) (3) (4)
Naive Controls Time FE Penalty FE
Panel A. Prison -0.117*** -0.093*** -0.093***
(0.010) (0.010) (0.009)
Observations 92,940 92,746 92,746
Pseudo R2 0.001 0.373 0.381
AUC 0.503 0.894 0.897
Panel B. Reoffend within 12 months -0.055** -0.044* -0.039* -0.030
(0.019) (0.018) (0.018) (0.019)
Observations 76,303 76,159 76,159 76,159
Pseudo R2 0.000 0.079 0.082 0.086
AUC 0.501 0.690 0.694 0.697
Panel C. Days to reoffence 124.661*** 120.721*** 63.431* 55.171*
(34.742) (34.775) (28.079) (28.007)
Observations 54,674 54,569 54,569 54,569
Pseudo R2 0.000 0.006 0.013 0.013
Controls N Y Y Y
Time FE N N Y Y
Penalty FE N N N Y
Note.PanelsAandBreportaveragemarginaleffectsderivedfromaLogisticregression.PanelCreportsaveragemarginaleffectsderivedfromaZero-TruncatedNegativeBinomialregression,AUC=AreaUnderthereceiveroperatingcharacteristicCurve,FE=FixedEffects,standarderrorsobtainedusingtheDeltamethodinparentheses,p<.001***,p<.01**,p<.05*.
PanelBexaminestherelationshipbetweenCSandtheprobabilityofatleastonere-offencewithin12monthsofsentencing.Column3indicatesthatCSisassociatedwitha3.9percentagepointreductionintheprobabilityofatleastonere-offencewithin12months.Inrelativeterms,expressedasafractionoftherecidivismrateofoffendersundergoingTS,thisequatestoadecreaseof9.6percent.Incolumn4weincludeasetofpenaltyFEs.Thatis,weconstrainourcomparisontooffendersreceivingthesamepenaltyandthenre-estimateEquation1.Interestingly,thecoefficientisnowaboutone-quartersmallerin(absolute)sizeandisstatisticallyinsignificant.ThissuggeststhatatleastsomeofthereductioninrecidivismassociatedwithCScanbeattributedtodefendantsreceivingdifferentpenalties.One
28 PanelsAandBreportaveragemarginaleffectsderivedfromaLogisticregression.PanelCreportsaveragemarginaleffectsderivedfromaZero- TruncatedNegativeBinomialregression.InterestedreadersaredirectedtoTableA4intheAppendix,whichreportsthefullsetofestimatescorrespondingtothese regressions.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 13
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
interpretationofthisfindingisthatcirclegroups,whichincludeamagistrate,areabletoassignmoreeffectivepenaltiesthanamagistrateworkinginisolation.Thisisdiscussedfurtherinthefinalsectionofthe bulletin.
PanelCexaminestherelationshipbetweenCSandthenumberofdaysbetweensentencingandtheoffender’sfirstre-offence.Column3indicatesthatCSisassociatedwithanadditional55daysbeforetheoffender’sfirstreoffence.Inrelativeterms,expressedasafractionofthenumberofdaysbeforeoffendersundergoingTSreoffend,thisequatestoanincreaseof10.3percent.InclusionofthemagistrateFEsincolumn4generatesareductioninsizeandstatisticalprecision,althoughthecoefficientremainsmarginallysignificantatthefivepercentlevel.
Recidivism for specific offences
OurinabilitytoaddresstheselectionbiasissuesoutlinedearliermeansweareunabletomakeanycausalclaimsregardingtheeffectofCSonincarcerationsorrecidivism.Thatsaid,the(significant)negativeassociation between CS and recidivism represents a substantive departure from prior research on CS (Fitzgerald,2008)andRJprogramsmoregenerally(Bergsethand2007;Poynton,2013;Strangetal.,2013;Smith&Weatherburn,2012).Assuch,theseestimateswarrantfurtherinvestigation.
Inordertounpackwhatmaybedrivingthisassociationfurther,wedivideourmeasureofrecidivismintosevenbinaryvariables.Thatis,werecode“Reoffendwithin12months”intosevenbinaryvariables.Eachofthesebinaryvariablestakesvalueoneiftheoffenderreoffendswithin12monthsandtheirfirstreoffenceisaparticulartypeofoffence.29Forexample,oneofthesevennewoutcomevariablestakesvalueoneiftheoffenderbothreoffendswithin12monthsandtheirfirstreoffenceisaviolentoffence.Wethen re-estimateEquation1overthesesevenoutcomes(i.e.,oneregression/outcome)andreporttheestimatesinTable5.FromthefirstrowinTable5wecanseethatoffendersundergoingCSare3.2percentagepointsmorelikelytoreoffendforaviolentoffence.FromTable5wecanalsoseethatthisincreaseismorethanoffsetbydecreasesinreoffendingforpropertycrime(2.2percentagepoints),drugoffences(1.6percentagepoints),trafficoffences(3.2percentagepoints)andoffencesagainstjusticeprocedures(1.1percentagepoints).
Table 5. Relationship between Circle Sentencing and reoffending by reoffence type
Crime category Estimate Standard error
Panel A. All crime
Violent crime 0.032* (0.014)
Property crime -0.022* (0.011)
Drug offences -0.016* (0.007)
Weapon offences -0.001 (0.004)
Public order & miscellaneous 0.000 (0.008)
Traffic offences -0.032*** (0.009)
Offences against justice procedures -0.011* (0.005)
Panel B. Crimes unaffected by reporting/detection bias
Selected violent and property crime -0.020* (0.010)
Selected violent crime 0.005 (0.005)
Selected property crime -0.025** (0.009)Note.ThistablereportsaveragemarginaleffectsderivedfromaLogisticregression,standarderrorsobtainedusingtheDeltamethodinparentheses,p<.001***,p<.01**,p<.05*.
29 Inthisanalysisthemethodusedtoclassifyoffencesintocrimecategorieswasconsistentwiththeclassificationsfortheprincipaloffencetype (seeprevioussectionondescriptivestatistics).DescriptivestatisticsforthesevariablesarereportedinTableA3oftheAppendix.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 14
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Oneissueweareyettoaddressisthepossibilityofreportinganddetectionbiascontaminatingourmeasuresofrecidivism.RecallfromPanelCofTable3thatoffendersundergoingCSaremorelikelytoreceivesomeformofsupervision(e.g.,abondorsuspendedsentence)thanoffendersundergoingTS.Itisreasonable,therefore,toquestionwhethertheapparentincreaseinviolentcrime(ordecreaseinothercrimecategories)isareflectionofenhanceddetectionofoffences,ratherthananincrease(ordecrease)intheactuallevelofoffending.Inordertobetterunderstandthisproblem,inPanelBofTable5welimitouranalysistospecifictypesofviolentandpropertycrimeconsideredtobelesssusceptibletoreportingand detection bias.30Theseviolentcrimesinclude:homicide;assaultoccasioninggrievousbodilyharm;androbbery.Thepropertycrimesinclude:breakandenter;theft;andmotorvehicletheft.Wethenre-estimateEquation1usingabinaryvariableequaltooneiftheoffenderre-offendswithin12monthsandtheirfirstreoffenceisoneoftheseselectedcrimes.FromPanelBofTable5wecanseethatCShasno(statisticallysignificant)associationwithviolentrecidivism,whiletheassociationbetweenCSandreoffendingforpropertycrimeislargelyconsistentwithitscounterpartsinTable4andPanelAofTable5.
DISCUSSIONInthisstudywesetouttoexaminetherelationshipbetweenCircleSentencing(CS)andlikelihoodofincarcerationandrecidivism.Wefoundthatnetofcontrolsandfixedeffects,offendersparticipatinginCSare9.3percentagepointslesslikelytoreceiveaprisonsentence.Inrelativeterms,thisequatestoareductionof51.7percent.
Thequestion,therefore,iswhetherthisreductioninincarcerationswasaccompaniedbyanincreaseinrecidivismforoffendersnotsenttoprison.Weusedtwomeasuresofrecidivismtoanswerthisquestion.First,theprobabilityofatleastonere-offencewithin12months;andsecond,thenumberofdaysbetweensentencingandtheoffender’sfirstre-offence.Withrespecttotheformer,wefoundthatoffendersparticipatinginCSare3.9percentagepointslesslikelytoreoffend(9.6%inrelativeterms).Withrespecttothelatter,wefoundthatoffendersparticipatinginCStake55dayslongertoreoffendwhentheydocommitanewoffence(arelativeincreaseof10.3%).
Thereare,however,twocaveatsthatneedtobeconsideredwheninterpretingourestimates.Thefirstisthatourestimatescannotbeinterpretedcausally(i.e.,selectionbiasmayberesponsibleforourresults).Thesecondisthat,evenifour(recidivism)estimatescouldbeinterpretedcausally,wealsofoundsome(limited)evidenceindicatingthatCSmaybeassociatedwithanincreaseinviolentrecidivism;althoughthisincreaseismorethanoffsetbyreductionsinnon-violentcrime.Whetherthebenefitofanetreductionin(non-violent) crime exceeds the cost of an increase in violent crime is beyond the scope of this paper but is an important avenue for future research.31
Inanyevent,ourrecidivismestimatesmeaningfullydepartfromthosereportedbyFitzgerald(2008).Oneexplanationforthisdepartureisteethingissuesduringtheearlyyearsoftheprogram.Forexample,DalyandProietti-Scifoni(2009)identifiedanumberoflimitationsregardingtheearlyimplementationofCS,includinginadequatedrugandalcoholsupportservicesinsomelocations.Therefore,it’spossiblethatCSwasnotoperatingasintendeduntilafterFitzgerald’sevaluation.Anotherrelatedexplanationissamplesize.OursampleissubstantiallylargerthanthesampleavailabletoFitzgerald(2008).Assuch,Fitzgeraldmayhavelackedsufficientpowertodetectaneffect(evenifonewaspresent).
Despiteitslimitations,ourstudydoeshaveseveralimportantimplicationsforresearchersandpolicymakers.Thefirstofwhichistobetterunderstandwhycirclegroupsassigndifferentpenaltieswhencomparedtoamagistrateworkinginisolation.Recallthatoncewelimitedourcomparisontooffenders
30 Thesecrimesareconsideredtobelesssusceptibletoreportinganddetectionbiasbecausevictimshavemoreincentivetoreportsuchoffencestopolice.DescriptivestatisticsforthesevariablesarereportedinTableA3oftheAppendix.31 Mayhew(2003)providesthemostrecentestimatesofthecostsofcrimeinAustralia.Wedonot,however,usetheseestimatestoconductacost-benefitanalysis as the information is likely to be out of date for the majority of our estimation sample
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 15
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
receivingasimilarpenalty,theassociationbetweenCSandrecidivismreducedinmagnitudeandstatistical precision. This suggests that at least some of the association between CS and recidivism is duetocirclegroupsassigningdifferent,potentiallymoreeffective,penalties.Thiscouldbebecausecirclegroupshaveadeeperinsightintothecircumstancesoftheoffenderandarethereforeabletoidentifymoreappropriatepenalties(e.g.,offendersparticipatinginCSarebothmoredisadvantagedthanoffendersparticipatinginTS,andlesslikelytoreceivemonetaryfine).ThesecondistoinvestigatethelinkbetweenCSandhealthoutcomes.GiventhatdrugandalcoholissuesareprevalentamongCSparticipants(CulturalandIndigenousResearchCentre,2008;Daly&Proietti-Scifoni,2009),futureresearch could investigate the relationship between CS and health outcomes by linking (drug and alcohol related)emergencydepartmentpresentationsandhospitalisationsdatawithBOCSAR’sReoffendingDatabase.Thethirdistodetermineif,andunderwhatconditions,CScanbeintroducedinotherlocalities.ExpansionoftheCircleSentencingprogramtootherlocationswouldrequire(1)localsupportfrommagistratesandpoliceprosecutors,(2)availableandaccessiblelegalaidandhealthservices(e.g.drugandalcoholtreatmentfacilities),and(3)arelativelylargeIndigenouspopulation.
Tosummarise,CSclearlyhasthepotentialtolowertheIndigenousincarcerationrate.IfCScanachievethisgoal,withoutadverselyaffectingrecidivism,thenetbenefittosocietyisdifficulttooverstate.Forexample,overthe2016-17financialyear,thecosttotheNSWgovernmentofincarceratinganindividualwas$253perday.32Overthissameperiod,therewere3,141IndigenousAustralianheldincustodyeachday.Hence,evenaone-percentagepointdecreaseequatesto31fewerincarcerationsperday.Thisimpliesasavingof$7,843perdayor$2,862,695peryear.Onthesegroundsalonefurtherresearch,ideallyintheformofalongrunningrandomisedcontrolledtrial,todeterminethetruecausaleffectofCSonreoffendingiscertainlyjustified.
ACKNOWLEDGEMENTSTheauthorswouldliketoexpresstheirgratitudetoMareeJenningsandEdwinaCrawfordfromtheAboriginalServicesUnitfortheirinvaluableadviceontheoperationofCircleSentencinginNSW.Theauthorswouldalsoliketothankthetwoanonymousreviewersfortheirconstructivefeedback,aswellascolleaguesatBOCSAR,inparticularSuzannePoynton,ClareRinglandandLilyTrimbolifortheircommentson earlier drafts of this bulletin.
REFERENCESAustralianBureauofStatistics.(2011a).Census of Population and Housing: Socio-Economic Indexes for Areas (SEIFA), Australia, 2011(Cat.No.2033.0.55.001).Retrieved20Jul2019fromhttp://www.abs.gov.au/websitedbs/censushome.nsf/home/seifa.
AustralianBureauofStatistics.(2011b).Australian and New Zealand Standard Offence Classifications (ANZSOC): Australia(Cat.No.1234.0).Retrieved20Jul2019from:http://www.abs.gov.au/ausstats/[email protected]/mf/1234.0.
AustralianBureauofStatistics.(2016).2016 Census QuickStats.Retrieved20July2019fromhttp://quickstats.censusdata.abs.gov.au/census_services/getproduct/census/2016/quickstat/036.
Bergseth,K.J.,&Bouffard,J.A.(2007).Thelong-termimpactofrestorativejusticeprogrammingforjuvenileoffenders.Journal of Criminal Justice,35,433–451.
32 TheNSWgovernment’stotalnetoperatingexpenditure(includingdepreciation)ontheincarcerationsystemwas$1.1billion.Anaverageof11,916individualswereheldincustodyeachday,3,141ofwhichwereIndigenous.AllinformationusedinthesecalculationsisderivedfromtheReportonGovernmentServices(ProductivityCommission,2018).
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 16
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Biddle,N.,&Markham,F.(2018).Indigenous identification change between 2011 and 2016: evidence from the Australian Census Longitudinal Dataset.Canberra:CentreforAboriginalEconomicPolicyResearch,AustralianNationalUniversity.
CulturalandIndigenousResearchCentre.(2008).NSW Attorney General’s Department Evaluation of Circle Sentencing Program Report.Retrieved20January2020fromtheIndigenousJusticeClearinghousewebsite:https://www.indigenousjustice.gov.au/resources/nsw-attorney-generals-department-evaluation-of-circle-sentencing-program-report/.
Daly,K.,&Proietti-Scifoni,G.(2009).Defendants in the Circle: Nowra Circle Court, the presence, and impact of Elders, and re-Offending.Retrieved10March2020fromhttps://www.griffith.edu.au/__data/assets/pdf_file/0020/227720/2009-Daly-and-Proietti-Scifoni-Defendants-in-the-Circle-updated-Aug-2015.pdf
Fitzgerald,J.(2008).Does circle sentencing reduce Aboriginal offending?(CrimeandJusticeBulletinNo.115).Retrieved10March2020fromNSWBureauofCrimeStatisticsandResearchwebsite:https://www.bocsar.nsw.gov.au/Documents/CJB/cjb115.pdf
Jones,C.(2009).Does Forum Sentencing reduce re-offending?(CrimeandJusticeBulletinNo.129).Retrieved10March2020fromNSWBureauofCrimeStatisticsandResearchwebsite:https://www.bocsar.nsw.gov.au/Documents/CJB/cjb129.pdf
Larsen,J.(2014).RestorativejusticeintheAustralian criminal justice system (Research and public policy seriesno.127).Canberra:AustralianInstituteofCriminology.Retrieved1January2020fromhttps://aic.gov.au/publications/rpp/rpp127.
Latimer,J.,Dowden,C.,&Muise,D.(2005).Theeffectivenessofrestorativejusticepractices:Ameta-analysis. The Prison Journal, 85(2),127–144.
Little,S.,Stewart,A.,&Ryan,N.(2018).Restorativejusticeconferencing:Notapanaceafortheover-representationofAustralia’sIndigenousyouthinthecriminaljusticesystem.International Journal of Offender Therapy and Comparative Criminology, 62(13),4067-4090.
MacKinnell,I.,Poletti,P.,&Holmes,M.(2010).Measuring offence seriousness(CrimeandJusticeBulletinNo.142).Retrieved10March2020fromNSWBureauofCrimeStatisticsandResearchwebsite:https://www.bocsar.nsw.gov.au/Documents/CJB/cjb142.pdf
Marchetti,E.&Daly,K.(2004).Indigenous courts and justice practices in Australia(TrendsandIssuesinCrimeandCriminalJusticeNo.277).RetrievedfromAustralianInstituteofCriminologywebsite:https://aic.gov.au/publications/tandi/tandi277.
Mayhew,P.(2003).Counting the costs of crime in Australia(TrendsandIssuesinCrimeandCriminalJusticeNo.247).Retrieved10March2020fromAustralianInstituteofCriminologywebsite:https://aic.gov.au/publications/tandi/tandi247
McGarrell,E.F.&Hipple,N.K.(2007).FamilyGroupConferencingandre-offendingamongfirst-timejuvenileoffenders:TheIndianapolisexperiment.Justice Quarterly,24,221-246.
Potas,I.,Smart,J.,Brignell,G.,Thomas,B.,&Lawrie,R.(2003).Circle Sentencing in New South Wales: A review and evaluation.Retrieved10March2020fromJudicialCommissionofNewSouthWaleswebsite:https://www.judcom.nsw.gov.au/wp-content/uploads/2016/07/research-monograph-22.pdf
Poynton,S.(2013).Rates of recidivism among offenders referred to Forum Sentencing(CrimeandJusticeBulletinNo.172).Retrieved10March2020fromNSWBureauofCrimeStatisticsandResearchwebsite:https://www.bocsar.nsw.gov.au/Documents/CJB/cjb172.pdf
ProductivityCommission.(2018).Report on Government Services 2018.Retrieved10August2019fromhttps://www.pc.gov.au/research/ongoing/report-on-government-services/2018/justice/corrective-services/rogs-2018-partc-chapter8.pdf
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 17
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Rahman,S.(2019).The marginal effect of bail decisions on imprisonment, failure to appear, and crime (Crime andJusticeBulletinNo.224).Retrieved10March2020fromNSWBureauofCrimeStatisticsandResearchwebsite:https://www.bocsar.nsw.gov.au/Documents/CJB/2019-Report-Bail-decisions-on-imprisonment-CJB224.pdf
Shapland,J.,Atkinson,A.,Atkinson,H.,Chapman,B.,Dignan,J.,Howes,M.,Johnstone,J.,Robinson,G.,&Sorsby,A.(2007).Restorative justice: The views of victims and offenders. The third report from the evaluation of three schemes.London:MinistryofJustice.
Smith,N.,&Weatherburn,D.(2012).Youth justice conferences verses Children’s Court: A comparison of re-offending.(CrimeandJusticeBulletinNo.160).Retrieved10March2020fromNSWBureauofCrimeStatisticsandResearchwebsite:https://www.bocsar.nsw.gov.au/Documents/CJB/cjb160.pdf
Strang,H.,Sherman,L.,Angel,C.M.,Woods,D.J.,Bennett,S.,Newbury-Birch,D.,&Inkpen,N.(2006).Victimevaluationsofface-to-facerestorativejusticeconferences:Aquasi-experimentalanalysis. Journal of Social Issues, 62(2),281–306.
Strang,H.,Sherman,L.,Mayo-Wilson.E.,Woods,D.,&Ariel,B.(2013).Restorativejusticeconferencing(RJC)usingface-to-facemeetingsofoffendersandvictims:Effectsonoffenderrecidivismandvictimsatisfaction. A systematic review. Campbell Systematic Reviews,12.Retrieved10March2020fromhttps://restorativejustice.org.uk/sites/default/files/resources/files/Campbell%20RJ%20review.pdf
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 18
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Tabl
e A1
. Cha
ract
eris
tics
of t
hose
ref
erre
d to
Cir
cle
Sent
enci
ng, b
y pa
rtic
ipat
ion
stat
us
Ever
yone
refe
rred
Refe
rred
but
tr
aditi
onal
ly s
ente
nced
Circ
le s
ente
nced
Diff
eren
ce
Obs
Mea
nSt
d. D
ev.
Obs
Mea
nSt
d. D
ev.
Obs
Mea
nSt
d. D
ev.
Estim
ate
Std.
Err
.(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)(9
)(1
0)(1
1)
Pane
l A. O
utco
me
vari
able
s
Pris
on80
50.09
70.29
614
90.24
80.43
365
60.06
30.24
2-0.186
***
(0.037
)
Reoff
end
with
in 1
2 m
onth
s gi
ven
no
pris
on72
70.36
90.48
311
20.47
30.50
261
50.35
00.47
7-0.124
*(0.051
)
V
iole
nt o
ffen
ce72
70.15
70.36
411
20.17
00.37
761
50.15
40.36
2-0.015
(0.038
)
P
rope
rty
offen
ce72
70.07
20.25
811
20.08
00.27
361
50.07
00.25
5-0.010
(0.028
)
D
rug
offen
ce72
70.02
50.15
511
20.04
50.20
761
50.02
10.14
4-0.024
(0.020
)
W
eapo
n off
ence
727
0.00
70.08
311
20.01
80.13
361
50.00
50.07
0-0.013
(0.013
)
T
raffi
c off
ence
727
0.05
80.23
311
20.10
70.31
161
50.04
90.21
6-0.058
(0.031
)
Publ
ic o
rder
& m
isce
llane
ous
offen
ces
727
0.03
90.19
311
20.03
60.18
661
50.03
90.19
40.00
3(0.019
)
Off
ence
s ag
ains
t jus
tice
proc
edur
es72
70.01
20.11
111
20.01
80.13
361
50.01
10.10
6-0.006
(0.013
)
Day
s un
til fi
rst r
eoff
ence
60
762
9.46
870
2.36
712
550
3.43
255
6.60
448
266
2.15
473
2.43
115
8.72
2**
(59.84
2)
Pane
l B. C
ontr
ol v
aria
bles
Age
805
30.181
9.41
414
930
.362
9.10
865
630
.140
9.48
8-0.222
(0.832
)
Age
at fi
rst c
onta
ct w
ith ju
stic
e sy
stem
805
19.770
7.12
514
919
.906
7.39
365
619
.739
7.06
8-0.167
(0.664
)
Mal
e80
50.70
40.45
714
90.73
80.44
165
60.69
70.46
0-0.042
(0.040
)
SEIF
A Q
1 (L
owes
t SES
)80
50.71
30.45
314
90.67
10.47
165
60.72
30.44
80.05
1(0.042
)
SEIF
A Q
280
50.16
30.36
914
90.14
10.34
965
60.16
80.37
40.02
7(0.032
)
SEIF
A Q
380
50.09
30.29
114
90.11
40.31
965
60.08
80.28
4-0.026
(0.028
)
SEIF
A Q
4 (H
ighe
st S
ES)
805
0.00
50.07
014
90.00
00.00
065
60.00
60.07
80.00
6*(0.003
)
Mis
sing
SEI
FA80
50.02
60.15
914
90.07
40.26
265
60.01
50.12
3-0.059
**(0.022
)
Inne
r reg
iona
l80
50.47
70.50
014
90.40
30.49
265
60.49
40.50
00.09
1*(0.045
)
Maj
or c
ities
805
0.15
90.36
614
90.11
40.31
965
60.16
90.37
50.05
5(0.030
)
Out
er re
gion
al80
50.14
50.35
314
90.19
50.39
765
60.13
40.34
1-0.060
(0.035
)
Rem
ote
805
0.19
30.39
514
90.21
50.41
265
60.18
80.39
1-0.027
(0.037
)
Mis
sing
Are
a80
50.02
60.15
914
90.07
40.26
265
60.01
50.12
3-0.059
**(0.022
)
No.
con
curr
ent c
harg
es80
51.95
40.89
914
91.86
60.98
465
61.97
40.87
80.10
8(0.087
)
APPENDIX
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 19
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Tabl
e A1
. Cha
ract
eris
tics
of t
hose
ref
erre
d to
Cir
cle
Sent
enci
ng, b
y pa
rtic
ipat
ion
stat
us (c
ontin
ued)
Ever
yone
refe
rred
Refe
rred
but
tr
aditi
onal
ly s
ente
nced
Circ
le s
ente
nced
Diff
eren
ce
Obs
Mea
nSt
d. D
ev.
Obs
Mea
nSt
d. D
ev.
Obs
Mea
nSt
d. D
ev.
Estim
ate
Std.
Err
.(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)(9
)(1
0)(1
1)
MSR
of p
rinci
pal o
ffen
ce80
568
.853
23.011
149
68.295
24.954
656
68.980
22.564
0.68
5(2.222
)
Bail
at fi
rst c
ourt
app
eara
nce
805
0.95
40.21
014
90.96
00.19
765
60.95
30.21
2-0.007
(0.018
)
Plea
d gu
ilty
805
0.84
80.35
914
90.78
50.41
265
60.86
30.34
40.07
8*(0.036
)
No
plea
ent
ered
805
0.10
70.30
914
90.14
10.34
965
60.09
90.29
9-0.042
(0.031
)
Plea
d no
t gui
lty80
50.04
50.20
714
90.07
40.26
265
60.03
80.19
2-0.036
(0.023
)
No.
prio
r cou
rt a
ppea
ranc
es (w
ith p
rove
n off
ence
s)80
57.14
26.69
814
98.45
08.10
765
66.84
56.30
5-1.605
*(0.707
)
No.
prio
r pris
on s
ente
nces
805
1.40
72.75
714
92.31
53.48
865
61.20
12.52
1-1.114
***
(0.302
)
Pane
l C. S
ente
ncin
g ou
tcom
es
Bond
with
sup
ervi
sion
805
0.30
60.46
114
90.27
50.44
865
60.31
30.46
40.03
7(0.041
)
Bond
with
out c
onvi
ctio
n w
ith s
uper
visi
on80
50.02
90.16
714
90.02
00.14
165
60.03
00.17
20.01
0(0.013
)
Bond
with
out c
onvi
ctio
n w
ithou
t su
perv
isio
n80
50.07
10.25
714
90.03
40.18
165
60.07
90.27
00.04
6*(0.018
)
Bond
with
out s
uper
visi
on80
50.07
70.26
714
90.05
40.22
665
60.08
20.27
50.02
9(0.021
)
Conv
ictio
n on
ly80
50.00
40.06
114
90.00
70.08
265
60.00
30.05
5-0.004
(0.007
)
Fine
805
0.03
70.19
014
90.08
10.27
365
60.02
70.16
3-0.053
*(0.023
)
Hom
e de
tent
ion
805
0.00
10.03
514
90.00
00.00
065
60.00
20.03
90.00
2(0.002
)
Impr
ison
men
t80
50.09
70.29
614
90.24
80.43
365
60.06
30.24
2-0.186
***
(0.037
)
No
conv
ictio
n re
cord
ed80
50.00
60.07
914
90.00
00.00
065
60.00
80.08
70.00
8*(0.003
)
No
pena
lty80
50.02
10.14
414
90.05
40.22
665
60.01
40.11
6-0.040
*(0.019
)
Oth
er p
enal
ties
805
0.00
90.09
314
90.00
70.08
265
60.00
90.09
50.00
2(0.008
)
Inte
nsiv
e Co
rrec
tion
Ord
er80
50.01
10.10
514
90.02
00.14
165
60.00
90.09
5-0.011
(0.012
)
Com
mun
ity S
ervi
ce O
rder
805
0.10
80.31
114
90.08
10.27
365
60.11
40.31
80.03
4(0.026
)
Susp
ende
d se
nten
ce w
ith s
uper
visi
on80
50.18
40.38
814
90.08
10.27
365
60.20
70.40
60.12
7***
(0.027
)
Susp
ende
d se
nten
ce w
ithou
t sup
ervi
sion
805
0.04
00.19
514
90.04
00.19
765
60.04
00.19
5-0.001
(0.018
)
Note.N=o
bservatio
ns,rob
uststand
arderrorsinparentheses,p<.00
1***,p<.01
**,p<.05
*.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 20
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Tabl
e A2
. Rob
ustn
ess
chec
ks
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)(9
)(1
0)
Perf
ect m
atch
esAl
l per
iods
All c
ourt
sFu
ll sa
mpl
eCo
urt F
ERo
lling
Prac
tical
ITT
Reoff
endi
ng
cuto
ffPS
M
Pane
l A. P
riso
n-0.103
***
-0.092
***
-0.098
***
-0.097
***
-0.100
***
-0.074
***
-0.087
***
-0.090
***
-0.093
***
-0.113
***
(0.011
)(0.009
)(0.009
)(0.009
)(0.009
)(0.019
)(0.011
)(0.009
)(0.009
)(0.018
)
Obs
erva
tions
92,613
102,29
211
6,08
112
6,93
712
6,83
13,70
252
,219
92,762
91,958
1,31
0
Pseu
do R
20.38
10.38
40.38
30.38
60.40
60.42
10.35
50.38
10.38
1
AUC
0.89
70.89
90.89
80.89
90.90
60.91
30.88
60.89
70.89
7
Pane
l B. R
eoff
endi
ng-0.051
*-0.039
*-0.034
-0.034
-0.022
-0.016
-0.049
*-0.035
-0.039
*-0.037
(0.020
)(0.018
)(0.018
)(0.018
)(0.019
)(0.040
)(0.020
)(0.018
)(0.018
)(0.027
)
Obs
erva
tions
76,041
84,103
94,467
103,43
910
3,43
43,33
543
,366
76,171
75,526
1,23
0
Pseu
do R
20.08
220.08
180.08
050.08
020.08
410.11
50.08
110.08
220.08
22
AUC
0.69
40.69
40.69
20.69
20.69
60.72
70.69
30.69
40.69
4
Pane
l C.
Days
unt
il ne
xt o
ffen
ce70
.331
*66
.501
*56
.500
*59
.192
*44
.453
113.52
192
.165
**61
.055
*61
.866
*63
.402
*
(31.26
0)(29.04
7)(27.73
4)(28.59
5)(28.57
6)(89.38
9)(33.94
7)(27.67
1)(28.21
8)(28.17
2)
Obs
erva
tions
54,474
60,750
67,489
74,433
74,430
2,68
030
,193
54,581
54,359
896
Pseu
do R
20.01
30.01
20.01
30.01
30.01
30.00
10.01
20.01
30.01
2
Cont
rols
YY
YY
YY
YY
YY
Tim
e FE
YY
YY
YY
YY
YN
Cour
t FE
NN
NN
YN
NN
NN
Note.N=observatio
ns,AUC=
AreaUnd
erth
ereceiverope
ratin
gcharacteristicCurve,FE=FixedEff
ects.Amon
goff
ende
rsund
ergoingCS
,colum
n1restrictsth
eestim
ationsampletooffe
nderswith
identicalsentencingdatesinbothRO
Dand
theAS
U
database.Colum
n2allowsoff
ende
rsparticipatinginTSinm
onth-yearswith
outC
Stobeinclud
edinth
eestim
ationsample.Colum
n3allowsoff
ende
rsparticipatinginTSincou
rtswith
CSavailabletobeinclud
edinth
eestim
ationsample.Colum
n4
impo
sesno
sam
plerestrictio
nregardingcourtsorm
onth-years.Colum
n5em
ployscourtFEs.Colum
n6lim
itsth
eestim
ationsampletocou
rtsthatwilleventuallyhaveCS
,and
thencom
paresoff
ende
rsund
ergoingCS
tooffe
ndersun
dergoingTSin
courtsyetto
introd
uceCS
.Colum
n7lim
itsth
eestim
ationsampletooffe
ndersthatpleadguiltyand
weregrantedbailatth
eirfi
rstcou
rtapp
earance.Colum
n8repo
rtsIntention-to-treate
stimates.Colum
n9lim
itsth
eestim
ationsampletooffe
nders
with
abaselinesentencingdateon
orb
efore31
July201
8.Colum
n10
repo
rtsthediffe
rence-in-m
eansbetweenparticipantsund
ergoingCS
and
am
atched
group
ofcon
trols.Theseoffe
nderswerematched
usingprope
nsityscorem
atchingon
theset
ofcon
trolvariablesdescribed
inPanelBofTable3,stand
arderrorsinparentheses,***p<.00
1,**p<
.01,*p<.05
.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 21
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Table A3. Descriptive statistics for reoffending variables
Full sample Traditionally sentenced Circle sentenced Difference
N Mean Std. Dev.
N Mean Std. Dev.
N Mean Std. Dev.
Estimate Std. Err.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Reoffend within 12 months for:
Violent offence 76,303 0.113 0.316 75,688 0.112 0.316 615 0.154 0.362 0.042** -0.015
Property offence 76,303 0.100 0.300 75,688 0.100 0.300 615 0.070 0.255 -0.030** -0.010
Drug offence 76,303 0.043 0.202 75,688 0.043 0.202 615 0.021 0.144 -0.022*** -0.006
Weapon offence 76,303 0.007 0.084 75,688 0.007 0.085 615 0.005 0.070 -0.002 -0.003
Traffic offence 76,303 0.079 0.270 75,688 0.079 0.270 615 0.049 0.216 -0.031*** -0.009
Public order & miscellaneous offences
76,303 0.038 0.191 75,688 0.038 0.191 615 0.039 0.194 0.001 -0.008
Offences against justice procedures
76,303 0.025 0.156 75,688 0.025 0.156 615 0.011 0.106 -0.014** -0.004
Reoffend within 12 months (selected crime)
76,303 0.071 0.257 75,688 0.071 0.257 615 0.042 0.201 -0.029*** (0.008)
Selected violent crime 76,303 0.005 0.070 75,688 0.005 0.070 615 0.008 0.090 0.003 (0.004)
Selected property crime
76,303 0.066 0.249 75,688 0.066 0.249 615 0.034 0.182 -0.032*** (0.007)
Note.N=observations,robuststandarderrorsinparentheses,p<.001***,p<.01**,p<.05*.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH 22
CIRCLE SENTENCING, INCARCERATION AND RECIDIVISM
Table A4. Raw maximum likelihood coefficients
(1) (2) (3)
Prison Reoffend Days
Circle Sentencing -1.348*** -0.183* 0.113*
(0.182) (0.089) (0.047)
SEIFA Q2 -0.083** 0.000 -0.009
(0.030) (0.020) (0.012)
SEIFA Q3 -0.202*** -0.028 -0.022
(0.035) (0.023) (0.014)
SEIFA Q4 -0.220*** -0.018 -0.046*
(0.057) (0.036) (0.022)
Missing SEIFA 0.001 -0.164 0.227
(0.531) (0.272) (0.171)
Major cities 0.066* 0.135*** -0.069***
(0.029) (0.019) (0.011)
Outer regional -0.124*** -0.043 -0.005
(0.037) (0.024) (0.014)
Remote -0.421*** -0.033 -0.000
(0.071) (0.041) (0.022)
Missing Area 1.754*** -0.264 -0.065
(0.532) (0.274) (0.172)
Age 0.010*** -0.031*** 0.016***
(0.003) (0.002) (0.001)
Age at first contact with justice system -0.027*** -0.011*** -0.004***
(0.003) (0.002) (0.001)
Male 0.420*** 0.118*** -0.051***
(0.028) (0.017) (0.010)
No. concurrent charges 1.053*** 0.137*** -0.047***
(0.013) (0.009) (0.005)
MSR of principal offence -0.025*** 0.004*** -0.002***
(0.000) (0.000) (0.000)
Bail at first court appearance -1.171*** -0.250*** 0.079***
(0.037) (0.039) (0.021)
No plea entered -0.876*** 0.327*** -0.144***
(0.030) (0.019) (0.011)
Plead not guilty -0.401*** 0.172*** -0.069***
(0.039) (0.025) (0.014)
No. prior court appearances (with proven offences) 0.041*** 0.081*** -0.030***
(0.003) (0.002) (0.001)
No. prior prison sentences 0.113*** 0.016*** -0.010***
(0.005) (0.005) (0.003)
Constant -1.353*** -0.211* 6.640***
(0.154) (0.105) (0.066)
Observations 92,746 76,159 54,569
Note.Columns1and2reporttherawcoefficientsfromaLogisticregression.Column3reportstherawcoefficientsfromaZero-TruncatedNegativeBinomialregres-sion.Robuststandarderrorsinparentheses,p<.001***,p<.01**,p<.05*.
NSW BUREAU OF CRIME STATISTICS AND RESEARCH - LEVEL 1, HENRY DEANE BUILDING, 20 LEE STREET, SYDNEY 2000
[email protected]•www.bocsar.nsw.gov.au•Ph:(02)83461100•Fax:(02)83461298ISSN2204-5538(Online)•ISBN978-1-925343-75-5
©StateofNewSouthWalesthroughtheDepartmentofCommunitiesandJustice2020.Youmaycopy,distribute,display,downloadandotherwisefreelydealwiththisworkforanypurpose,providedthatyouattributetheDepartmentofJusticeastheowner.However,youmustobtainpermissionifyouwishto(a)chargeothersforaccesstothework(otherthanatcost),(b)includetheworkinadvertisingoraproductforsale,or(c)modifythework.