An analysis of juvenile recidivism (in Juvenile crime and by lindahy


An analysis of juvenile recidivism (in Juvenile crime and

More Info
                                               An Analysis of

                                               M ICHAEL CAIN

                                                            juvenile criminal careers within the study
In 1996, the NSW Department of Juvenile
Justice published a report, entitled
                                                                  These selection processes resulted in a
“Recidivism of Juvenile Offenders in New
                                                            final sample of 91 230 records relating to the
South Wales” 1 , which is based on a study
                                                            proven appearances of 52 935 individual
of 52 935 juvenile offenders who appeared
                                                            juveniles, of whom 43 331 (81.9 per cent)
before the Children’s Court from 1986 to
                                                            were male and 9604 (18.1 per cent) were
                                                            female. It should be noted that these data
     The recidivism report examined the
                                                            are far larger, more detailed, more
pattern and characteristics of juvenile
                                                            representative, and more timely than is
offending and re-offending. Its findings are
                                                            usual in criminal justice research.
a “good news” story. The report destroys
                                                                  Within the final sample, 36 723 records
some commonly held beliefs concerning
                                                            pertain to juveniles who had one and only
juvenile crime and juvenile re-offending,
                                                            one proven appearance, that is, that
and the media, the public and politicians
                                                            number of non-recidivist juvenile offenders.
should not disregard the findings.
                                                            The remaining 54 507 records refer to the
                                                            first and subsequent proven appearances of
The study                                                   16 212 recidivist offenders.

Description of the sample                                   Limitations of the study
The recidivism study examined those                         As is the case with most research, the
records for juvenile offenders who me t the                 recidivism study did have a number of
following conditions:                                       limitations and restrictions:
•    they were first convicted of a criminal                •    the NSW Children’s Court does not
     offence in the Children’s Court on or                       routinely collect information on a
     after 1 January 1986; and,                                  young offender’s family life,
•    they had reached the age of 18 years                        socioeconomic conditions, education,
     by the end of 1994.                                         employment or unemployment, drug
     Expressed another way, the study                            and alcohol use, culture or
captured all juvenile offenders who had                          Aboriginality. These factors, despite
commenced and effectively ended their                            their recognised importance to an
                                                                 understanding of juvenile crime, were
                                                                 not available for analysis;
           This report is available from the NSW
Department of Juvenile Justice, Level 5, 24 Campbell
Street, Sydney 2000.

An Analysis of Juvenile Recidivism

    •    the study was not able to access a                •    males are one-third more likely to re-
         young person’s prior involvement in                    offend than females;
         police cautioning or pre-court                    •    the younger an offender is at first
         diversionary schemes;                                  court appearance the greater the risk
    •    the study only examined juvenile                       of future offending;
         recidivism as there is no easy way to             •    juveniles who commit, as their first
         track juvenile offenders into the adult                offence, a common assault, break and
         justice and correctional systems.                      enter, or motor vehicle theft are more
                                                                likely to re-offend, whereas those that
    Findings                                                    first commit a sexual offence, drug
                                                                offence or offence against good order
    If there are two things we can safely say                   are unlikely to re-offend;
    about juvenile crime they are that:
                                                           •    juveniles given a custodial sentence,
    •    the majority of juvenile offenders will                community service order, or
         not reappear after their first proven                  supervised probation as their first
         offence; and,                                          penalty are more likely to re-offend
    •    it is a small proportion of juvenile                   than juvenile first offenders given
         offenders that accounts for a large                    lesser penalties, such as fines or
         percentage of juvenile offences.                       nominal penalties;
         The recidivism study found that                   •    juveniles from western and eastern
    around 70 per cent of juvenile offenders                    Sydney, the Hunter area, and Western
    had one, and only one, proven criminal                      NSW were more likely to return to
    appearance in the Children’s Court. A                       crime after their first court appearance.
    further 15 per cent had just two proven                     On the other hand, juveniles from
    appearances. That is, 85 per cent of juvenile               southern Sydney, and northern and
    offenders had reasonably limited involve-                   southern country areas of NSW were
    ment with the criminal justice system.                      less likely to re-offend;
    Nevertheless, the study also uncovered
                                                           •    juveniles who were dealt with by a
    some concerning figures, including that:
                                                                non-specialist Children’s Court were
    •    9 per cent of juvenile offenders were                  also more likely to re-offend than
         responsible for almost one-third of all                juveniles who first appeared before a
         criminal appearances; and,                             specialist magistrate.
    •    less than 2 per cent of juvenile                       The recidivism study also identified a
         offenders were responsible for almost             number of general characteristics of juvenile
         10 per cent of all criminal appearances.          crime and juvenile recidivism:
         These findings show that there would              •    persistence in juvenile crime is marked
    be great value in identifying those juveniles               by progressively shorter periods to the
    who are most at risk of re-offending,                       next offence;
    particularly if this could be done at the time
                                                           •    the majority of offences (86 per cent)
    of their first court appearance. This, in fact,
                                                                for which juveniles appear and re-
    is what this study attempted to do.
                                                                appear in the Children’s Court are
                                                                property offences and not crimes of
    The characteristics of juvenile re-                         violence;
    offenders                                                   • juveniles who re-offend, including
                                                           persistent offenders, do not escalate to
    Who are the juvenile offenders who are
                                                           more serious and violent crimes. Even those
    most likely to become repeat and persistent
                                                           juveniles who first committed a violent
    offenders? Logistic regression methods
                                                           offence, when they re-offended, were more
    were applied to the data, and yielded the
                                                           likely to commit a subsequent property
    following findings (see table 1):
                                                           crime. The differential association of these
                                                           factors with re-offending allows a model to

                                                                                                                              MICHAEL CAIN

A model for predicting juvenile recidivism risk
      Example 1: For a 14-year-old boy from the Sydney west metropolitan area given a custodial order by a
children’s magistrate for a first offence of assault (note, this combination of factors is bordering on a worst case
scenario), the predicted risk of re-offending may be estimated by including the appropriate logistic coefficients
in the model:

     Estimated prob (recidivism)      = 1/(1 + e-(6.1449 + 14(-0.4540) + 0.2828 + 0.1777 + 0.2872 + 0.1429 + (-.0636) )
                                      = 0.649
     That is, this juvenile’s estimated risk of re-offending is around 65 per cent.

     Example 2: The predicted risk of re-offending for a 16-year-old female from the south coast of NSW given
an unsupervised recognizance by a local court for an initial drug offence is 18.9 per cent, which is calculated as:

    Estimated prob (recidivism)               = 1/(1 + e-(6.1449 + 16(-0.4540) + 0 + (-0.2007) + (-0.1011) + (-.0342) + 0 )
                                              = 0.189.

be developed to predict recidivism risk at
                                                                                    of re-offending, valuable program resources
time of first court appearance. The two
                                                                                    can be diverted and better directed to
examples above indicate how the logistic
                                                                                    those juveniles identified as being at
regression model may be applied.
                                                                                    greatest risk of re-offending. One may ask,
     Notably, the logistic regression model
                                                                                    what is the point of intervening in the lives
has an overall accuracy of 72 per cent in
                                                                                    of the 70 per cent of young offenders who
predicting which juvenile offenders will or
                                                                                    are unlikely to re-offend anyway? There is
will not re-offend. It is 91 per cent accurate
                                                                                    certainly value, both financial and social, in
in predicting which juveniles will not re-
                                                                                    reducing the number of young people who
                                                                                    are unnecessarily involved in the juvenile
The beauty of the recidivism risk model is                                          justice system.
not only its impressive degree of predictive                                              Thirdly, targeting high-risk juveniles at
utility, it has other benefits.                                                     first court appearance will mean not only a
      Firstly, because it is based on factors                                       more rational and economic use of juvenile
which are determined through the court                                              justice resources. It will also mean that many
process, there is good reason to believe that                                       troubled youth, in need of help, will have
many of the important social factors not                                            their problems and issues identified and
contained in the model have been                                                    addressed much earlier in the process. Early
considered in arriving at the sentencing                                            intervention by justice and welfare agencies
decision. Expressed another way, the                                                into the troubled lives of young offenders
penalty a juvenile first receives is a proxy                                        and their families is the preferred strategy
for a whole range of personal and social                                            for attempting to address the juvenile crime
conditions which affect a young person’s                                            problem.
criminal propensity, and thus the sentence                                                It is important to remember that it is a
they receive. Children’s Courts regularly                                           small proportion of juvenile offenders who
request a background report on a juvenile                                           are responsible for a large percentage of
prior to sentencing. These reports may flag                                         offences. The identification of these high
problems in a juvenile’s home life or                                               risk juveniles at their first court appearance,
education, and other things such as their                                           and selectively targeting these young
association with known offenders, drug                                              offenders with programs and services to
and alcohol problems, and so on. Courts                                             stop or minimise their further offending will
generally will consider these factors in                                            have immense social and economic benefits
arriving at their sentencing decision.                                              for the people of NSW.
                                                                                          Furthermore, there is every indication
Secondly, because the model identifies non-
                                                                                    that the findings of this research are equally
recidivists as well as those at increased risk

An Analysis of Juvenile Recidivism

    applicable to juvenile justice administrators                          and their families, the less effective such
    in other States and Territories. All govern-                           interventions are likely to be in changing
    ments must recognise that the later they                               antisocial attitudes and criminal behaviours
    leave their attempts at addressing the social                          that are at risk of becoming entrenched.
    and personal problems of young offenders
        Table 1: Logistic regression analysis (main effects modela) of the relationship between juvenile recidivism
                     and sex, age at first proven CA, first offence, first penalty, place of residence and type of court,
                     all juveniles (n=52,935), Children’s Court, 1986 to 1994

        Juvenile recidivism (DV) b
        Independent variables (IVs)                 Β                 SE                Sig           Ratio             99% CI

        Age 1st                                 -.4540             .0073              .0000

        (Male)                                   .2828             .0144              .0000            1.33           1.28 - 1.38

        First offence                                                                 .0000
          Serious person                        -.0393             .0482              .4145            0.96           0.84 - 1.08
          Robbery                                .1113             .0765              .1453            1.12           0.92 - 1.36
          Sexual                                -.3719             .0982              .0002            0.69           0.61 - 0.78
          Assault                                .1777             .0395              .0000            1.19           1.08 - 1.32
          Drug                                  -.2007             .0482              .0000            0.82           0.72 - 0.93
          Break & enter                          .1868             .0289              .0000            1.21           1.12 - 1.30
          Steal motor vehicle                    .2141             .0317              .0000            1.24           1.20 - 1.28
          Theft                                  .0342             .0256              .1816            1.03           0.97 - 1.11
          Justice & good order                  -.1121             .0269              .0000            0.89           0.83 - 0.96

        First penalty                                                                 .0000
          Custodial orders                       .2872             .0713              .0001            1.33           1.11 - 1.60
          CSOs                                   .2022             .0648              .0018            1.22           1.04 - 1.45
          Supervised orders                      .2258             .0297              .0000            1.25           1.16 - 1.35
          Fines                                 -.2260             .0252              .0001            0.80           0.73 - 0.87
          Unsupervised orders                   -.1011             .0331              .0000            0.90           0.85 - 0.96
          Nominal penalties                     -.3881             .0255              .0000            0.68           0.64 - 0.72

        NLGA - place of residence
        (Departmental region)                                                         .0000
          Eastern Sydney                         .0794             .0258              .0020            1.08           1.03 - 1.14
          Western Sydney                         .1429             .0271              .0000            1.15           1.08 - 1.24
          Southern Sydney                        .0452             .0244              .0641            1.05           0.98 - 1.11
          Hunter                                 .1112             .0264              .0000            1.12           1.04 - 1.20
          Northern NSW                          -.0142             .0307              .6446            0.99           0.91 - 1.07
          Western NSW                            .1614             .0335              .0000            1.18           1.08 - 1.28
          Southern NSW                          -.0342             .0273              .2096            0.97           0.90 - 1.04
          Interstate/unknown                    -.4917             .0406              .0000            0.61           0.55 - 0.68

        Court Type
        (Specialist Children’s Court)           -.0636             .0126              .0000            0.94           0.91 - 0.97

        Constant                               6.1449              .1170              .0000

    a       The model does not contain an analysis of interaction terms because of limitations in computer processing capacity.
    b       Juvenile recidivism is a dichotomous variable.
    c       Age 1st is a continuous variable. Sex, First offence, First penalty, NLGA, and Court Type are categorical variables.
             CI = Confidence Interval.


To top