The area of youth offender risk assessment and treatment is one that has received a lot of attention, both from the public and from correctional authorities. Violent crime by delinquent youth has increased markedly from the 1980s, with juvenile offenders often described as a vulnerable, multi-problem population, with high co-occurring rates of psychiatric; substance abuse, and abuse issues (Loeber & Farrington, 2000).

Youth have the highest rate of recidivism compared to other age brackets with 24.5% being reimprisoned within six months after release and 71.5% reimprisoned within five years (Spier, 2002). Violence amongst youth is also on the rise. Spier (2002) found that the number of youth apprehended for violent offences increased by 21% between 1994 and 2001 and approximately a quarter of all proven cases in New Zealand involving youth are violent (Spier, 2002).

Loeber and Farrington (2000) identified that a number of developmental pathways existed for delinquent youth, with most attention being paid to those with early onset offending. A younger age at first conviction has been linked to an increased risk for violent recidivism (Loeber & Stouthamer-Loeber, 1996; Moffitt, 1997). Generally, offenders who begin their criminal careers earlier and are introduced to the justice system at a young age are more likely to commit further acts of violence and criminality than those who become criminally active later in life. A large number of studies confirm the link between early onset and chronic criminal behaviour, including the Dunedin longitudinal study, which established persistent antisocial behaviour prior to age 13 as a key risk indicator (Moffitt, 1993). In another long-term study of criminal behaviour using a sample of 282 male aboriginal offenders, Bonta, Lipinski, and Martin (1992) found that criminal recidivists had a significantly younger mean age at first conviction (17.8 years) than non-recidivists (19.5 years). Moreover, in a sample of 322 male inmates followed-up from 1973, Martinez (1997) found that an offender’s age at his first arrest was predictive of future criminal activity. Finally, Lattimore, Visher, and Linster (1995) further identified age at first arrest as being a significant risk predictor for future violent crime, using multivariate competing hazards analysis to identify salient risk predictors for violent recidivism among young offenders.

The development of risk assessment stems from the need to identify those individuals who have a higher risk of future offending, this is so that precious resources can be allocated to where there is the most need and thus make the most impact on reducing crime. This has become what is known as the ‘best practice’ approach and guides intervention towards needs that reduce the risk of recidivism. There is a growing body of literature around identifying chronic adult offenders so that intervention strategies may reduce the risk of reoffending (Andrews & Bonta, 1998). However, by the time an adult is identified as a chronic offender they have already amassed a large and devastating criminal history.

In a sample of 120 inmates released from a maximum-security psychiatric institution, Villeneuve and Quinsey (1995) found that repeat violent offenders had a substantially greater history of serious juvenile delinquency than non-recidivists. In addition, Bonta, Law and Hanson’s (1998) meta-analysis revealed that juvenile delinquency correlated strongly (r = .27) with violent recidivism. Gendreau, Little, and Goggin (1996) also found that a history of pre-adult antisocial behaviour was highly predictive (mean weighted r = .16) of general recidivism. Further documentation of the importance of early behaviour to later offending comes from Rice and Harris (1996) who examined several predictors of violent recidivism in a sample of 243 mentally disordered fire setters, finding several variables reflecting childhood antisocial behaviour was a significant predictor of violent recidivism.

There is therefore a need to apply the advancements made in risk assessment to youth offender populations, as it is among the delinquent youth of today that we are likely to find the chronic offenders of tomorrow.

Actuarial versus clinical assessment of risk

In keeping with the ‘best practice’ approach identified by Andrews and Bonta (1998), there was a need to be able to accurately identify those individuals with a higher risk of future offending in order to enable a focus on addressing the dynamic criminal risk factors that maintain recidivism risk. Bonta (1996) reviewed the literature on offender risk assessment and identified that up until the 1980’s the majority of risk assessment had relied on clinical judgement which had never been empirically validated. In fact, it was found that clinicians were susceptible to judgement errors, stereotypical biases, and cognitive heuristics in making decisions about risk. Grove and Meehl’s (1996) meta-analysis found that clinical judgement outperformed actuarial approaches (summation of factors that related to recidivism) in only 6% of their sample of 136 studies. A further study by Bonta et al. (1998) found that clinical judgement had a low correlation with violent recidivism compared to actuarial risk measures (clinical, r = .09; actuarial, r = .30). It is noted that the majority of actuarial measures were heavily weighted for static predictors such as prior criminal history rather than dynamic predictors such as employment or marital status (Andrews & Bonta, 1998). In light of the poor ability to predict risk based on clinical judgement there is a need to establish reliable actuarial measures of risk for youth offenders.

Validated risk measures for youth offenders

One of the first actuarial measures was developed by Burgess in 1928. He looked at 3,000 parolees and created an index that combined twenty-one items for the purpose of predicting parole success or failure. In 1972 the federal parole board of the United States started using the Salient Factor Score as a part of their decision making guidelines (Hoffman, 1994). The problem with this and similar kinds of instruments such as the General and Statistical Information on Recidivism (GSIR) and the Violence Risk Appraisal Guide (VRAG) was that they contained mainly static items (Loza, 2003). This meant they were limited in their scope by been unable to guide rehabilitative interventions or assess an individual’s change in risk over time. This lead to a new generation of instruments which included both static and dynamic factors (Andrews & Bonta, 1998). These are generally risk/need assessments and are based on the social learning theory of criminal conduct (Andrews & Bonta, 1998). There are a number of instruments like this which are used for a variety of purposes. For example: The Wisconsin model developed by Baird in 1981 (Andrews & Bonta, 1998); and the Historical, Clinical, and Risk Management violence risk assessment scheme (HCR-20) (Douglas, Ogloff, Grant, & Nicholls, 1999). The most widely documented is the Level of Service Inventory (LSI) and more recently the Level of Service Inventory Revised (LSI-R). The literature has documented reliability and validity in a number of different areas including general offenders, juvenile, sex offenders, and violent offenders (Gendreau, Little & Goggin, 1996; Loza, 2003). Its purpose is to aid in the decision making process regarding supervision requirements for probation or parole. It is comprised of 54 items that assess 10 areas of risk/need: criminal history, education/employment, financial, family/marital, accommodations, leisure/recreation, companions, alcohol/drugs, emotional/personal, and attitudes/orientation.

In 1999 the LSI-R was modified to focus more specifically on risk and need factors of juveniles. What resulted was the Youth Level of Service/Case Management Inventory (YLS/CMI). This new assessment comprises of 42 items that uses essentially the same risk domains as the LSI-R: Prior and Current Offences; Family Circumstances and Parenting; Education/Employment; Peer Relations; Substance Abuse; Leisure/Recreation; Personality and Behaviour; Attitudes and Orientations (Flores, Travis & Latessa, 2003). Flores, Travis and Latessa (2003) looked at the validity of the YLS across three different settings: The Ohio Department of Youth Services – a custodial juvenile institutional programme; The Butler County Juvenile Rehabilitation Centre – a residential programme that provides treatment services to adjudicated youth; and the Clermont County Juvenile Probation Department. They found that it significantly predicted which youths were at a higher risk of recidivating using a number of different recidivism measures including program completion, institutional violations, technical violations, re-arrest, re-arrest seriousness, and reincarceration (Flores, Travis & Latessa, 2003). They also found that it was significantly related to all of the recidivism measures, the strongest of which were with technical violations and re-arrest (Flores, Travis & Latessa, 2003).

Considering that the YLS was developed from the LSI-R which has a large amount of literature attesting to its validity and utility within the correctional setting, the YLS has a great deal of promise for use within the New Zealand correctional system for guiding the management of offenders, especially because of its emphasis on risk/needs the offender. It is important though that the YLS be validated on a New Zealand Juvenile population and compared with other important prediction instruments in order to fully evaluate its effectiveness within the New Zealand context.

Another important actuarial measurement that has received an enormous amount of attention since its development by Hare in 1980 is the Psychopathic Checklist (PCL). This is a rating scale designed to measure the personality construct of psychopathy. It is based on two different constructs that define a psychopath: Factor one characteristics consist of personality traits such as Manipulative and deceitful, glib and superficial, lack of empathy, and egocentric and grandiose (Hare, 1993); Factor two characteristics are the social deviant behaviours that consist of being impulsive, the need for excitement, having poor behaviour controls, a lack of responsibility, early behaviour problems and adult antisocial behaviour (Hare, 1993). The checklist consists of 20 characteristics that are rated on a 3 point scale and are added to give a total score that ranges from 0 to 40 (Catchpole & Gretton, 2003). In order to conduct this assessment the examiner must review records and administer a lengthy interview and must be specially trained in administering the instrument. This makes it a complex and time consuming instrument to administer (Murrie & Cornell, 2002). However, it has demonstrated a high predictive ability compared to other measures in a number of different correctional settings for a variety of different purposes (Loza, 2003). Like the LSI-R, the PCL-R has also been revised to accommodate a juvenile population. In the Psychopathic Checklist-Youth Version (PCL: YV) a number of items have been changed to reflect the different presentation of psychopathic traits between adolescents and adults (Kosson et al., 2002). Kosson, Cyterski, Steuerwald, Neurmann, & Walker-Mathews (2002) found that the PCL: YV had high internal consistency and inter-rater agreement in a community adolescent sample. They also found that it predicted antisocial behaviour, childhood psychopathology, interpersonal behaviours associated with adult psychopathy, and a lack of attachment to parents (Kosson et al., 2002).

New Zealand risk measures

Corrections Policy Development has supported Dr Graham Scott (2002) in developing an instrument to provide a measure of risk that used four sections (30 items): Initiating events before 2 years of age (e.g., early abuse or neglect), Early behaviour disorder 4-10 years (e.g., inattentive, hyperactive), Early delinquent behaviour 8-12 years (e.g., fighting, shoplifting), and Early criminal offending 10-15 years (e.g., criminal offences). The measure now titled Risk Serious Youth Offending (RSYO) allows the estimation of chronic versus acute recidivism profiles. One unique aspect of the instrument is the inclusion of a crosscheck interview with a nominated significant other who knew the youth offender. The instrument in initial validation research (research version called the SOR-Y, 36 items) has found using Factor Analysis that one factor overwhelmed the six possible factors, namely, Childhood behavioural disorders (aspects of ADHD, ODD, or CD). Predictive and convergent validity are still to be established for the RSYO, however, initial results indicate that the instrument should identify a higher risk group based on static developmental risk factors.

The Risk of Reconviction/Risk of Reimprisonment Scale (RoC/RoI) was developed in 1994 for use within New Zealand offender populations (Bakker, O’Malley, & Riley, 1998). Its purpose is to assess the probability of a particular offender incurring new convictions based on statistically significant risk factors. It consists of static items which are weighted and added up to produce an overall measure of the risk of reconviction. This instrument assists in assigning inmates to one of five categories of risk so that their sentence and rehabilitation is planned appropriately. Bakker, O’Malley, and Riley (1999) plotted the predicted probability of reconviction against the actual proportion in the sample who were reconvicted. The plot shows a high linear relation, indicating high accuracy of the reconviction model. The reimprisonment model was also represented, this plot also showed a high linear relation, until predicted probabilities reached 80% chance of reimprisonment, after that the model over-predicts the proportion of offenders reimprisoned after five years. The authors put this fluctuation down to smaller amounts of individuals within these higher groups (Bakker et at., 1998).

Overall, there are four actuarial measures that were chosen for the purposes of this study: The PCL: YV, the YLS, the RSYO, and the RoC*RoI. The PCL: YV and the YLS were chosen because they are both actuarial measures based on well researched adult instruments; they have both demonstrated good initial psychometric properties on juvenile samples overseas; and both need to tested on a representative juvenile sample in order demonstrate how effective they would be in guiding management and intervention strategies within a New Zealand context. Not to mention that the PCL: YV is based on personality factors, whereas the YLS is risk/needs based. The RSYO is also an instrument based on the risk/needs principle, however, it has recently been developed within a New Zealand context and its psychometric properties have yet to be tested. And, the RoC*RoI has been included because it is based on static, historical factors and is already currently used for the management of offenders in New Zealand.

Strategic benefits from this research

  • This will be the first research that has sought to gather information on a New Zealand incarcerated youth offender population from actuarial measures containing both static and dynamic risk factors. This will allow the Department to develop a risk profile of youth offenders within the Young Offender Units to enable targeted intervention to be developed.

Research Objectives

  1. To establish the risk profile of a representative sample of youth offenders from Young Offender Units
  2. To collect, analyse and report on three actuarial measures of risk containing developmental, static, and dynamic factors.
  3. To provide a report on the risk profiles of the study participants and the convergent validity of the RSYO for Policy Development. It is hoped that information on criminogenic needs and responsivity issues for the sample will be used to make policy recommendations for managing youth offenders based on the results from the study