Aim: To investigate whether completing the Custody-based Intensive Treatment (CUBIT) program for moderate to high risk/needs sex offenders reduces re-offending.
Method: The study sample includes 386 male offenders identified by custodial staff as suitable for participation in CUBIT who were released from NSW custody during the period 2000 to 2010. An Instrumental Variables (IV) approach is adopted to control for potential unobservable selection bias in treatment outcomes (a key concern in the treatment effects literature), with the CUBIT completion rate amongst commencements around the time of an offender’s own potential participation in CUBIT employed as the preferred IV. This method is statistically inefficient, and has limited power to assess the impact of treatment on sex re-offending (which is relatively rare) specifically. As such, three separate two-stage least squares models (including a range of control variables) are used to estimate the impact of completing CUBIT on whether or not an offender will re-offend with a sex, violent, or general offence, respectively, within five years of free time following their release from custody.
Results: Within five years free time following release from custody 12 per cent of offenders who completed CUBIT committed a proven sex offence, 27 per cent re-offended with a new violent offence and 41 per cent committed a proven offence of any type. Multivariate models controlling for a range of important observable risk factors found that CUBIT completers had, on average, a 5-year general recidivism risk that was 13 percentage points lower than a similar cohort of offenders suitable for CUBIT but who did not participate. No significant differences between the treated and untreated groups were found for violent or sex re-offending.
Conclusion: There is some evidence to suggest that completing CUBIT results in a considerable reduction in general recidivism risk. No evidence is found to suggest that CUBIT completion reduces sexual or violent re-offending. However, it is difficult to draw any conclusion from the null results, since the power of the statistical methods employed to detect a treatment impact in this study is limited given the relatively small sample size.