Author Imogen Halstead and Suzanne Poynton
Published July 2016
Report Type Crime and Justice Bulletin No. 192
Subject Evaluation reports; Recidivism / Re-offending; Statistical methods and modelling
Keywords alcohol and other drug treatment, re-offending, propensity score matching, impact evaluation, legislative evaluation

Download this publication



To investigate whether referral to the Intensive Drug and Alcohol Treatment Program (IDATP) reduces re‑offending and/or returns to custody.


The study sample included 1,285 offenders who were released from NSW custody on or after 1 January 2013, all of whom satisfied IDATP eligibility criteria. An intention-to-treat (ITT) design was employed to protect against unobservable selection bias in program commencement and completion, with the treatment group comprising the complete sub-sample of 340 offenders who were referred to IDATP (59.4% of whom commenced treatment). Propensity score matching is used to construct matched samples of treated and untreated offenders across a range of observable characteristics related to recidivism risk. Separate matchings were constructed for offenders with follow-up periods of 3, 6, and 12 months. Rates of re-offending and/or return to custody were compared for the matched samples of treated and untreated offenders.


IDATP referrals were younger than the similarly eligible offenders in the comparison group, had a more extensive criminal history, and had been classified according to official risk assessment measures as being at higher risk of recidivism including for reasons related to their drug and/or alcohol use. Nonetheless, after matching on the estimated propensity score, the samples of offenders were similar across these and a broader range of observable characteristics. Although rates of re-offending and/or return-to-custody were lower for the matched treatment than matched comparison group at follow-up periods of 6 and 12 months, there was no statistically significant difference in recidivism outcomes.


The power of the statistical analyses was severely limited by the small sample of IDATP participants and the ITT research design. The power to detect a treatment impact will improve as more offenders engage with IDATP over time.

Download this publication