Author Sara Rahman, Suzanne Poynton and Wai-Yin Wan
Published August 2018
Report Type Crime and Justice Bulletin No. 216
Subject Assault; Offenders; Recidivism / Re-offending; Statistical methods and modelling; Evaluation reports
Keywords Violent offending, re-offending, instrumental variables, regression

Download this publication



To identify the impact of the Violent Offender Treatment Program (VOTP) on re-offending and return to custody outcomes at 24 months of free time post release.


Data were obtained for all offenders referred to VOTP between 2007 and 2014 and released from prison, yielding a sample size of 587 referrals. Ordinary least squares and two-stage-least-squares (2SLS) linear probability models were used to estimate the differences between those who started the program and those who did not, on four outcome variables measured at 24 months free time post release:

1) re-offending with any offence;
2) re-offending with any offence or a return to custody;
3) re-offending with a violent offence; and
4) re-offending with a violent offence or a return to custody.

Similar analyses were also undertaken comparing outcomes for offenders who completed VOTP with those who did not complete the program.


Starting VOTP was associated with significantly lowered probability of general re-offending (by 9 percentage points), general re-offending or returning to custody (7 percentage points) at 24 months free time post release. Similar differences in the probability of general re-offending (9 percentage points) and general re-offending or returning to custody (7 percentage points) were found in relation to completing VOTP. We also find non-significant results for violent re-offending. The latter finding may be related to loss of statistical power due to sample attrition.


VOTP appears to be associated with lower rates of general re-offending and return to custody, however the
estimates obtained are based on a simple regression approach and may not represent causal effects. Replication with more robust techniques and/or a larger sample size is recommended.

Download this publication