3. The Montgomery House violence prevention programme
The Montgomery House violence prevention programme began in 1987 as a joint project between the New Zealand Department of Corrections and the New Zealand Prisoners’ Aid and Rehabilitation Society. The initial design, development and early implementation of the programme are documented in detail by Dixon and Wikaira (1988), Dixon and Polaschek (1992) and Polaschek and Dixon (2001). In 1994 the programme was thoroughly reviewed following concerns about the quality and integrity of the programme delivery. The programme was restructured in accordance with the review recommendations (Berry, 1995) and recommenced after a short hiatus in late 1994. In brief, the programme is an 8-week group based intervention established upon social learning and cognitive behavioural principles (e.g., Andrews & Bonta, 1988; Cullen & Gendreau, 1989). Montgomery House is unique in that the programme structure and delivery strongly reflect traditional kawa (Māori protocol and customs). Broadly based on the Te Whare Tapa Wha model (Durie, 1994) the programme emphasises balance across a number of personal dimensions. For example, the programme specifically seeks to address te taha tinana (physical), te taha hinengaro (psychological), te taha wairua (spiritual), and te taha whanau (familial) needs of all residents (Montgomery House, 1994). The majority of referrals to Montgomery House are for Māori offenders. For example, in the study by Berry (1999) reported below, 81% of the treatment group were Māori, 13% Pacific Island, and 6% Caucasian.
Outcome results presented in the current discussion relate to the first 10 programmes that were run following the 1994 review (January 1995 to December 1996) and are as documented in Berry (1999) 1. A total of 64 residents successfully graduated from the 10 Montgomery House violence prevention programmes run during 1995 and 1996. These 64 men were matched to a control sample on a large number of relevant variables (e.g., ethnicity, age at first violent offence, number of previous violent and non-violent offences, amount of time spent in prison, maximum seriousness of offending, probability (risk) of re-conviction). There were no statistically significant differences between the treatment and control groups on any variables, although it was not ultimately possible to determine if any of the control group had had any other forms of treatment (Berry, 1999).
The treatment group was followed-up for a mean of 17 months before re-conviction data was gathered (the control group was followed-up for only slightly shorter – mean of 16 months). During this period the treatment group (N=64) accumulated 33 violent convictions compared to 51 for the control group (N=64) 2; . Treatment completers thus registered 35.3% fewer convictions for violence than the control group. However, somewhat paradoxically, re-convictions for non-violent offences were 22% higher amongst the treatment group compared to the control group (despite being matched on risk variables). Berry (1999) suggests this result is because of the higher percentage of the control group who were re-imprisoned and hence not at liberty to commit non-violent offences for the remainder of the follow-up period. A more thorough evaluation of recidivism is provided by the survival analysis results (Berry, 1999). As previously noted, survival analysis allows variable periods of follow-up to be statistically controlled for. It also allows computation of statistics on the speed of re-conviction. When adjusted for rate of pre-programme violence, survival analysis showed that 25% of the treatment group had been reconvicted of a further violent offence compared to 42% of the control group after a mean of 16-months at liberty. In terms of the speed of violent re-conviction, treatment completers were significantly slower to re-offender than the matched control sample (Cox’s F-Test, p=.011).
Berry (1999) also provides a comparative analysis of the performance of the 18 individuals who dropped out of treatment (treatment non-completers) and their matched controls 3. The 18 treatment non-completes accumulated 127% more violent convictions than their matched controls during the 16-month follow-up period. While the actual figures presented are small (25 violent convictions for the 18 treatment non-completers verses 11 violent convictions for their 18 matched controls) the same trend of treatment non-completers performing worse than all other groups (both completed treatment groups and non-treatment control groups) is evidenced in re-offending seriousness statistics, speed to re-offending, and survival analysis (Berry, 1999).
In terms of the overall effect size of the Montgomery House violence prevention programme, McLean and Grace (1998) used a conservative approach (in that they combined the treatment completed and treatment not-completed groups verses a matched control group), which yielded an effect size of r=+.12. This effect size is comparable to international findings on the treatment of adult violent offenders which, in the case of the Dowden and Andrews (2000) meta-analysis report a mean effect size of r=+.07 from 35 separate studies.
In a further investigation of the effectiveness of the Montgomery House violence prevention programme, Wilson (personal communication, 2 April 2002) has followed up 60 of the original Berry (1999) sample for a total of 6 years. Wilson has reported that preliminary results indicate a difference of 20 percentage points between the violent recidivism rates of the treatment and control groups after 6 years (58% violent recidivism by the treatment group verses 78% by the control group). These results provide further support that Montgomery House is effective at reducing the rate of violent recidivism and that the treatment effect remains stable even after an extended period of time at liberty.
1 Other evaluation data relating to earlier programmes (programmes completed between 1987 and 1989) is thoroughly documented by Polaschek and Dixon (2001).
2 Note that these data represented just the frequency of reconvictions, and for example do not necessarily equate to 33 individuals re-offending, just 33 re-offences (by an unknown number of individuals).
3 Non-completer controls were matched on the same variables used to compare the completers and their controls.