Brief 35: Reducing Reconvictions Among Released Prisoners

This study evaluated a Risk-Need-Responsivity (RNR) evidence-based offender supervision program, Citizenship, using a randomized controlled trial. Overall, there was a non- significant 20% effect of Citizenship in reducing reconvictions. However, controlling for risk, the hazard with higher risk offenders was 34% lower than for the control group. Results therefore support RNR-based probation supervision.

Link to Full Study

Category: Public Service Provision

Tags: behavior, corrections, probation, stepped wedge cluster design, rehabilitation

Date of Publication: Thursday, January 7, 2016

EGAP Researcher: David Torgerson

Other Authors: Dominic A.S. Pearson, Cynthia McDougall, Mona Kanaan, Roger A. Bowles

Partners: Photo credit: Flickr user Oinkyliciously, available via Creative Commons BY-NC-ND 2.0

Geographical Region: Europe

Research Question:

Primary: Does the Citizenship probation supervision program in the U.K., effective in reducing reconvictions?

Secondary: Does the programme's effectiveness vary by the risk-level of the offenders?

Preparer: Alicia Cooperman

English

Background:

Probation supervision is a crucial element of the criminal justice system, yet little is known about the effectiveness of different programs. This study evaluates the Citizenship program of probation supervision, which is based on the Risk-Need-Responsivity (RNR) principles. The principles adjust features of the intervention to each offender’s profile: the intensity is based on risk of reoffending, the content is based on individual needs, and the delivery style is based on the learning capabilities and responsiveness of the offender.

Most prior evaluations of probation and parole supervision programs have not used rigorous research design techniques. This paper is among the first studies to use a randomized controlled trial to study supervision in the U.K. criminal justice system.


Research Design:

To study the effectiveness of the Citizenship program on offenders of different risk profiles, the researchers used a stepped wedge cluster randomized controlled trial (RCT). The Citizenship program was introduced to six offices in the new probation area of Teeside, U.K. during 2007-2008. The sample encompassed 1,091 offenders, with 395 assigned to Citizenship and 696 assigned to control.

The stepped wedge cluster design includes a sequential roll-out of the program to each office, where the order of receiving treatment is randomly assigned. At first, all offices were in the control condition. After two months, the first randomly-selected office received the treatment of implementing the Citizenship program, with only newly sentenced/released offenders entering the treatment. Two months later, the second randomly-selected office received the treatment, with the design continuing at two-month intervals until all six offices had been treated. Each office was considered a “cluster,” where the treatment was assigned to the entire office at once.

This type of design is most effective where interventions are known to be beneficial, and a prior study by the same authors found that the Citizenship program reduced reoffending in a different area using an observational study (i.e. without random assignment). The stepped wedge design addresses many ethical considerations of RCTs in social science, since all subjects receive treatment eventually. In addition, it is rarely feasible to introduce a program across all units simultaneously, so the randomization of roll-out is both practical and analytically rigorous.

 

Results:

Overall, the Citizenship group had a 20% lower reconviction rate than the control group, though the effect was not statistically significant. However, the effect varied by risk profile of the offenders. Among high risk offenders, the Citizenship group had a 34% reduction in reconvictions relative to the control group. The analysis used statistical models to measure whether offenders had reconvictions and the amount of time before each subject reoffended.

It is important to note that the treatment effects presented are based on “intention to treat” – those assigned to the Citizenship program, even if they did not comply with the treatment. This is the case where the Citizenship intervention was not fully implemented and the prior system of probation continued. In the end, the researchers found that only 38% of those assigned to Citizenship actually complied with treatment. This suggests that the reported treatment effects may be underestimating the impact of the program among who comply with it, though the results are representative of the real-world implications that must consider the challenges of implementing probation programs. Low compliance levels were likely due to insufficient training (length of training and follow-up) and possible practitioner resistance to implementing a randomized control trial.

 

Policy Implications:

The Citizenship probation supervision program, which used Risk-Need-Responsivity (RNR) principles in the U.K. criminal justice system, was most effective among high risk offenders. While it had a 20% reduction in reconvictions overall, there was a 34% reduction among high risk offenders.  The differences in effectiveness between the two risk groups, whilst suggestive of a differential effect were not statistically significantly different.  

The study highlights two areas of consideration going forward:

→ Need for better policy support.

Thorough training and clinical supervision follow-up is necessary to increase compliance rates when implementing a new and time-intensive program. In addition, the analysis should be based on “intention to treat” and not just those who complied, in order to accurately reflect the broader social impacts of this type of program.  Although it is important to note that implementation of the programme in a neighbouring areas was less effective because implementation was undertaken at the same time.  The stepped wedge design as well as allowing for a rigorous evaluation also allowed for better implementation.1

→ Need for randomized control trials (RCT) in criminal justice work.

Observational studies are prone to bias from unobserved factors out of control of the researchers. For example, a prior study by the same group used observational data, with the challenge of separating the reduction in reconvictions due to the program from nationwide reductions during the same time period. Studies based on voluntary selection into new programs are also biased, since the offenders and probation officers who seek out new interventions are also likely to be the type of people who are more motivated to reduce reconviction.

Instead, with a RCT, the control and treatment groups are measured contemporaneously, and the only difference on average between the groups is the random assignment of the Citizenship treatment. The results from this RCT provide rigorous evidence of the effectiveness of the Citizenship program in reducing reconvictions, especially for high-risk offenders.

  • 1. Pearson D, Torgerson D, McDougall C, Bowles R.  Parable of two agencies, one which randomizes.  Annals of the American Academy of Political and Social Science 2010; 628: 11-29.