Thursday 24 May 2012

Experiment on prisoners!

National's announced what sounds like a decent measure to reduce alcohol-related crime: better rehab treatment in prisons for offenders.
Budget 2012 will contribute to a 25 per cent reduction in reoffending by 2017, and 18,500 fewer victims of crime every year from 2017, Corrections Minister Anne Tolley and Associate Corrections Minister Dr Pita Sharples say.
The moves are part of the Prime Minister’s expectations for a more efficient and results-driven public service.
A boost in alcohol and drug treatment, alongside increased education, skills training and employment programmes for prisoners, including remand prisoners, will lead to safer communities and better value for money for taxpayers.
From 2017, there will also be 600 fewer prisoners in jail than in 2011, and 4,000 fewer community offenders.
“It’s time to get serious about breaking this vicious cycle of prison and reoffending,” Mrs Tolley says.
There have been a few stories out over the last few years about lack of availability of treatment options for offenders who have wanted to seek treatment; increasing availability is likely to help those offenders. But it would likely be wrong to extrapolate from results achieved by those seeking treatment to those that could be achieved among those who would be compelled to seek treatment. So while I'm not convinced that Tolley's projections around reductions in reoffending are right, it still seems a policy worth trying.

Even better, it's a policy possibly worth trying as a randomized control trial. If the total amount of funding available isn't sufficient to give drug and alcohol rehabilitation treatment to everyone they might wish to have it, or to provide employment and reintegration support to all prisoners leaving prison, randomize who gets to participate. Here's one potential approach.

Set up three groups for each type of intervention. The first is a control group - no treatment. The second is compelled to participate, but they get a lotto. Those wishing not to have treatment can ask for it, and some of those wishing to opt out will be able to opt out. The third gets a lotto: those wishing to select into treatment can ask for it, and some of those opting for it will get it.

What does this kind of design let you do?

1. What's the value of treatment for those who want to have treatment? Compare outcomes for those who indicated they want treatment but didn't get it with outcomes for those who won the lottery.

2. What's the value of treatment for those compelled to have treatment? Compare outcomes for those who are forced into treatment against their lottery-expressed wishes with those who are allowed to opt-out of compelled treatment.

3. What's the effect of changing the default option? Compare average outcomes between the two lottery treatments.

4. What's the average effect of compulsory treatment? Get the average rate of "wanting to be treated" across groups 2 & 3, the effect of treatment on the "want to be treated" groups in 2&3, the average rate of "not wanting to be treated" across 2 & 3, the effect of treatment on that group, take the weighted average outcomes across both, and compare to the control group. This would be what we'd expect as effect of a blanket "must be treated" rule.

It's great that the government's looking at targeting one of the real sources of alcohol and drug related external cost. It would be even better to set it up so we could learn something!

Alas, the Ministry of Justice and the Corrections Minister would likely have a hard time getting this done under the auspices of any of the universities due to likely quibbling from Human Ethics Review panels. But I'm sure there'd be a few folks around who'd find it interesting enough to do as a side project.

No comments:

Post a Comment