Constructing baselines
The MoJ has been busy working on a number of fronts to take forward the Transforming Rehabilitation competition.
Last Thursday, 10 October, the MoJ published its latest thoughts on the payment mechanism for new providers of probation services.
The next day they also published a rather technical report describing:
“the approximate size, composition and re-offending rates of offender cohorts over time in each of the 21 new Contract Package Areas (CPAs)”.
Accompanying the report, the MoJ published modelled data for the 6 years from 2005 to 2010, showing the number of offenders in each PbR cohort and the 1-year re-offending rates among those offenders.
The report provides an historical picture of probation performance in reducing reoffending aimed at those organisations interested in winning the new probation contracts.
So, it presents performance on a Contract Package Area, rather than Probation Trust, basis.
And it looks specifically at the group of offenders for whom the new Community Rehabilitation Companies will be responsible.
The method
The MoJ has sought to include those on short custodial sentences who will receive supervision and support the first time under Transforming Rehabilitation (although there is no explanation as to how this part of the cohort was calculated, presumably the MoJ used court data).
These are a very important group since their overall binary reoffending rate is much higher than those supervised on community sentences and a proportion are likely to be prolific offenders, with a high frequency of reoffending too.
The short sentence prisoner group is modelled as comprising a quarter (25.2%) of the overall population of offenders that CRCs will be responsible for, although this proportion varies markedly from 14% of the total caseload in Northumbria to 36% in London.
The MoJ has tried to exclude high risk offenders from its modelling (since this group will be the responsibility of the National Probation Service).
High-risk offenders are not easily identifiable, therefore the MoJ has used a proxy indicator for these offenders – they define this group as “anyone who had, at any time in the past, been convicted of a serious offence” (the overall figure been adjusted to reflect the proportion of offenders expected to be defined as high risk under the new arrangements).
It has also excluded:
- Offenders sentenced to a stand-alone electronic monitoring
- Offenders sentenced to stand-alone committee payback
- Foreign nationals subject to deportation
The data
The tables provided with the report (see the bottom of this page) provide information about the number of offenders who would have been in the payment by results cohort for each CPA, broken down into men and women, and also into the three categories of:
- Community Sentence/Suspended Sentence Supervision Order
- Custodial sentence less than 12 months
- Custodial sentence 12 months or more
The MoJ has calculated how many offenders would be expected to re-offend and the total number of expected re-offences.
The MoJ also models each CPA’s actual reoffending performance against their predicted performance.
The predicted rate takes into consideration the profile of offenders in the current cohort. Over recent years, the probation caseload has been increasingly made up of offenders with a high risk of reoffending. This is one of the reasons that comparing historical reoffending rates can be misleading.
Therefore measuring actual reoffending against predicted reoffending provides a fair measure of performance.
By comparing the binary rate (column 3) with the average OGRS score (column 6) in the table below you can see whether CPAs performed better or worse than expected.
Using these figures:
Staffordshire & West Midlands has performed best with a binary reoffending rate of 39.8% against an expected rate of 43%.
London has performed worst with a binary reoffending rate of 45.3% against an expected rate of 41.2%.
Perhaps the most important issue in regards to Transforming Rehabilitation is where the MoJ will choose to set the baseline for caculating reoffending rates – and, of course the associated payments under the PbR system.
I presume they might refine this modelling exercise and update it some time in 2014, although, to date, MoJ has made no announcements about what the baseline will be.
One Response
Fact is criminals are individuals not cohorts!
Statistics have their uses but ultimately EACH person who the state requires to change is more likely to make positive changes if they are individually assessed and engaged in a constructive relationship with one person who takes a personal responsibility for stimulating and aiding the changes to be made.