Nine percent reduction in re-offending
It’s hard fully to remember now what a big deal the ONE Service – — the Peterborough prison resettlement scheme, led by St Giles Trust — was when it was launched in 2010. Much heralded as the world’s first Social Impact Bond (SIB), it promised a new funding model which would repay investors for achieving outcomes on a payment by results basis.
The SIB was intended to invest £5m in reducing reoffending in three tranches of 1,000 prisoners who were leaving Peterborough Prison after serving sentences of less than 12 months. It was intended to operate over seven years. However the SIB was abandoned after the second tranche, because of the launch of the government’s new probation model, Transforming Rehabilitation, which involved PbR contracts to reduce reoffending among all short-term prisoners, and removed the ability to judge the SIB against a control group.
At the end of last month (27 July 2017), the government published the outcomes from the second (and final cohort), some 12 months late and with no accompanying press release, despite positive findings.
You can see details of the ONE Service model here and of the structure of the SIB here.
The report, mainly consists of very technical details on the propensity score matching approach used to ensure that prisoners at Peterborough were compared to similar (in terms of demography and offending profiles) prisoners elsewhere so that the analysis of reoffending outcomes is reliable. (Personally, it seems bizarre to me that the MoJ has commissioned a number of different reconviction studies of the Peterborough SIB when it could just have asked its in-house Justice Data Lab to do the work which would have guaranteed that the reoffending outcomes would be directly comparable to almost 200 other interventions which the JDL has already analysed.)
Under the terms of the SIB, investors are paid according to how successful the One Service is in reducing reconviction events (one or more offences committed at the same time). If a minimum threshold of a 7.5% reduction in reconviction events was reached across the pilot, payment would be triggered. Additionally, there was an option to trigger an early payment if a 10% reduction in the number of reconviction events in individual cohorts was achieved.
Here’s the official finding:
We estimate that the One Service reduced the number of reconviction events among those discharged from HMP Peterborough by 9.7% for cohort 2. The reduction across both cohorts is estimated to be 9.0%, which reached the minimum threshold of 7.5% across all cohorts. This reduction is sufficient to trigger an outcome payment.
A nine percent reduction in reoffending is clearly an extremely positive outcome, with the slight caveat that one would expect a very positive finding when comparing a well-respected intensive and holistic intervention with short term prisoners elsewhere who received no resettlement support.
The lack of fanfare surrounding the results can be attributed to the fact that the ONE Project was supposed to be a model for resettlement work at every prison post-Transforming Rehabilitation. In effect, the new contracts provide very few funds and no incentives for the new private probation providers to deliver resettlement. This led the probation inspectors to conclude in a recent damning report that the new “through-the-gate” services had made no impact at all. No-one doubts the effectiveness of the Peterborough approach but the local Community Rehabilitation Company could not afford to commission it.
Looking at the broader perspective, the SIB model has been widely adopted, with 89 examples in 19 countries worth over £300m. It has been praised because it provides social bodies with freedom to innovate, allows the development of long-term solutions and early intervention, and transfers risk away from charities and public commissioners. But it also remains controversial because it is seen as complex, bureaucratic and expensive. SIBs have also struggled to create sufficiently accurate measures to assess whether social issues are being addressed, and are sometimes criticised for commodifying social problems.
Payment by results also has a very mixed experience so far with the PbR elements of the Work Programme and Troubled Families programmes contracts widely accepted to be poorly designed. As far as we know, the PbR elements of the new probation contracts for the 21 Community Rehabilitation Companies remain in place, although any payment details are shrouded in secrecy owing to “commercial sensitivities”.
Given the current poor performance of the CRCs, I wouldn’t be surprised if the PbR elements of their contracts are quietly put to bed. Whether the public is informed about this sort of change to the contracts is unclear although what happens next is almost sure to be a litmus test of the government’s oft-repeated commitment to transparency.
My personal view is that while the payment by results approach can be effective — focusing on outcomes and allowing new approaches without imposing burndensome reporting requirements — funding new approaches via Social Impact Bonds can often be inefficient. I compare SIBs with Private Finance Initiatives, the state gets money up front but still has to pay up in the long term. In the case of Peterborough, approximately 5% of the overall funding of £5m (£250k) went to Social Finance to organise the legal requirements of the funding structure, construct the payment mechanism etc. and the MoJ had to pay a 3% per year premium return to the investors (who in this case were a range of philanthropic trusts). This leads me to question why the MoJ did not just contract directly with St Giles Trust with a requirement that any success dividend (the 3% per year) be reinvested in the next iteration of the scheme? To be fair, in a SIB PbR scheme the investors bear all the risk and had the scheme not reduced reoffending by the agreed target, it would have cost the MoJ nothing and the investors would have borne the costs. So perhaps SIBs should be reserved for genuine new approaches to entrenched social problems.
What a waste of time and money
Good analysis Russell – the St Giles model creates a lot of good practice learning opportunities. Recovery and mentoring is the way forward. I’d also be interested to know if socio-economic factors of the geographic areas “liberated to” were in the assessment model. The social and economic determinants of crime are important factors. There’s a big difference in being liberated to Glasgow and Glastonbury.
Thanks for your comment Richard. I’m not sure how wide Peterborough’s catchment area is, but certainly people from a number of different areas. Not just crime patterns locally but availability of accommodation etc. are important factors. But to answer your question, the matching was not done by area but by demographics and offending profiles and history.
Hi Russell, Thanks for this analysis… Have a couple of questions which I can’t seem to find answers to and wondered whether you could shed any light?
1) What number of individuals was the 5 million spent on? i.e. what was the unit cost of an intervention which brought down the rate by 9.7%?
2) Was this measuring the frequency of offenses only? (If so, does that answer your JDL question, which I believe is only set up for the binary measure?)
3) If this was about frequency of offending, was there any weighting used? Or did an “offense” score 1 – regardless of whether is was rape or possession of cannabis?
Thanks again for your work on this and also the blog in general – a great resource for our sector… Best, Nathan
Thanks for your kind words about the blog.
There were originally supposed to be 3 x cohorts of 1,000. The report doesn’t make it clear how many Peterborough prisoners received the ONE service although there were only 2 cohorts in the end. I don’t know whether the £5m was reduced accordingly.
Frequency of reoffending was measured as number of reconviction events in 12 months follow up but there was no weighting of offences. The Justice Data Lab is able to do this.
St Giles Trust or Social Finance are best placed to answer detailed questions.
Useful blog as always Russell. I just wanted to highlight that the Justice Data Lab does routinely provide frequency of offence reoffending analysis, as well as other metrics such as time to offending. It was the case that at the start of the JDL, only rate of reoffending was provided, but this has been updated now- see the latest reports https://www.gov.uk/government/statistics/justice-data-lab-quarterly-statistics-july-2017
This makes me wonder whether the JDL was not used because the MoJ or perhaps the investors wanted independence for the results- you could argue there would be a conflict of interest for MoJ to provide the reoffending statistics that determined whether it should pay out £5million plus!
Thanks Tracey. I’ve updated my previous reply to clarify that the JDL does measure frequency of reoffending. I think the decision not to use the JDL may well be about independence and also historical timing but it does seem important to use them in the future to reduce costs and encourage spread of best practice.
With the prisoner earning and spending money in a Prison I have published on You Tube Money Cat Planner.
This I feel would be of help and would be pleased to explain further if asked.
The prisoner does not need to have literacy or numeracy skills.