In 2010, the world’s first Social Impact Bond was launched at Her Majesty’s Prison Peterborough. Social impact bonds (SIBs) are an innovative way of financing social interventions. Under a SIB, private investors pay for the intervention, and then receive payments based on the results achieved.
This first SIB funded The One Service, which aimed to reduce prisoner reoffending by offering client-led ‘through-the-gate’ support including, befriending, helping clients to secure housing, manage mental health problems and provide skills training. Payback to the investors would depend on the reduction in reconvictions, measured over 12 months. Payment would be triggered if reconvictions were reduced by 7.5 per cent across two cohorts of prisoners. Additionally, early payment could be triggered by a 10 per cent reduction in either of the individual cohorts.
A 2014 report showed an estimated reduction in reoffending of 8.4 per cent. Today, the cohort two report has been published, written by Dr Jake Anders, Senior Research Fellow at UCL and Professor Richard Dorsett, Professor of Economic Evaluation at University of Westminster. They estimated a reduction in reoffending of 9.7 per cent. Together with the previous published result this implies a reduction in reoffending across both cohorts of 9.0 per cent; sufficient to trigger payment.
Professor Dorsett argues that The One Service is an important case study providing several insights into the role of evaluation and the design of SIBs.
In policy evaluation, there is typically an acceptance that impact estimates are subject to some degree of uncertainty, and that different samples will yield different results. However, when used to rule on payments, such uncertainty becomes more of a problem. Variation that might arise purely by chance can mean the difference between hitting a target and missing it. Even with the same sample, evaluation choices can be important. To illustrate this, Anders and Dorsett describe their inability to precisely replicate the cohort one results, despite using the same approach.
With stakes high, the need for transparency and openness is paramount. The Ministry of Justice, who commissioned all the reports, took great care to maintain this by withholding information on reoffending outcomes until the point at which it could no longer influence any evaluation design decisions.
SIBs transfer risk to investors and, chiefly, this risk hinges on the unknown effectiveness of the intervention. In practice, impact must be estimated and that introduces additional uncertainty. Effects that should trigger payment may be missed and equally, weak effects may wrongly be estimated as strong enough.