#721 2010 · Social Finance (Peterborough Prison, 'One Service') · Criminal justice / public finance
No one would fund a rehab program for short-sentence prisoners that might not work, so investors fronted the cash and got paid only if reoffending actually fell
问题
A rehabilitation program that might not work can't get funded upfront
背景
Prisoners released after serving under 12 months in the UK received almost no statutory support at all, and national reoffending among that group ran around 60%. Everyone involved could name programs likely to help — intensive, personalized support through the first months after release — but government wouldn't commission them: an unproven intervention is a budget risk no official wants to own, and a normal grant or service contract pays out whether or not the program actually reduces reoffending. Charities that believed in such programs had no way to prove the model at scale, because proving it required money nobody would advance without proof.
Social Finance, a UK nonprofit, proposed breaking the deadlock by moving the risk off the public budget entirely rather than trying to argue government into taking it on. In March 2010 it signed a contract for what became known as the world's first Social Impact Bond: private investors, not the state, would supply the working capital, and the state would only pay if an independent evaluator confirmed the program actually worked against a real counterfactual.
换别人会怎么做
Government had two ordinary levers — fund the program as a normal grant or contract, which pays out regardless of whether reoffending actually falls, or decline to fund an unproven intervention at all, which is what had happened for years and is exactly why short-sentence prisoners had no support to begin with. Neither lever lets an official commit public money to something that might not work.
他们看到了什么
Social Finance saw that the obstacle wasn't a lack of belief in the intervention, it was that ordinary financing forces the payer to accept the execution risk before results exist. Separating who fronts the capital from who eventually pays for it — investors take the risk, government pays only for a verified outcome — let each party do the thing only they were actually positioned to do: investors could tolerate losing money on an experiment, and government could commit to paying for a result it never had to bet on in advance.
那一手
Seventeen investors, mostly charitable foundations, put up £5 million to fund Social Finance's 'One Service' — intensive support delivered by St Giles Trust, Ormiston Families, YMCA and other providers to three planned cohorts of 1,000 prisoners released from Peterborough after short sentences. The Ministry of Justice agreed to repay investors their capital plus a return only if reconviction events fell by 10% per cohort, or 7.5% across all cohorts, measured against a matched national control group of similar offenders rather than against Peterborough's own past performance.
为什么管用
Because investor capital, not public budget, is at risk during delivery, government can authorize an unproven program without owning the downside if it fails, which removes the single biggest reason such programs never got funded. Because repayment is contingent on an independently measured result against a genuine control group rather than the program's own before/after story, the payer is buying a verified outcome rather than an intention, which is also what let investors demand — and get — a real return once the 9% reduction actually cleared the bar.
值了多少
The program ran two of its three planned cohorts before a separate national policy overhaul, Transforming Rehabilitation, extended payment-by-results contracting to all short-term prisoners nationally and eliminated the ability to hold a clean control group, ending the SIB early. On the results actually measured, the Ministry of Justice's independent evaluation found a 9% reduction in reoffending against the control group, beating the 7.5% target, and in 2017 investors were repaid their full capital plus a return of roughly 3% a year.
什么时候会失灵
The mechanism needs an outcome that can be defined narrowly and measured against a trustworthy counterfactual — Peterborough only worked because a matched national control group existed to compare against, and the program's own early termination shows how fragile that requirement is: a single surrounding policy change that disrupts the control group can end the contract regardless of how well the intervention itself is performing. It also needs investors patient enough to wait years for repayment and willing to lose their capital outright if the outcome isn't hit, and it costs more to administer than a plain grant, which is why the model has stayed a niche instrument rather than a general replacement for public funding.
后来呢
Peterborough is credited as the first Social Impact Bond in the world, and the structure was replicated into roughly 89 SIBs across 19 countries with combined investment over £300 million by the time Peterborough's own results were announced. It also became the model's own cautionary case: the very policy instability the structure was meant to work around is what cut the pilot short, and critics have since pointed to SIBs' administrative complexity and cost relative to a plain grant as a real ceiling on how far the mechanism travels.
资料来源
- [1]World's first social impact bond sees resultsGovInsider, 2017govinsider.asia
- [2]Peterborough social impact bond investors repaid in fullCivil Society Media, 2017civilsociety.co.uk
- [3]HMP Peterborough (The One Service) — INDIGO Impact Bond DatasetGovernment Outcomes Lab, University of Oxford, 2017golab.bsg.ox.ac.uk