Topic 28 of 64· Governmental performance and delivery metrics
Payment by Results and Social Impact Bonds (PbR/SIBs)
Payment by results (PbR) pays a provider based on verified outcomes achieved, not activities performed. A social impact bond (SIB) is a specific PbR financing structure in which private or philanthropic investors fund service delivery up front and are repaid — with a return — by a government commissioner only if independently measured outcomes hit agreed thresholds, shifting the delivery risk from the taxpayer to the investor.
Why it matters
The world's first SIB launched at HMP Peterborough in September 2010: Social Finance raised £5 million from 17 investors to fund the "One Service," working with short-sentence prisoners (under 12 months) to cut reoffending, with the Ministry of Justice and the Big Lottery Fund agreeing to repay investors only if reconviction events fell by at least 7.5% against a matched national comparison cohort. The final cohort of the Peterborough pilot recorded an 9.7% reduction in reconvictions, comfortably above the threshold, and investors were repaid with a return. The mechanism mattered because it solved a specific commissioning problem: government wanted to pay for outcomes rather than inputs, but couldn't absorb the financial risk of an intervention that might not work, so the SIB structure moved that risk onto investors willing to underwrite it. The Government Outcomes Lab (GO Lab) at Oxford's Blavatnik School of Government now maintains the most complete public evidence base on PbR and SIB performance worldwide, tracking well over 200 impact bonds globally and publishing the research on what design features correlate with success or failure. The lesson the evidence base repeatedly returns to is that the outcome metric chosen, and who bears the risk of missing it, determines almost everything else about how a PbR contract actually behaves in practice.
The math
PbR payment = base payment (if any) + Σ (outcome achieved × unit price per outcome)
Social impact bond investor return:
Investor outlay = upfront capital funding service delivery
Outcome payment = commissioner pays only if outcome ≥ threshold, scaled by
how far above threshold performance lands
Investor return = outcome payments received − investor outlay
(a rate of return, often capped, reflecting risk taken)
Key design parameters that determine the whole contract's behavior:
Outcome metric — must be an outcome, not an output (see outcomes-vs-outputs)
Comparison/counterfactual — usually a matched cohort (see counterfactual-analysis)
Payment threshold — minimum improvement before any payment triggers
Payment curve — linear, stepped, or capped above the threshold
Attribution/deadweight discount — see additionality-and-deadweight
Worked example
Peterborough One Service (illustrative figures drawn from published evaluations):
Investor capital raised: £5,000,000
Cohort: ~3,000 short-sentence male prisoners over two cohorts
Threshold: ≥7.5% reduction in reconviction events vs matched
national comparison group, or no payment
Cohort 1 result: 8.4% reduction — below the contractual bar for that
cohort alone under the original rules
Combined/final cohort result: 9.7% reduction — above threshold
Outcome payment: government (Ministry of Justice / Big Lottery Fund)
pays per percentage point above threshold, funding
investor repayment plus a return
Local authority PbR contract (illustrative): a family-intervention service is commissioned at £4,000 per family referred (activity payment) plus £6,000 per family with no further child protection referral 12 months after closure (outcome payment). 200 families referred, 150 cases closed, 96 remain referral-free at 12 months:
Activity payment = 200 × £4,000 = £800,000
Outcome payment = 96 × £6,000 = £576,000
Total contract cost = £1,376,000 for 96 confirmed sustained outcomes
Cost per confirmed outcome ≈ £14,333 (see cost-per-outcome)
Software engineering connection
Payment by results is an incentive-alignment problem before it is a data problem, and the data system is where that alignment either holds or breaks. Independent, tamper-evident outcome verification is the whole game: the commissioner and provider have opposing incentives on how an ambiguous case gets coded, so the system recording outcomes needs an audit trail, a data-sharing agreement with the independent verifier (often a different body from the provider, sometimes an official statistics body matching against police or benefits records), and immutable versioning of the outcome definition — the PbR equivalent of the "redefining the metric" pitfall in public sector KPIs. Attribution calculations depend on counterfactual-analysis matched-cohort methods, which need reproducible, auditable code, not a one-off spreadsheet. And the metric itself must be a genuine outcome, not a proxy activity — see outcomes vs outputs — because a PbR contract paying for an output just relabels business-as-usual funding with extra transaction cost. Where a SIB's social return is being modeled prospectively, that appraisal typically borrows directly from social return on investment methodology.
Pitfalls
- Paying for an easily-gamed proxy outcome: "attendance at sessions" is an activity dressed as an outcome; insist on a measure that reflects the actual change sought (reoffending, employment, housing stability).
- No credible counterfactual: without a matched comparison group, an improvement could be regression to the mean or a wider trend, not the program's effect — see counterfactual-analysis and additionality-and-deadweight.
- Underestimating transaction and evaluation costs: independent verification, data linkage, and contract administration for PbR/SIB schemes routinely run into double digits as a percentage of contract value — GO Lab's evidence base documents this as a recurring driver of scheme discontinuation.
- Cherry-picking or "parking": providers paid per outcome have a direct incentive to prioritize clients most likely to succeed anyway and deprioritize the hardest cases — design payment tiers or case-mix adjustment to counter it.
Sources
- Government Outcomes Lab, University of Oxford, Blavatnik School of Government. https://golab.bsg.ox.ac.uk/
- Social Finance, "Peterborough Social Impact Bond" evaluation summaries. https://golab.bsg.ox.ac.uk/case-studies/peterborough-social-impact-bond/
- Ministry of Justice, "Peterborough Social Impact Bond HMP Doncaster: Final Reconviction Results for the Peterborough Social Impact Bond."