Investors Get Paid Only if the Programme Actually Works
A social impact bond funds a social intervention with private capital, repaid by government only if agreed outcomes are achieved. It transfers delivery risk away from the public purse and creates a measurement problem.
The Problem It Addresses
Governments spend heavily on responding to problems and comparatively little on preventing them. Prison, emergency healthcare, and homelessness services are funded because the need is immediate. Programmes that would reduce those needs years later compete for money against them and generally lose.
Part of the reason is budgetary. Prevention costs money now and produces savings in a future budget cycle, possibly in a different department, possibly after the current officials have left.
A social impact bond, also called a pay for success contract, restructures that. Private investors fund the intervention upfront. The government repays them, with a return, only if independently verified outcomes are achieved.
The Structure
| Party | Role |
|---|---|
| Investors | Provide capital upfront, bear delivery risk |
| Service provider | Delivers the intervention |
| Government | Pays only on verified outcomes |
| Independent evaluator | Determines whether outcomes were achieved |
| Intermediary | Structures the transaction and manages performance |
The instrument is not a bond in any conventional sense. There is no fixed coupon and no principal repayment obligation. It is an equity like exposure to whether a social programme achieves a measured result.
The government is not buying a service. It is buying an outcome, and declining to pay for effort that did not produce one. That is genuinely different from ordinary procurement and it is the whole point.
What It Is Supposed to Fix
Three things, and each is a real problem in public service commissioning.
Risk transfer. If the programme fails, investors lose money rather than taxpayers. Government pays only for results.
Evidence. Because payment depends on measurement, the intervention must be evaluated rigorously, which is far more than most social programmes receive.
Flexibility. Providers are paid for outcomes rather than for delivering a specified activity, which permits adapting the approach during delivery.
What Actually Happened
Enough of these have now been executed to assess them, and the honest summary is mixed.
The most cited early failure was a programme aimed at reducing reoffending among short sentence prisoners at a British prison. The intervention did not meet the threshold on the agreed measure and investors lost their capital, which is the structure working exactly as designed. A separate American programme aimed at the same objective similarly failed to reach its target and produced no payment.
Several others achieved outcomes and produced returns, particularly in early childhood education and homelessness, where the interventions had stronger existing evidence.
The pattern that emerges is that these instruments have worked best where the intervention already had good evidence behind it, which raises an obvious question about what the structure added.
The Costs Nobody Advertises
The transaction costs are substantial and are the strongest practical criticism.
Each deal requires legal structuring, an intermediary, an independent evaluator, and negotiation over outcome definitions and measurement methodology, frequently taking years before any service is delivered.
For a programme of modest size, that overhead can consume a meaningful share of the total. Several practitioners have concluded that the structure only makes sense above a scale that few social interventions reach.
The Measurement Problem
Paying for outcomes requires knowing what would have happened otherwise, which is the same counterfactual difficulty that appears in carbon offsets and in demand response.
Rigorous designs use randomised assignment or matched comparison groups. Those are expensive, take time, and are sometimes ethically or practically impossible for the intervention in question.
Weaker designs use historical baselines, which are vulnerable to general trends. A reoffending programme that appears successful during a period when reoffending fell nationally has not demonstrated anything.
The negotiation over the counterfactual is therefore the substantive commercial negotiation, and it is conducted between parties with opposite interests in where the threshold sits.
Where the Model Went
The concept has largely evolved rather than scaled. Elements of it appear in outcomes based contracting without the investor layer, where government simply pays a provider on results and the provider bears the risk.
That removes the transaction cost of assembling investors and removes the risk transfer, since a small provider cannot bear the loss that a fund can.
Development focused variants, where a donor rather than a government pays on outcomes, have continued and address a genuine gap, since a donor funding a programme in another country has limited ability to supervise delivery and paying on verified results substitutes for that.
The Bottom Line
Social impact bonds pay investors only when an intervention achieves measured outcomes, which transfers delivery risk from taxpayers and forces evaluation that social programmes rarely receive. They have worked, they have failed, and the failures functioned as designed with investors bearing the loss. The durable criticism is transaction cost and the difficulty of establishing what would have happened anyway, and the durable contribution is that they made outcome measurement a contractual requirement rather than an afterthought.