Failures and lessons
Failure (stated outcome not achieved; programme ended)
Rikers Island Social Impact Bond: the first US SIB, and the first to pay out nothing
Failure (stated outcome not achieved; programme ended). What was attempted, what happened, what is documented and what can be learned.
- Last reviewed
- Related chapter
- Chapter Four, case 6 (the failure story); Chapter Five, step 6 (milestones and release conditions) and the accountability guardrail
- Sectors
- Criminal justice, youth
- Geography
- New York City, United States
- Vehicles
- Outcomes-based contract; public-private platform
What was attempted
In 2012 New York City launched the first social impact bond in the United States. Goldman Sachs provided USD 9.6 million (reported in some sources as USD 7.2 million; the sources differ, and the library reports both) to fund a cognitive behavioural therapy programme (the Adolescent Behavioral Learning Experience) for 16- to 18-year-olds detained at Rikers Island, delivered by the Osborne Association with Friends of Island Academy. MDRC was the intermediary. Bloomberg Philanthropies guaranteed the majority of Goldman's exposure (USD 7.2 million per MDRC; USD 6 million per Nonprofit Quarterly). The Vera Institute of Justice ran an independent quasi-experimental evaluation.
The deal: if recidivism among participants fell by 10 percent, the city would repay Goldman in full; larger reductions would earn a capped return (up to around USD 2.1 million); a reduction below 8.5 percent would mean no repayment.
What happened
In July 2015 Vera reported that the programme had not reduced recidivism among the young people it served. The programme ended in August 2015, at the close of its scheduled term. The city paid nothing. Goldman's loss was limited by the Bloomberg guarantee; the net exposure after the guarantee has been reported at around USD 1.2 million, though the sources do not state Goldman's loss directly.
What is documented and what is not
Documented: the structure, the parties, the target, Vera's headline finding, the programme's end, and that the city owed no repayment. Not documented in the sources reviewed: the detailed evaluation findings (Vera's full report followed the summary), the precise loss to each party, and whether the programme would have performed differently with a longer run or a different population.
What the structure got right
Philanthropy took the first-loss position (Bloomberg's guarantee), a commercial investor took the senior position (Goldman), an independent evaluator held the parties to a pre-agreed metric, and when the metric was not met, no public money was spent. From the taxpayer's perspective, the structure did what it was designed to do.
What it got wrong, according to its critics
Nonprofit Quarterly's analysis (August 2015) made three points. The structure added costs that are easy to overlook: an intermediary, an evaluator, and substantial in-kind time from city officials, none of which show up in the headline numbers. Second, a single intervention was asked to shift an outcome (recidivism among detained teenagers) that is driven by many things outside the programme's control, and the evaluation could only say that this intervention, in this setting, did not move it; it could not say what would. Third, and most relevant to this book: with Bloomberg guaranteeing most of Goldman's exposure, the claim that SIBs "shift risk to the private sector" was weak. An MDRC report cited by the authors acknowledged that philanthropy may need to "smooth the curve" for commercial investors, which undercuts the premise.
Lesson for catalytic philanthropy
The book's guardrails apply directly. Additionality: Goldman was not taking much risk, so its participation was not very additional; the catalytic capital was Bloomberg's. Proportionality: a guarantee covering most of the senior layer is not a buffer, it is the layer. Accountability: the structure had it, and that is why the failure is clean and documented. A failure that is measured honestly is worth more to the field than a success that is not.
Sources
Educational material, not investment, legal or tax advice. Reported figures and review dates reflect the cited sources and editorial review; they are not live data.
