A Social Impact Bond is a way to get private investment behind new public programs with the potential to save governments money or otherwise produce a positive social result. This funding mechanism makes the most sense when governments want to adopt large-scale initiatives that are backed by a solid body of evidence. A government entity and a private investor agree beforehand on the types of outcomes that will constitute success, and then the private investor funds the program’s delivery and operation. When a reliable third-party evaluation confirms that the services delivered the agreed-upon outcome, the government reimburses the investor. The mechanism had been used previously in the United Kingdom, but the MDRC project was the first time it had been tried in the United States.

In this case, the goal was to reduce the recidivism rate among adolescents (those 16 to 18 years old) incarcerated at Rikers Island. Historically, nearly half of these young people return to Rikers within a year of their initial release, and the typical adolescent released from Rikers Island will spend more than 200 additional days in jail over the next six years. Because incarceration is so expensive — costing the City of New York more than $85,000 per inmate per year — cutting the recidivism rate could save the city a great deal of money. It could also reduce crime, improve the lives of those young people, and lessen jail’s corrosive effect on them, their families, and their communities.

The study evaluated a program called the Adolescent Behavioral Learning Experience (ABLE). It was part of the Young Men’s Initiative launched by former Mayor Michael Bloomberg, which aimed to improve the lives of young men of color. Under the negotiated terms for the Social Impact Bond, Goldman Sachs was to invest $9.6 million over four years in the ABLE program, of which $7.2 million was guaranteed by Bloomberg Philanthropies. If the project had reduced recidivism by 10 percent, the city would have paid back Goldman Sachs that $9.6 million; if it had reduced recidivism by more than that amount, the city would have paid an additional return on a capped, sliding scale. Even taking those additional payments into account, reductions that beat the 10 percent target would have still saved the city significant amounts of money.