Social Impact Bonds: Overview and Considerations CLASP

Acknowledgements: CLASP's work on Social Impact Bonds as a financing mechanism for human services was funded by the John D. and Catherine T. MacArthur Foundation. We thank them for their generous support but acknowledge that the recommendations in this brief are those of the author alone, and do not necessarily reflect the opinions of the funder. To sign up for CLASP updates, please visit: www.clasp.org Social Impact Bonds: Overview and Considerations March 7, 2014 By Elizabeth Lower-Basch One of the hottest topics in human services is “pay- for-success” approaches to government contracting. In this era of tight budgets and increased skepticism about the effectiveness of government-funded programs, the idea that the government could pay only for proven results has a broad appeal. And those who have identified prevention-focused models that have the potential to improve long-term outcomes and save the government money are deeply frustrated that they have been unable to attract the funding needed to take these programs to scale. Some advocates for expanded prevention efforts are confident that these programs could thrive under pay for success and see such an approach as a way to break out of the harmful cycle where what limited funds are available must be used to provide services for those who are already in crisis, and there are rarely sufficient funds to pay for prevention. While performance-based contracting has existed for years in a range of human services areas — including job training and placement, welfare-to- work activities, and child welfare — pay for success, and in particular, the version referred to as a “social impact bond” (SIB) has drawn a great deal of attention at all levels of government in recent years. The Obama Administration has already carved out funding to support pay-for-success models in both workforce and ex-offender programs— and in the 2014 and 2015 budgets, proposed a $300 million fund at the Treasury to support state SIB initiatives as well as specific pay-for-success activities in the areas of job training, education, criminal justice, and housing.1 While only a couple of SIBs are currently underway in the United States, at least 14 states are currently at various stages of exploring SIBs in domains including criminal justice, health care, and early childhood education.2 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 2 Because of this high level of interest, many policymakers, practitioners, funders and advocates may need to respond to the question of whether a SIB would be a good way to expand funding for a particular intervention or population in a given state or community. This paper provides background information and a framework to help answer this question. It is based on CLASP’s review of the literature on SIBs, as well as on our extensive knowledge of the literature and experience with performance measurement systems, performance- based contracting, and strategies to link public policy and implementation with research evidence for programs serving low-income and other disadvantaged populations. What is a Social Impact Bond? A “Social Impact Bond” or SIB is a new way to finance the expansion of prevention-focused social services that are expected to save government money in the future. There is not a single agreed-upon definition for a SIB, but certain elements are common across most of the descriptions that are in use, as well as in the initial efforts to implement SIBs. Elements of the Prototypical SIB One of the most comprehensive definitions of a SIB is offered by the Nonprofit Finance Fund (NFF) on its Pay for Success Learning Hub. NFF states “in this SIB’s model, an intermediary organization (sometimes described as a Social Impact Bond- Issuing Organization or SIBIO) raises capital from private investors to fund multi-year delivery of preventative or early intervention social service programs traditionally funded by government agencies on an annual basis. If social service providers are successful in achieving contractually agreed targets for performance and achievement outcome metrics, the government pays the investors, through the SIBIO, a return on their investment. This return on investment is funded from the savings produced in the population receiving the preventative or early intervention services by comparison to a defined population that has not. If the outcome targets are not achieved, the government does not pay.”3 This definition helps identify several characteristics of the prototypical SIB that distinguish it from other 100% performance-based contracts:  Private investment: Under a SIB the up- front funding to support the services is provided by non-governmental private investors, who will be repaid, with a bonus (or return on their investment), if the desired outcomes are achieved and who bear the risk of losing their investment if these outcomes are not achieved. Under other entirely performance-based contracts, the service providers themselves must cover the up-front costs of the intervention and bear the risk of non-performance.4 The investors in a SIB could be profit-motivated entities, philanthropic organizations, or a mixture, but many of those promoting SIBs have emphasized the potential to bring in new funding from profit-motivated investors.  Focus on prevention or early intervention: Because SIBs are funded by investors with long-term horizons for receiving repayment, they are able to support on a pay-for- performance basis, preventive activities that will not achieve the desired outcomes for several years. As the service provider does not have to float the up-front costs, the period before outcomes are measured can be longer than is generally possible with a performance-based contract. For example, several current SIBs are looking at recidivism rates by ex-offenders over a multi-year follow-up period. By contrast, under existing 2  Figure 1: Social Impact Bonds: Overview and Considerations P. 3 Source: Laura Callanan, Jonathan Law, Lenny Mendonca, From Potential to Action: Bringing Social Impact Bonds to the US, McKinsey & Company, May 2012, http://mckinseyonsociety.com/downloads/reports/Social-Innovation/McKinsey_Social_Impact_Bonds_Report.pdf. Reprinted with permission. 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 4 performance-based contracts, workforce programs typically make performance payments at the point of job placement and after 3 or 6 months of job retention; longer term outcomes are rarely tracked.  Multi-year service delivery: In spite of the focus on prevention, there are still limits to how long investors will wait for repayment. An analysis by McKinsey & Company of the market for SIBs suggests that the longest period a SIB could reasonably run before payments began is 4-6 years, with 2-3 years of service delivery and 2-3 years of follow- up, which represents a balance between the desire to capture the effects of programs over time and the need to begin repayments to investors in a timely manner.5 If the initial results are promising, investors may be willing to support additional years of services. Even 2-3 years of consistent funding would be an improvement for many service providers, who are currently subject to year-to-year fluctuations in funding. However, whether a SIB would provide more certainty to the service providers than standard contracting depends on the specifics of the arrangement between the intermediary and the service providers.6  Impact measurement: NFF’s definition explicitly states that performance will be measured based on a comparison between the outcomes of participants and those of a comparison group that has not received the services. This implies that SIBs should only pay for impacts, or the effects of the program on participant outcomes, compared to what would have happened in the absence of the services. In other performance-based contracts, payments have almost always been made based solely on the outcomes achieved by the participants without regard to whether the outcomes without the investment might have been. For example, welfare-to-work providers are typically paid based on each participant who enters work, even though some of these participants presumably would have found jobs on their own if they had not received any services. Even though the distinction between outcomes and impacts is well known in the evaluation literature7, the cost of formal evaluations and the need to provide quick feedback for both payment and program management purposes has typically prevented programs from using impacts as performance measures.  Cashable savings: In NFF’s prototypical SIB, the government undertaking the contract will not have to increase expenditures when the performance payments come due, because the governmental savings that have accrued as a result of the program’s performance will exceed the amount owed to the investors. This is only true if the savings are “cashable,” meaning that expenditures actually decrease, and the agency that must make the payments can access these savings.8 This may not be possible even for a highly cost-effective program if the savings are dispersed among different agencies or different levels of government or if the savings would take a longer period to accrue than investors are willing to wait. Other organizations have suggested that SIBs may be an appropriate tool even if the cashable savings do not cover the performance payments, or for funding programs that are beneficial to society but do not result in net governmental savings.9 This issue is discussed in more detail later.  Intermediary: In the prototypical example of a SIB, the government’s pay-for- performance contract is not directly with the organization(s) that provide the services- but 4  with an intermediary organization, which takes on much of the program management role, including recruiting investors, and selecting service providers and overseeing their work in order to ensure that the desired outcomes are achieved. The intermediary’s contracts with the actual service providers may be traditional fee-for-service, pay-for- performance, or some hybrid of the two. In the stylized example, the government’s contract with the intermediary does not specify exactly what services will be provided in order to meet the desired outcomes but gives the intermediary the power to contract with a blend of service providers and to make mid-course corrections if the services are not performing as desired. SIBs in Practice While many SIBs are under discussion, only a few are far enough along that they can be compared to this list of characteristics. The first SIB to be developed is the One Service project, which began providing services to ex-offenders being released from short (under 1 year) sentences at Peterborough Prison in the United Kingdom in 2010.10 While the Ministry of Justice has released preliminary data suggesting that the One Service project is having positive impacts,11 the project has not yet reached the first point at which impacts are formally measured and the first payments will potentially be made. The first U.S.-based SIBs — one in New York City to provide cognitive behavior therapy to young offenders in the Rikers Island jail and one in Salt Lake City to expand the availability of preschool for low-income and at-risk children — are at even earlier stages of implementation. Projects in juvenile justice and services for chronically homeless individuals in Massachusetts and workforce services for ex- offenders in New York are just launching now. Other projects are even earlier in the planning process. As shown in the overview of SIBs in Table 1, these early SIBs all share most of the features of the prototypical SIB but differ in some respects from the idealized description. Some of the differences that are worth noting:  So far, up-front funding for SIBs has mostly been provided by foundations, or by private investors backed with a guarantee from a philanthropic source, rather than by purely profit-motivated investors. If this continues to be true, it will constrain the potential of SIBs to bring in new funding, although SIBs may alter the relationship between philanthropy and service providers. However, some have suggested that for-profit investors will be more interested in SIBs when they have more of a track record. This is a question to pay attention to as SIBs go forward.  The U.S. SIBs are all based on a specified intervention, rather than giving the intermediary the flexibility to meet the desired outcome goals through whatever means it chooses. In the two Massachusetts examples, moreover, the state selected the providers directly through a competitive request for responses process, rather than leaving it to the intermediary to select them. The chosen organizations were given the opportunity to negotiate over the model and contract provisions.  It is difficult to determine from the materials that have been released whether the payment schedules for these SIBs are in fact based on cashable savings. According to MDRC, the payment schedule for the Rikers Island SIB is based on a projection of cashable savings, with most of the benefits accruing if the need for jail beds can be reduced by at least 100 beds.12 However, the cost that the UK Social Impact Bonds: Overview and Considerations P. 5 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 6 government determined it would pay for reduced incarceration in the Peterborough SIB was not based solely on cashable savings from having fewer trials and fewer weeks of imprisonment, but also included an estimate of the broader societal benefits of reduced crime.13  Of the SIBs examined, only the two that received federal funding from the U.S. Department of Labor — the New York State SIB for transitional jobs for ex-offenders and the Massachusetts SIB for juvenile justice services — are using a full experimental design with random assignment to determine the payment points. Peterborough is evaluating the project by comparison to a matched cohort from other prisons, while New York City and Utah are comparing outcomes to historical performance. (Details on the evaluation plan for the Massachusetts homelessness services project have not been released) At this point, it remains to be seen whether these differences between theory and practice are the result of the novelty of the approach, and will go away over time, or if these differences will persist. For example, in the prototypical SIB, the contract between the intermediary and the government specifies the outcomes to be achieved for a population and how performance will be measured, but does not require the use of a pre-determined provider or services. In the U.S. examples, however, the contracts have specified service strategies and providers. Neither governments nor funders have the experience with SIBs to give them confidence in an open-ended approach. It is not yet known whether they will develop such confidence with more experience with SIBs. What are the potential benefits of a Social Impact Bond? Proponents of SIBs have suggested they offer a wide range of potential benefits.14 Among the benefits that have been claimed for SIBs are that they will:  Save money, by increasing investments in prevention.  Expand available resources, and thus break the cycle in which investment in prevention gets squeezed out by current needs.  Allow providers increased flexibility to provide customized interventions to respond to client needs, with more time to demonstrate outcomes.  Embed rigorous evaluation and focus on outcomes into program operations, increasing learning, and ultimately performance. Given that no SIBs have yet reached the payout stage, it is not yet possible to assess whether these potential benefits will materialize. Moreover, because of the differences between the first round of SIBs and the prototypical model, and because the early implementers have additional start-up costs (such as the need for policymakers and contracting staff to learn about the model), it is possible that these particular projects will not have all the benefits that are promised for SIBs, even if the underlying model is sound. In these cases, one of the benefits of the project may be simply building the capacity and knowledge base for future efforts. 6  Social Impact Bonds: Overview and Considerations P. 7 Table 1: SIBs in Practice Social Impact Bond Contracting entity Intermediary Investor Service Provider Independent Evaluator Peterborough Prison, UK15 U.K. Ministry of Justice Social Finance Consortium of private investments and charities through Social Finance (SF) St. Giles Trust, The Ormiston Children and Families Trust, SOVA and YMCA QinetiQ and University of Leicester Rikers Island Jail, New York, NY16 City of New York: Department of Corrections MDRC Goldman Sachs, with guarantee from Bloomberg Foundation Osborne Association and Friends of Island Academy Vera Institute of Justice Preschool Expansion, Salt Lake City, UT17 Salt Lake County United Way of Salt Lake and Voices for Utah Children Goldman Sachs, J.B. Pritzker Granite School District, UT Dr. Mark Innocenti, Utah State University Youth Services, Massachusetts18 Commonwealth of Massachusetts, with additional support from U.S. Department of Labor Third Sector Capital Partners in partnership with New Profit, Inc. Consortium of commercial and philanthropic organizations that includes the Goldman Sachs Social Impact Fund, The Kresge Foundation, Living Cities, Laura and John Arnold Foundation, New Profit, and The Boston Foundation Roca with Chapin Hall at the University of Chicago Public Consulting Group and Sibalytics LLC Homelessness Services, Massachusetts19 Commonwealth of Massachusetts Massachusetts Housing and Shelter Alliance, the United Way of Massachusetts Bay and Merrimack Valley, and the Corporation for Supportive Housing Consortium of private investors through United Way of Massachusetts Bay and Merrimack Valley and Third Sector Capital Partners Massachusetts Housing and Shelter Alliance Under contract negotiation New York City and Monroe County formerly incarcerated individuals20 New York State Department of Labor Pay for Success program, with additional support from U.S. Department of Labor Social Finance Private and institutional investors including the Robin Hood Foundation through Bank of America Merrill Lynch Center for Employment Opportunities (CEO) Chesapeake Research Associates 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org SIBs in Practice (Continued) 8 P. 8 SIB Status Intervention Goal and Evaluation Method Funding Repayment Provider Peterborough Prison, UK21 Active: Operational since September 2010 Flexible mix of pre- and post-release mentoring and services for about 3,000 adult male offenders completing short prison terms Reduced frequency of reconviction events; intervention group is compared to matched group released from other prisons Investment of £5 million is provided by a consortium of private investments and charities through Social Finance (SF) Repayment by the Ministry of Justice and the Peterborough Big Lottery Fund Rikers Island Jail, New York, NY22 Active: Operational since August 2012 Cognitive behavioral therapy while in jail for young offenders between 16 and 18 years old Reduced readmission bed days (RBDs, also called “future days in jail”) in DOC custody over 2 years, number of participants served; comparison to benchmarks based on historical performance Investment of $9.6 million is provided by Goldman Sachs, but 80 percent of investment is guaranteed by Bloomberg Philanthropies Repayment of up to $11.7 million by the New York City Department of Correction Preschool Expansion, Salt Lake City, UT23 Active: Operational since August 2013 Expansion of preschool services for up to 3,500 at- risk children in Granite School District to reduce achievement gap, decrease use of remedial and special education services, and increase kindergarten readiness Reduced use of remedial and special education services by high risk subgroup identified based on low scores on PeabodyPicture Vocabulary test compared to historical benchmark for similar population Investment of up to $7 million is provided by Goldman Sachs and J.B. Pritzker Salt Lake County has committed $350,000 for repayment; sponsors are seeking additional funding from the State of Utah  Social Impact Bonds: Overview and Considerations P. 9 SIBs in Practice (continued) SIB Status Intervention Goal and Evaluation Method Funding Repayment Provider Youth Services, Massachusetts24 Intermediary and service provider selected August 2012; terms of contract announced January 2014. received Federal Department of Labor Pay for Success funding Sept. 2013 Seven-year project providing basic and post-secondary education and vocational training services to increase employment outcomes and decrease recidivism for 929 at-risk men between 17 and 23 years old who are aging out of the juvenile justice and probation systems; additional federal funding from the Department of Labor could extend project to include 391 additional young men Goals are reducing recidivism by 20% – 60% and improving education and employment outcomes by 10% - 20% over several years; program participants avoid recidivism for 48 continuous months after being released from juvenile justice system. Project will use a Randomized Control Trial to determine outcome payments Investment of at least $18 million is provided by consortium of commercial and philanthropic funding; RFR required applicants to demonstrate a track record of raising at least $10 million in capital The Commonwealth of Massachusetts will make up to $27 million in success payments; In addition to this, the US Department of Labor has granted Massachusetts $11.67 million for use as additional success payments to expand the program; $50,000 paid per foregone incarceration defined as two years of prison post-aging out Homelessness Services, Massachusetts25 Intermediary and service provider selected July 2012; contract under negotiation Increase housing options through MHSA’s ‘Home & Healthy Good’ program based on the Housing First model to create 380 additional housing units and provide stabilizing impact on chronically homeless people Reduced Medicaid and other service costs; evaluation details have not been released Investment of $10 million ($7 million of which is a SIB) of private investors through United Way of Massachusetts Bay and Merrimack Valley and Third Sector Capital Partners Repayment by the Commonwealth of Massachusetts New York City and Monroe County formerly incarcerated individuals26 Received federal Department of Labor Pay for Success funding Sept. 2013 Integrated services of life skills, transitional jobs, job placement and post- placement support to 2,000 individuals Goals are to increase employment outcomes by at least 5 percent and reduce recidivism by at least 8 percent; project will use a Randomized Control Trial to determine outcome payments Investment of $13.5 million from private investors (including $300,000 from the Robin Hood Foundation) through Social Finance and Bank of America/Merrill Lynch. Rockefeller Foundation has committed a 10 percent first-loss guarantee of the total investment Repayment of up to $12 million from the NYDOL through the Governor’s Pay for Success Initiative Fund 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org 10 P. 10 SIBs have the potential to save money by increasing investments in prevention SIBs are often promoted as a way to save money by expanding a preventive activity which is so effective that it results in cashable savings that are large enough to cover the costs of repaying the investors and leave money left over. For example, a presentation explaining the SIB model by Steven Goldberg, then affiliated with Social Finance, one of the organizations that is involved in the creation of SIBs, includes a diagram that shows a $60 reduction in ongoing remediation costs as result of a SIB-funded intervention. Even after accounting for the $25 cost of the intervention, this leaves $35 in cost savings, plus additional unquantifiable social benefits, such as reduced crime or improved family well-being.27 However, it is important to understand that this benefit is the result of the highly cost-effective preventive activities, not the SIB structure. As discussed below, SIBs have financing and implementation costs in addition to the cost of the service delivery. Figure 2: Source: Steve Goldberg, Social Impact Bonds: A New Tool for Scaling Effective Nonprofits, Social Finance, PowerPoint, September 18, 2012, http://www.socialinvestingforum.com/Images/Social%20Solutions%2 0-%20Steve%20Goldberg.pdf. It is also unclear how many of the areas where SIBs have been proposed as a financing mechanism could result in this kind of cashable savings. For an activity to pay for itself, a number of factors are required to align:  Populations can be identified with a reasonable amount of precision for which government agencies are likely to incur significant remediative costs within a few years in the absence of intervention;  These costs can be reduced by an amount sufficient to cover the cost of the program (and of the SIB) if the number of people needing services declines; and  The government agency that will incur the costs of remediation is able to identify the avoided costs and pass the funds on to the agency that entered into the SIB relationship. This may be particularly challenging in the United States, due to the distribution of public activity across different levels of government. For example, the savings from a local program may be primarily realized by Medicaid, which is funded by the federal and state governments. Based on its consultation with experts, McKinsey & Company suggests that the areas that seem most promising for short-term savings are criminal justice, which has the potential to save money by preventing incarceration among high risk groups, and preventive services for chronically homeless individuals, which has the potential to cut both shelter costs and the cost of health care and other emergency services.28 Many of the early SIBs are in fact in these areas. For example, given the track record of reincarceration rates for juvenile offenders held at Rikers Island, New York City is confident  that it will save the costs needed to repay the SIB if it is successful. It is possible that additional types of services will also meet these criteria. For example, the Utah early childhood SIB is based on the expectation of saving money by reducing the need for special education services. The intermediary has identified a population of children who, based on their scores on a standardized early childhood assessment (the Peabody Picture Vocabulary test), are highly likely to need special education services in the absence of intervention. The costs of special education services are closely linked to the number of children served. The City of Salt Lake and the state of Utah will share in the cost savings; so far, only Salt Lake City has committed to participating in the repayment of the SIB, but the sponsors of the effort are hopeful of obtaining state support for additional years.29 However, each proposed project will need to be examined individually to determine whether there are likely to be direct cashable government savings that exceed the costs of the program. It is not clear whether all of the initial SIBs meet this test. For example, as noted before, in the Peterborough prison SIB, while the outcome payments are based on Ministry of Justice calculations about the costs of reconviction, these calculations took into account the broader societal benefits of reduced reoffending, not just the direct savings from reduced reincarceration.30 Most public services are designed to improve societal well-being, not to save taxpayer money directly. The primary beneficiaries of social services are typically the participants themselves, who benefit from higher earnings, better physical or mental health, etc. Taxpayers may benefit indirectly from reduced income support or social services, and increased taxes, but in most cases this is not sufficient to cover program costs, at least not in a 3-5 year time horizon.31 Many governmental activities, including public education, social security, and police and fire, are not intended primarily to save money but instead to achieve important social goals – a secure old age, an educated citizenry, safety and security. In fact, the same intervention may or may not result in cashable savings depending on how tightly it is targeted to a high-risk population. Roca, the service provider for the Massachusetts youth services SIB, has long provided similar services to a broader population of disadvantaged youth in Boston through more traditional grants and contracts. They are quite confident of their service model and its effectiveness in improving participants’ educational and employment outcomes, as well as in reducing recidivism. However, since the payments in this SIB are based on the expected governmental savings from reduced incarceration, they had to negotiate a contract under the SIB that limits services to a group with a very high probability of incarceration.32 There is not a consensus among SIB proponents as to whether SIBs should be limited to those cases that can result in cashable savings. Some have suggested that SIBs could still be valuable for the reasons discussed below even when the programs do not pay for themselves. In such cases, the governmental agency would have to make a value decision about how much it is willing to pay for the specified outcome.33 If projects will not lead to cashable savings, it is critical for policymakers to be clear about this and to have a clear plan for funding the outcomes payments when they come due. SIBs have the potential to expand available resources, and thus break the cycle in which investment in prevention gets squeezed out by current needs Even without generating cashable savings, SIBs could be transformative if they are effective in breaking through the cycle of underinvestment in Social Impact Bonds: Overview and Considerations P. 11 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 12 preventive activities. There is widespread frustration with the lack of government support for preventive services, even with those models that have been demonstrated to create public value far beyond the cost of the services. For example, research on early childhood home visiting, a rigorously studied intervention, has documented outcomes in areas including child health, school readiness, family economic self-sufficiency, reduced child maltreatment.34 And yet, federal funding for this program is only enough to support services to about 3 percent of those who could benefit— and is currently scheduled to expire at the end of federal fiscal year 2014.35 It is not entirely clear why society underinvests in prevention. The type of interventions best suited for SIBs – preventive activities that have cashable benefits in the relatively short term – seem like the economist’s proverbial $20 bill on the sidewalk – if it was that easy, why wouldn’t someone have picked it up already? If rigorous studies proving the benefits of prevention exist and are sufficient to convince profit-motivated private investors that SIBs supporting such prevention efforts would be sound investments, it is hard to believe that it would not be possible to get direct public investment in these areas. But the reality on the ground is that in many cases, governments have not been able to make this case. It remains to be seen whether SIBs will be able to overcome the problem of underinvestment in prevention. SIBs seem particularly well designed to address this issue when underinvestment is driven by skepticism about whether the programs will actually have the effects that are claimed, because under SIBs governments will not have to pay for the interventions until the benefits are actualized.36 SIBs may be particularly helpful in cases where services have been chronically underfunded and have had disappointing results. In some cases, this is because highly disadvantaged individuals are only receiving a portion of the supports that they need to succeed. For example, even a highly effective job training program may not result in increased employment if a participant does not have access to the reliable child care needed to show up at work on time each day. However, if a program has not demonstrated good results, it is often difficult to make the case for increased funding. If SIBs can show that increased funding for services can result in improved outcomes, this could significantly change the political dynamic. Some have suggested that SIBs could go further— and allow society to avoid having to make the tough tradeoffs between serving people who are already in dire situations and investing in services to prevent future hardships. This is only the case for the subset of interventions that produce cashable savings sufficient to cover the costs of the SIB. In other cases, SIBs only postpone the need to either cut services or increase spending until the date when outcome payments must be made. SIBs allow providers increased flexibility to provide customized interventions to respond to client needs, with more time to demonstrate outcomes. The possibility of flexible multi-year funding appeals to program operators who chafe at the restrictive rules of various funding streams and the need to blend multiple programs — and comply with multiple reporting requirements — in order to provide families with the full range of services that they need to thrive. One of the potentials of SIBs is to offer service providers funding without these constraints. For example, under the Peterborough SIB, the foundations had a lot of confidence in the intermediary— and were willing to trust it with flexible funding known as the “One Service” in order to achieve the desired outcome. Coverage of the One Service project highlights the value of this flexible funding.37 12  Flexibility also allows the intermediary to select a mixture of service providers in order to respond to the diversity of clients and their needs. Many smaller community-based organizations do not have the institutional capacity to compete for public funding, let alone the increased sophistication and financial support needed to be the lead organization in a pay-for-success contract. However, some of these organizations may have cultural competencies and community connections that enable them to reach clients who would not respond to outside agencies. Under a flexible SIB, the intermediary could contract with such community organizations as appropriate, and assist them in improving their capacity if needed. However, it is unclear whether other funders, especially profit-motivated investors, will be willing to support this sort of flexibility and uncertainty. As noted above, none of the U.S. SIBs have anywhere near the One Service project level of flexibility in providing services; in order to make the case for the likelihood of success, the intermediaries have had to specify exactly what services will be provided, and what entities will be delivering them. Some have suggested that SIBs will accelerate the adoption of promising but untested strategies; this assumes that investors will be less risk averse than governments, because they will have the capacity to spread risk over multiple projects.38 At this stage, there are not enough SIB projects in the works for this to be the case, and the whole SIB structure is sufficiently novel that the investors appear to be sticking to models that have already been extensively evaluated. In addition, it is not clear that governments will actually be less risk averse because of the SIB structure. This would be true if governmental risk aversion were primarily driven by the financial risks of supporting untested programs— and thus would be removed when these risks are transferred to the investors. However, the political consequences of failure may not be completely mitigated even if the government does not have to pay for the unsuccessful services. If voters still hold them responsible for failure to achieve the desired outcomes, elected officials are likely to be cautious about supporting untested innovations. Moreover, even if the government contract with the intermediary allows for flexible multi-year funding, this does not ensure that the intermediary’s contract with the service provider will be equally flexible. The roles and responsibilities of each organization will need to be negotiated for each project. SIBs can help embed rigorous evaluation and a focus on outcomes into program operations, increasing learning, and ultimately performance. Another potential benefit of SIBs is that they embed both rigorous outcome measurement and performance improvement into programs on an ongoing basis. At the most basic level, if a SIB- funded project does not achieve the specified outcomes, the investors will not be repaid and taxpayer funds will be preserved. Because the investors and the intermediary have a strong interest in ensuring success, programs will assessed, tweaked, and re-assessed along the way. To this extent, SIBs can be thought of as a way of purchasing improved project management capacity for program oversight, using data for performance measurement and making mid-course corrections. Most supporters of SIBs suggest that the potential benefits of SIBs go beyond the individual projects to promote more widespread use of evidence and improved performance elsewhere. The process of developing a SIB project and identifying the outcomes to be measured and what governments are willing to pay for them can help generate a more robust public dialogue that clarifies the goals of programs and how we value their outcomes. This dialogue can also create the impetus for funders, services providers and communities to focus their energy and attention on the outcomes that they wish Social Impact Bonds: Overview and Considerations P. 13 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 14 to achieve, rather than on process measures of performance. More concretely, some have suggested that SIBs could catalyze the use of administrative data for performance management. Outcome measurement is less costly and thus more feasible at scale when it is based on existing data (e.g. unemployment insurance earnings records, health care system use, correctional systems outcomes), rather than requiring new data to be collected. Use of administrative data also produces more representative samples than surveys with varying response rates and is not subject to distortions from self-reporting. However, legal and technical barriers have, so far, limited the use of data in this way. The hope is that SIBs will bring different agencies and evaluation experts together to address these barriers and bring down the cost of outcome measurement. Intermediaries and investors will also want to have ongoing access to data in order to monitor program performance along the way. Because many potentially beneficial interventions have not yet been rigorously evaluated, some have highlighted the value to society of more evaluations of promising programs. Indeed, rigorous evaluations can benefit policymakers and program operators in other states and localities, not just those who are supporting and operating the specific program that is evaluated. However, the more flexible the program intervention, and the more mid-course corrections that a program takes, the harder it becomes to specify the intervention that is being tested and what the implications are for other programs. Therefore, there may be tradeoffs between the power of the SIB to improve services in the specific location and the value of the evaluation for external audiences. While the clarity that random assignment evaluations provide is desirable, such experiments are often costly, can be challenging to implement, and are better suited to small-scale programs than to large-scale interventions that are designed to change broad systems. They also are not well-suited at times when ongoing adaptation to the needs of the population and the context are priorities. Given these challenges, most of the SIBs that are currently under way use comparison groups but not full random assignment. In these programs, outcomes are compared to similar individuals at different locations, or in a prior time period. What are the potential downsides of a SIB? There are also potential downsides to the SIB model. Some of these are inherent to the model while others are simply the issues that arise whenever high stakes are attached to outcome measures. Policymakers should pay attention to these concerns both in assessing whether a SIB is an appropriate mechanism for funding a package of services and in negotiating the terms of the contract if they decide to proceed with a SIB. SIBs have costs above and beyond the cost of providing services. It is important to understand that the SIB mechanism imposes costs beyond the cost of providing services. It will always be more expensive to support a SIB — which must pay a return to the investors, as well as cover the costs of the intermediary and the evaluators — than to support the exact same program with direct government funding. Given the low interest rates at which governments can generally borrow funds, this is true even if the government would need to borrow in order to pay for the service. Another way to think of this is to remember that the risk of non- performance does not disappear in a pay for performance contract but is simply shifted from the government agency to the entity with which it contracts. Basic finance tells us that investors will demand higher returns for more risky investments. 14  Higher costs are the inevitable consequence of shifting risk away. The best explanation we have found of how this might play out in practice is McKinsey’s pro forma analysis of a hypothetical SIB involving providing Functional Family Therapy (FFT) to juvenile offenders. This model has been studied with rigorous evaluations, and there is good data available on the costs of service provision and the expected benefits to society. McKinsey calculates that while government recoups the $3,191 cost of providing one young offender with FFT within 8 years, with the added costs of a SIB, it would have to capture savings over 12 years in order to break even. McKinsey concludes “SIBs are a more expensive way to finance the scaling up of preventive programs than if the government simply went to service providers and paid them to expand an intervention to more constituents... The ‘premium’ inherent in scaling programs through SIBs is justified only if conventional options aren’t working, and if the SIB structure is adding value commensurate with its cost.” 39 Particularly at this early point of development, SIBs also involve significant up-front investments of time, attention and money in order to identify the partners and negotiate a contract that works for everyone. Massachusetts released its solicitation of applications to participate in two pay-for-success demonstrations in January 2012 and announced the organizations that had been selected to participate in August 2012.40 The details of the juvenile justice project were released in January 2014, and the details for the homelessness services project, including the key question of how performance will be measured and rewarded, have not been released.41 The U.S. Department of Labor published its solicitation under pay for success in June 2012, with applications originally due in December 2012. That deadline was later extended to January 2013, and the awards were made in September 2013.42 In response to the Request for Information from the Department of Treasury regarding pay-for-success initiatives, a contracting officer with the Los Angeles Unified School District reported that staff had estimated the minimum up-front costs for personnel and legal fees associated with the development of a SIB at approximately $100,000 over 6-9 months.43 While acknowledging that the transaction costs of the Rikers Island SIB are high, MDRC suggests that these costs will come down in future SIBs “as the model becomes increasingly known and accessible, standards for evidence develop, and investors come to see SIBs as less new and risky.”44 In fact, one justification for a federal role in supporting SIBs and other pay-for-success contracts at this stage is the recognition that these early demonstrations are building a knowledge base for further activity, and that the state or local government agency entering into the contract should not bear the full cost of this knowledge development. However, even when SIBs are a fully mature model, they will still have incremental costs above the cost of service delivery, including the costs of the intermediary and evaluation, as well as the profit to the investors. SIBs involve loss of government control over programs and providers The flip side of increased flexibility for service providers is loss of government control over programs and providers. Considering a SIB forces a hard look at the question of whether society truly only cares about outcomes, or if it turns out that there are actually process factors that matter as well. For example, many people would find it unacceptable to use the threat of physical punishment to force people to participate in a training program, even if it turned out to be effective. Similarly, when it comes to vulnerable populations, such as children receiving child welfare services, or core governmental functions, such as education, policymakers and the public may want all individuals to receive a minimum amount Social Impact Bonds: Overview and Considerations P. 15 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 16 of services— and would not find it acceptable if a provider under a SIB chose to exclude some people from services. A SIB intermediary could also withdraw entirely from providing services in midstream. It is possible that midway through a performance period, an intermediary could realize that the odds are highly unlikely for reaching the performance threshold needed to receive the payment under the contract. Under that condition, it would be fiscally prudent for the intermediary to cut its losses by withdrawing from the program. If this occurs, the government would have to decide whether it is okay to stop providing services to this population, or find another way to provide such services (possibly through a more traditional fee-for-service contract, or a pay-for-performance contract with a lower threshold for performance). A related issue is that the prototypical SIB contract allows the intermediaries to select service providers and does not require an open competition. The outcome incentives under SIBs ensure that an intermediary will not select providers whom it does not think can do the job, but there are no protections against what Tammany Hall boss George W. Plunkett memorably described as “honest graft.”45 He argued that someone was going to get government contracts and make money, so you might as well pick your friends, as long as they are competent and capable of delivering the services the public needs. While this is common practice in the private sector, since Tammany Hall days the U.S. public sector developed rules that are designed not only to ensure that the public pays a fair price for goods and services, but also to ensure that all companies have a fair chance at competing for that business. Past experience with outcome-based performance measures offers cautionary lessons about the unintended consequences of tying high stakes to such measures There is a long history of performance-based contracting in many government services, as well as other strategies for using both rigorous research and ongoing administrative data on outcomes to guide policy and program choice as well as implementation The extensive literature about this experience offers a number of cautionary lessons about unwanted side effects of attaching significant consequences to failure to achieve designed targets on outcome-based performance measures, even well-intentioned ones. Among the harmful consequences of tying high stakes to such measures are disincentives to serve the most needy populations (often referred to as “creaming” or “cream-skimming”), skewing services to focus on the selected outcomes of interest at the expense of other aspects of a program (“tunnel vision”) and winding up with programs that affect the measured indicators but not the underlying outcomes (“teaching to the test”). Those undertaking SIBs should pay attention to these lessons, and include efforts to mitigate these negative effects. Creaming Creaming is a concern whenever the target population for an intervention varies significantly in their likelihood of achieving the desired outcome, and the service provider is likely to be able to predict with reasonable accuracy which individuals are more or less likely to succeed. When high stakes, such as loss of the investment, are tied to failure to achieve the desired outcomes, service providers have an incentive to focus their efforts on individuals who are more likely to succeed (including those who are likely to succeed even in the absence of the program) and to deny services to 16  those who have significant barriers to participation or are otherwise less likely to succeed. Creaming is a particular issue whenever performance is measured only for the individuals who are actually served by a program (as opposed to everyone who is eligible for a service) and when the program has control over which individuals are served- and thus included in the outcome measure. For example, there has been a great deal of concern that the Workforce Investment Act outcome measures have contributed to programs being less likely to serve individuals with low literacy levels or who have other barriers to employment. Many TANF programs have imposed up-front job search or community service requirements that have the effect of screening out individuals who are less likely to participate in work activities and thus will drag the state’s work participation rate down. This reduces the value of the work participation rate as a measure of the program’s effectiveness in engaging recipients in work activities, because non- participants are removed from the denominator. In a number of areas, programs have reported much better outcomes for participants than non- participants, only to be found to have much more modest impacts when subjected to a formal evaluation.46 One key reason for this is that even when program participants share characteristics that are generally considered to make them “hard to serve” — such as criminal records, low basic skills and lack of credentials, or substance abuse history — as long as the program is voluntary, participants are likely to be more motivated than non- participants. Therefore, it is important that programs’ success be measured based on either impacts, or outcome measures for a population as a whole, not just for those it serves. Even when the intermediary or service provider does not have the ability to influence whose outcomes are counted, it can still be challenging to design an outcome-based performance contract in such a way that contractors have incentives to serve even the most disadvantaged participants. For example, if contractors are paid a fixed amount for every individual who achieves the desired outcome, and they can accurately assess who is more or less likely to succeed, it would make business sense to provide services only to those who are closest to the margin for success. There is no incentive for them to serve those who are likely to succeed without help – or those who are likely to require help that costs more than the agreed-upon payment. Some of the existing SIBs have thresholds for performance, below which the investors get no repayment. One advantage of this approach is that it ensures that intermediaries cannot simply collect windfall payments for individuals who would likely have achieved the desired outcomes on their own. However, it is not clear that this is sufficient to ensure that even the most challenging participants receive appropriate services.47 Moreover, many of the initial SIBs set a cap on the total performance payment that may be collected by the intermediary. This was done to reduce the governmental uncertainty over the total cost of the SIB, but it has the potential to increase the disincentive to serving the most at-risk segment of the target population. In order to address concerns about creaming, it may be appropriate to require certain levels of access and participation as well as outcomes. When it is possible to identify the characteristics that make certain individuals more challenging and costly to serve effectively, it can be helpful to provide tiered levels of outcome payments, where providers can earn additional amounts for succeeding with more disadvantaged individuals. It may also be appropriate to provide payments for incremental success when serving very disadvantaged groups. When a SIB contract sets limits on the services that can be provided, and the provider has reason to believe that some individuals will not benefit from those particular services, there should be a mechanism to encourage providers to connect these Social Impact Bonds: Overview and Considerations P. 17 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org 18 P. 18 individuals with other services, rather than to simply exclude them. Tunnel vision Outcome measures can also distort the specific services that are offered. This is particularly an issue when services are intended to result in multiple outcomes, but only some are measured and rewarded. Performance measures that only capture one aspect of a program can result in that aspect being emphasized and others neglected. For example, high-quality preschool supports children’s development and well-being across domains, including social-emotional, physical and cognitive development. However, there have been efforts that would measure the performance of early childhood education programs largely on student performance on a standardized literacy and math test.48 There is justifiable concern that such a narrow measure of effectiveness would create an incentive to devote class time to drilling on these measures, at the cost of time available for promoting broader social-emotional development. Many believe that the focus on the results of standardized tests in K-12 education has already taken away learning time from subjects that are not tested. In some areas, such as child welfare, there may actually be direct competition between desired goals, such as minimizing the amount of time children spend in out-of-home placement and ensuring that children are not returned to unsafe environments. Both of these goals are critical to the well-being of the children who are served, and so an effective child welfare system must balance these competing goals in order to meet children’s needs. Performance measures that focus only on one of these goals could have serious harmful effects on children. Teaching to the test If multiple desired outcomes are highly correlated, there may be less reason to worry about only measuring some of them. For example, a report commissioned by South Carolina to analyze the feasibility of using SIBs to support an expansion of early childhood home visiting programs argues that payments could be made conditional on reductions in pre-term births. The researchers argue that because pre-term births are predictors of many other negative outcomes, this would be a fair indicator of the programs’ effectiveness and long-term savings potential.49 However, the evaluation literature notes that tying high stakes to a predictive measure can have distortionary consequences. Even if the correlation between a performance measure and the underlying outcomes of interest is strong at the beginning, this correlation tends to decrease over time as people figure out ways to game the system by improving results on the specified measures without actually changing the core outcomes in the desired direction.50 The best known example of this is “teaching to the test;” by focusing on test-taking skills, schools may be able to improve student performance on standardized tests without actually improving the student’s literacy or math comprehension. In response to these challenges, some of the governmental agencies using outcome-based performance payment systems have adjusted their systems over time, replacing measures that appear to have lost predictive value and re-setting the balance between competing values.51 However, by the nature of a SIB, it is essential to establish the outcomes that will be measured and the payments associated with them at the beginning — and not to alter them during the period of the contract.  A related issue is that in many cases, important outcomes cannot be measured directly but require proxies, such as test scores for knowledge, crime reports for actual crimes. No outcome is ever perfectly measured, for a variety of reasons — people work off the books, give survey answers that they believe to be socially desirable, fail to report crimes, etc. This does not mean that these measures can’t be useful — perfection is unachievable, and even noisy data can provide useful information, as long as the “noise” is not correlated with the intervention. It is important to be aware of the potential differences between the measure and the underlying reality- and to spend some effort thinking about the possible causes and ways to address them. It is also critical that the data on which performance payments are based should be collected and reported by an independent entity, in order to minimize the possibility of bias. When is a SIB Appropriate? Because of the high level of interest in SIBs, many policymakers, practitioners, funders and advocates are facing the question of whether a SIB would be a good way to expand funding for a particular intervention or population in a given state or community. A wide range of areas have been suggested as possibilities for pay-for-success contracts. Many of the early SIB projects or proposals have focused on reintegration services for ex-offenders and juvenile justice. Other areas that have been proposed for SIBs include housing for homeless individuals and families, child welfare, early education, early childhood home visiting, job training, and health care. In this section, based on the information presented so far, we offer a preliminary overview of the questions that need to be considered for a SIB to be plausibly offered as a financing mechanism. This is not a detailed checklist of the components needed to implement a successful SIB, but should rather be treated as a preliminary screen to narrow the field of programs to consider.  Does Everyone Agree on What Success Would Look Like? While implicit in the very term “pay for success,” it is essential to highlight the importance of both agreement on the desired outcome and ability to measure these outcomes in ways that reflect the value added by services. Is the government agency indifferent to the means by which the service provider will achieve the desired outcomes, or are there expectations about a minimum level and type of services that all eligible participants will receive? These questions should be addressed very early in the process of exploring a SIB transaction. Moreover, bringing stakeholders together for these conversations can be a very valuable activity for a community, even if a SIB is not the end result.  What is the Funding Mechanism? Is the expectation that achieving the desired outcomes will result in cashable savings sufficient to repay the investors with interest? If so, will these cashable savings apply to any success or only occur if outcomes are achieved at scale? Does it require other levels of government or other government agencies to identify and contribute the funds that they will save as a result of the intervention? If the project will not result in cashable savings, how will the agency determine the amount that it is willing to pay for the projected outcomes?  Is There a Plausible Evidence Base to Suggest the Intervention Will Succeed? Because one of the benefits of a SIB is the ability to provide flexible services, it does not make sense to limit SIBs only to Social Impact Bonds: Overview and Considerations P. 19 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 20 interventions that have been previously tested through a rigorous controlled evaluation. However, there must be a sufficient evidence base in support of the intervention to make it plausible that the project will succeed. One of the most promising ways to use SIBs is to build flexible additional services upon a core intervention that has already shown some success. In practice, the investors —with input from the intermediaries and other parties —will be the de facto arbiters of what evidence is sufficient, as they will have their funding at risk if the project fails.  Can the Proposed Intervention be Taken to Sufficient Scale to Achieve Outcomes? In order for a SIB to succeed, there must both be an identifiable pool of individuals who could benefit from the services, and service providers who are able to expand or replicate the intervention. Given the overall degree of underinvestment in prevention, this is usually assumed not to be a problem. However, when a SIB is being justified based on the ability of intervention to prevent a harmful outcome, it is important to be able to identify a group of participants who would be highly likely to have this outcome in the absence of the service. For example, under the Massachusetts youth SIB, Roca, proposed to serve a specific population of young offenders who were highly likely to be re-incarcerated. Since their work is geographically specific, they had to check to ensure there were enough people meeting these criteria in the targeted communities to make the project feasible.52  Does the Project Represent an Expansion of Funding and Services? While some have suggested that SIBs should be used to redirect funding towards more effective programs, we are concerned about the possibility that they will simply shift funding towards more “SIB-able” project areas, at the expense of other areas, where outcomes are harder to measure, or the benefits are more diffuse. Because the SIB model is largely untested, we believe that at least for now, it should be limited to new funding. Given the additional costs involved in a SIB compared with direct contracting (whether fee-for-service or outcome-based), the case has not been made for taking funding away from existing programs. In addition, SIBs should expand services to individuals who are not already being served, rather than replacing existing services. This is important because intermediaries may stop providing services if they are not going to be able to achieve the performance targets. This is only acceptable if the program is serving people who would not otherwise have received services. CLASP applauds the growing interest in improving outcomes in a range of areas where governments provide services. SIBs and other pay-for-success mechanisms could potentially add value as a means of expanding investment, supporting prevention-focused activities, focusing on outcomes, and improving the evidence base for what works. However, expanding use of SIBs should not be a goal; they are only worth supporting to the extent that they contribute to the goal of expanding the reach of high-quality prevention-oriented services, including to the most disadvantaged groups. We hope that these questions will be of help to those considering SIBs in supporting their efforts to think strategically about whether SIBs are the right vehicle to use to expand this particular service, in this place, at this time, in order to further that goal. 20  Social Impact Bonds: Overview and Considerations March 7, 2014 1 See “Opportunity for All: Building and Using Evidence to Improve Results” FY 2015 Fact Sheet, http://www.whitehouse.gov/sites/default/files/omb/bud get/fy2015/assets/fact_sheets/building-and-using-evidence-to- strengthen-results-in-government.pdf and “Using Evidence to Get Better Results,” Budget of the United States Government, Fiscal Year 2014, Office of Management and Budget, http://www.whitehouse.gov/sites/default/files/omb/budget/fy2 014/assets/budget.pdf. For further explanation of the Adminstration’s proposals, see Sonal Shah and Kristina Costa, “Social Impact Bonds: White House Budget Drives Pay for Success and Social Impact Bonds Forward,” Center for American Progress, April 23, 2013, http://www.americanprogress.org/issues/economy/news /2013/04/23/61163/white-house-budget-drives-pay-for- success-and-social-impact-bonds-forward/. While Congress did not support this new fund in the FY 2014 appropriations bills, it did provide for Pay for Success models in both ex- offender services and national and community service programs. 2 Sonal Shah and Kristina Costa, Social Impact Bonds: Social Finance - A Primer Understanding Innovation Funds, Impact Bonds, and Impact Investing, November 5, 2013, http://www.americanprogress.org/wp- content/uploads/2013/11/SocialFinance-brief.pdf. Updated lists of SIB activity are also maintained by Social Finance at: http://www.socialfinanceus.org/social-impact- financing/social-impact-bonds/history-sib-market/united-states and Instiglio at: http://www.instiglio.org/sibs-worldwide/. 3 “About Pay for Success Social Impact Bonds,” Nonprofit Finance Fund, http://payforsuccess.org/about. 4 In Social Impact Bonds: A Promising New Financing Model to Accelerate Social Innovation and Improve Government Performance, Center for American Progress, February 2011, Jeffrey Liebman notes the fact that many service providers do not have the capacity to provide the up-front funding for services or to bear the risk of non-performance, as a reason why 100% performance based contracts are relatively rare in the human services world. When they exist, as in New York City’s welfare to work program, they tend to be taken up by large multi-site for-profit service providers. More common are partially performance-based contracts, where the government makes up-front payments to cover at least a portion of the costs of service delivery and providers can earn additional payments when participants achieve the desired outcomes. The ability to engage small nonprofit service providers who may not have the financial capacity to taken on 100% performance based contracts is sometimes cited as a benefit of SIBs. 5 Laura Callanan, Jonathan Law, Lenny Mendonca, From Potential to Action: Bringing Social Impact Bonds to the US, McKinsey & Company, May 2012, http://mckinseyonsociety.com/downloads/reports/Social- Innovation/McKinsey_Social_Impact_Bonds_Report.pdf. 6 Even though certainty of funding is sometimes listed as one of the benefits of a SIB, this does not appear to be inherent to the model. The intermediary could choose to change service providers in mid-stream if it determined that the initial provider chosen was failing to deliver adequate services. 7 Because the quality of a program is only one factor in determining outcomes, there is not necessarily a correlation between the programs with the best outcomes and the programs with the largest impacts, if programs are in different economic contexts or serve different populations. In particular, there is an extensive literature that shows that programs serving highly disadvantaged populations often achieve lower levels of outcomes than ones that serve more advantaged population, even when they result in larger changes in participants’ outcomes, compared to what would have happened in the absence of the program. 8 "Guidance - Social Impact Bonds: A Data Gathering How-to Guide for Commissioners," Cabinet Office, Government of the United Kingdom, February 5, 2013, https://www.gov.uk/government/publications/social-impact- bonds-how-to-guide-for-commissioners. 9 Timothy Rudd, Elisa Nicoletti, Kristin Misner, Janae Bonsu, Financing Promising Evidence-Based Programs: Early Lessons from the New York City Social Impact Bond, MDRC, December 2013, http://www.mdrc.org/sites/default/files/Financing_Prom ising_evidence-Based_Programs_FR.pdf. 10 “Peterborough Social Impact Bond Fact Sheet,” Social Finance, One Service and the U.K. Big Lottery Fund, 2011, http://www.socialfinance.org.uk/sites/default/files/SF_Peterbo rough_SIB.pdf. 11 Jenna Pudelek, “HM Prison Peterborough Social Impact Bond Has Led to a Fall in Reconvictions, Official Figures Show,” Third Sector News, June 14, 2013, http://www.thirdsector.co.uk/news/1186265/. 12 Rudd et al., 2013, page 15. The city believes it can save $4,600 per bed for reductions of less than 100 beds, but $28,000 per jail bed for reductions of 100 beds or more. 13 Emma Disley, Jennifer Rubin, Emily Scraggs, et al., Lessons Learned From the Planning and Early Implementation of the Social Impact Bond at HMP Peterborough, Ministry of Justice and RAND Europe, May 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 22 2011, https://www.gov.uk/government/uploads/system/upload s/attachment_data/file/217375/social-impact-bond-hmp- peterborough.pdf. 14 See for example, Jeffrey B. Liebman, Social Impact Bonds: A Promising New Financing Model to Accelerate Social Innovation and Improve Government Performance, Center for American Progress, February 2011, http://www.americanprogress.org/issues/2011/02/pdf/social_i mpact_bonds.pdf and Steve Goldberg, Social Impact Bonds: A New Tool for Scaling Effective Nonprofits, Social Finance, PowerPoint, September 18, 2012, http://www.socialinvestingforum.com/Images/Social%20Solut ions%20-%20Steve%20Goldberg.pdf. 15 Disley et al., 2011; Siôn Cave, Darrick Jolliffe , et al., Peterborough Social Impact Bond: An Independent Assessment, Ministry of Justice, May 2012, https://www.gov.uk/government/uploads/system/uploads/attac hment_data/file/217392/peterborough-social-impact-bond- assessment.pdf; and Social Finance, "Criminal Justice," September 2010, http://www.socialfinance.org.uk/work/sibs/criminaljustice. 16 Esmé E. Deprez and Michelle Kaske, “Goldman Sachs Inmate Bet Fuels Social-Impact Bonds: Muni Credit,” Bloomberg, Aug 21, 2012, http://www.bloomberg.com/news/2012-08-21/goldman-sachs- inmate-bet-fuels-social-impact-bonds-muni-credit.html; George Overholser, Case Study: Preparing for a Pay for Success Opportunity, The Rockefeller Foundation and Third Sector Capital Partners, April 2013, http://www.thirdsectorcap.org/wp- content/uploads/2013/04/Third-Sector_Roca_Preparing-for- Pay-for-Success-in-MA.pdf; John Olson and Andrea Phillips, Rikers Island: The First Social Impact Bond in the United States, Goldman Sachs, Federal Reserve Bank of San Francisco, Pay for Success Financing, volume 9, issue 1, April 2013, http://www.frbsf.org/community- development/files/rikers-island-first-social-impact-bond- united-states.pdf; and Rudd et al., 2013. 17 Emilee Eagar, “Public-Private Partnership Pumps Dollars into Preschool Program,” Deseret News, July 16 2013, http://www.deseretnews.com/article/865583232/Public- private-partnership-pumps-dollars-into-preschool- program.html and Kelli Fratto and Austin Isbell, United Way of Salt Lake Announces Results-based Financing for Low- income Preschool Students, United Way of Salt Lake, June 13, 2013. 18 Overholser, 2013; Elkins, 2013; Massachusetts Executive Office of Labor and Workforce Development Project Name: The Massachusetts Juvenile Employment and Recidivism Initiative, PFS Grantee Summary, Workforce Innovation Fund; and Press Release, Office of Governor Deval L. Patrick, 2014. 19 Overholser, 2013; Kristina Costa and Jitinder Kohli, Social Impact Bonds: New York City and Massachusetts to Launch the First Social Impact Bond Programs in the United States, Center for American Progress, November 5, 2012, http://www.americanprogress.org/issues/economy/news/2012/ 11/05/43834/new-york-city-and-massachusetts-to-launch-the- first-social-impact-bond-programs-in-the-united-states/; Lili Elkins, Supporting At-Risk Youth: A Provider’s Perspective on Pay for Success, Roca Inc., Federal Reserve Bank of San Francisco, Pay for Success Financing, volume 9, issue 1, April 2013, http://www.frbsf.org/community- development/files/supporting-youth-provider-perspective-pay- for-success.pdf; Massachusetts Executive Office of Labor and Workforce Development Project Name: The Massachusetts Juvenile Employment and Recidivism Initiative, PFS Grantee Summary, Workforce Innovation Fund; “The Massachusetts Juvenile Justice Pay for Success Initiative,” Third Sector Capital Partners, http://www.thirdsectorcap.org/massachusetts-juvenile-justice- pfs/; and Press Release, Massachusetts Launches Landmark Initiative to Reduce Recidivism Among At-Risk Youth  $27 million Initiative is Largest Financial Investment in a Pay for Success Contract in the Country, Office of Governor Deval L. Patrick, January 29, 2014, http://www.thirdsectorcap.org/wp- content/uploads/2014/01/1.29.2014-SIF-Press-Release.pdf. 20 Center for Employment Opportunities, 2013; New York State Department of Labor Project Name: New York State Pay for Success Projects: Employment to Break the Cycle of Recidivism, PFS Grantee Summary, Workforce Innovation Fund; Press Release, Bank of America Merrill Lynch, 2013; and Press Release, Office of Governor Andrew M. Cuomo, 2013. 21 Disley et al., 2011; Cave et al., 2012; and Social Finance, 2010. 22 Deprez et al., 2012 and Costa et al., 2012. 23 “FACT SHEET: The Utah High Quality Preschool Program,” Goldman Sachs, J.B. and M.K. Pritzker Family Foundation and United Way of Salt Lake, 2013, http://www.goldmansachs.com/what-we-do/investing-and- lending/urban-investments/case-studies/impact-bond-slc- multimedia/fact-sheet-pdf.pdf; Sean Meehan, “Goldman Sachs Makes Investment, Gets Returns on Utah Preschool Program,” Education Week, June 25, 2013, http://www.utahchildren.org/newsroom/what-s-new/item/375- goldman-sachs-makes-investment-gets-returns-on-utah- preschool-program; and "Urban Investments: Social Impact Bond for Early Childhood Education," Goldman Sachs, http://www.goldmansachs.com/what-we-do/investing-and- 22  Social Impact Bonds: Overview and Considerations P. 23 lending/urban-investments/case-studies/salt-lake-social- impact-bond.html. 24 Overholser, 2013; Elkins, 2013 25 Joe Finn and Jeff Hayward, Bringing Success to Scale: Pay for Success and Housing Homeless Individuals in Massachusetts, Massachusetts Housing and Shelter Alliance and United Way of Massachusetts Bay and Merrimack Valley, Federal Reserve Bank of San Francisco, Pay for Success Financing, volume 9, issue 1, April 2013, http://www.frbsf.org/community-development/files/bringing- success-scale-pay-for-success-housing-homeless-individuals- massachusetts.pdf; Mary Moore, “Social Impact Bonds are Advancing,” Boston Business Journal, July 12, 2012, http://www.bizjournals.com/boston/print- edition/2012/07/13/social-impact-bonds-are- advancing.html?page=all; and “Social Impact Bonds Hit Massachusetts,” Mission Investors, August 28, 2012, https://www.missioninvestors.org/news/social-impact-bonds- hit-massachusetts. 26 “CEO Awarded Pay for Success Grant,” Center for Employment Opportunities, September 24, 2013, http://ceoworks.org/ceo-awarded-pay-success-grant/; New York State Department of Labor Project Name: New York State Pay for Success Projects: Employment to Break the Cycle of Recidivism, PFS Grantee Summary, Workforce Innovation Fund; Press Release, “Bank of America Merrill Lynch Introduces Innovative Pay-for-Success Program in Partnership With New York State and Social Finance Inc.,” Bank of America Merrill Lynch, December 30, 2013, http://newsroom.bankofamerica.com/press-releases/global- wealth-and-investment-management/bank-america-merrill- lynch-introduces-innovat; and Press Release, “Governor Cuomo Announces New York the First State in the Nation to Launch Pay for Success Project in Initiative to Reduce Recidivism,” Office of Governor Andrew M. Cuomo, December 30, 2013, http://www.governor.ny.gov/press/12302013-pay-for-success- project. 27 Goldberg, 2012. 28 Callanan et al., 2012. 29 H.B. 96, a bill providing state funding for performance- based contracts with private entities, was passed by the Utah House in February 2014, but had not yet been taken up by the Senate at the time of this writing, http://le.utah.gov/~2014/bills/static/HB0096.html. 30 Disley, et al., 2011. 31 Liebman, 2011. 32 Overholser, 2013. 33 Liebman, 2011 and Rudd et al., 2013. 34 "Home Visiting Evidence of Effectiveness," U.S. Department of Health & Human Services, http://homvee.acf.hhs.gov/document.aspx?rid=5&sid=20&mid =1. 35 Pew Center on the States, States and the New Federal Home Visiting Initiative: An Assessment from the Starting Line, The Pew Charitable Trusts, 2011, http://www.pewstates.org/uploadedFiles/PCS_Assets/2011/ass essment_from_the_starting_line.pdf. 36 In practice, states will probably need to obligate funds for SIBs up front, and may even need to place them into reserve accounts in order to enter into a multi-year contract. See Kristina Costa, Sonal Shah, Sam Ungar et al., “Social Impact Bonds: Frequently Asked Questions,” Center for American Progress, December 5, 2012, http://www.americanprogress.org/wp- content/uploads/2012/12/FAQSocialImpactBonds-1.pdf. 37 “Why the Peterborough Prison Social Impact Bond is Working for Ex-offenders,” Pioneers Post TV, June 11, 2013, http://www.pioneerspost.com/pp-tv/20131106/why-the- peterborough-prison-social-impact-bond-working-ex- offenders. 38 Liebman, 2011. A more indirect version of this argument is the suggestion that foundations will be more willing to support rigorous evaluations of promising practices if they know that such evaluations will lead to ongoing support for these interventions through SIBs. 39 Callanan et al., 2012. 40 Press Release, “Massachusetts First State in the Nation to Announce Initial Successful Bidders for 'Pay for Success' Contracts,” Massachusetts Executive Office for Administration and Finance, August 01, 2012, http://www.mass.gov/anf/press- releases/fy2013/massachusetts-first-state-in-the-nation-to- announce-ini.html. 41 Press Release, Office of Governor Deval L. Patrick, 2014. 42 “US Labor Department Awards Nearly $24 Million in Pay for Success Grants,” United States Department of Labor, News Release Number 13-1936-NAT, September 23, 2013, http://www.dol.gov/opa/media/press/eta/ETA20131936.htm. 43 “Comment from Galen van Rensselaer,” Department of the Treasury (TREAS) Notice: Strategies To Accelerate the Testing and Adoption of Pay for Success (PFS) Financing Models, Dec 2, 2013, http://www.regulations.gov/#!documentDetail;D=TREAS- DO-2013-0006-0010. 44 Rudd et al., 2013, page 53. 45 George W. Plunkitt, Plunkitt of Tammany Hall: A Series of Very Plain Talks on Very Practical Politics, uploaded December 29, 2008, http://www.gutenberg.org/ebooks/2810. 1200 18th Street NW • Suite 200 • Washington, DC 20036 • p (202) 906.8000 • f (202) 842.2885 • www.clasp.org P. 24 46 See, for example Burt S. Barnow. Exploring the Relationship Between Performance Management and Program Impact: A Case Study of the Job Training Partnership Act, Institute for Policy Studies, Johns Hopkins University, 1999 and Stephen Freedman et al. National Evaluation of Welfare-to-Work Strategies: Evaluating Alternative Welfare-to-Work Approaches: Two-Year Impacts for Eleven Programs, U.S. Department of Health and Human Services and U.S. Department of Education, 2000. 47 This is generally an issue because of the asymmetry of information, where the service provider usually knows more about the likelihood of success with a given individual than the contracting government agency. Mathematically, if the intermediary or service provider is able to predict the expected cost for a given individual of providing services needed to achieve the specified outcome, in a straight pay-per-outcome contract, the intermediary should rationally only serve those whose expected cost to serve is less than the payment level. If the contract requires a threshold level of performance of X% of participants, the intermediary should serve the (X+n)% easiest to serve participants, with n being an additional number that covers the uncertainty in the prediction. 48 Richard Rothstein, “Too Young to Test,” American Prospect, October 17, 2004, http://prospect.org/article/too- young-test. 49 “Social Impact Bond RFI #1: Request for Information (RFI),” South Carolina Department of Health and Human Services, September 17, 2013, https://www.scdhhs.gov/sites/default/files/Social_Impact_Bon d_RFI_%231.pdf. 50 Carolyn J. Heinrich and Gerald Marschke, “Incentives and Their Dynamics in Public Sector Performance Management Systems,” Journal of Policy Analysis and Management, Vol. 29, No. 1, 183–208 (2010). 51 Heinrich, 2010 and Swati Desai, Lisa Garabedian and Karl Snyder, Performance-Based Contracts in New York City: Lessons Learned from Welfare-to-Work, Rockefeller Institute of Government, June 2012, http://www.rockinst.org/pdf/workforce_welfare_and_social_s ervices/2012-06-Performance-Based_Contracts.pdf. 52 Overholser, 2013. 24