Latest posts by Sanjay Basu (see all)
- Converting EpiAnalysis for new media - Sep. 26, 2013
- Industry ties to medical expert panels - Aug. 26, 2013
- The Stanford Health 4 America Fellows Program - Aug. 07, 2013
- The Disabled States of America: regional disparities in healthy life… - Jul. 31, 2013
Government budgets are tight during the recession, with cuts to public health budgets being announced on almost daily basis. What strategies are available to enhance revenues for public welfare programs–for the kinds of health and education expenses that won’t “pay for themselves”(at least in the short term), and therefore are often the first to get slashed in hard times? Raising tax rates among the wealthy, and introducing new taxes like a Robin Hood Tax, have been widely discussed. But some researchers have also studied entirely new revenue-generating strategies for social welfare programs that don’t rely on taxes—including a popular pay-for-performance scheme based on “social impact bonds” (SIBs).
How they work
A SIB is one of many “payment by results” plans. Just like other types of bonds (for instance, the municipal bonds we invest in to fund a local community college), SIBs involve private investors paying for a particular program that funds some social welfare operation. But SIBs are organized such that if the social welfare program is successful, there should be some net savings to the government and benefits to society.
For example, if a public health program prevents diabetes by successfully sustaining a weight loss intervention, the government should save money that would have otherwise been spent through Medicaid or Medicare on future hospitalizations caused by diabetes. As part of a SIB, the government agrees to pay a portion of these savings back to the investors who funded the weight loss program. And just like any investment, if the program fails, the investors lose money—theoretically attracting investors towards the most effective social welfare programs.
The SIB mechanism has a few interesting limitations and poses some potential adverse consequences. First, there must be some way that the government and investors can agree on the savings rate from the social program. Hence, in our diabetes example, it may be challenging and debatable how to determine the number of diabetes cases that would have occurred if the program hadn’t happened (i.e., what is the counterfactual? How can we really know what would have happened if the program hadn’t existed?). Similarly, methadone replacement programs may reduce incarceration rates and unemployment or disability expenditures for the government, but someone needs to determine how many heroin users would have gone to prison or claimed unemployment and disability checks if the methadone program hadn’t existed. Any estimate of these end-points could be highly contested. Hence, SIBs may be limited to situations where the savings generated by a prevention program are very proximal in time to the intervention, and discrete enough to easily catalogue.
But SIBs also have an attractive market power. In theory, if social welfare programs are ineffective at improving outcomes, then investors won’t invest in them–reducing the prevalence of ineffective government programs and improving funding available for good programs. SIBs also allow providers of the social program to be paid in advance, relieving the up-front infrastructure costs of a social program from the government, and shifting some of the risk to private investors. The providers of the services don’t bear financial risk, allowing non-profit organizations to potentially try higher-risk projects than they would have otherwise, even during critical periods like the recession when government grants are hard to find. For example, if food bank programs are especially needed during the recession, then food service organizations could find investment funds even when government grants decline due to budget deficits.
This all sounds good in theory, but we’ve heard about market failures (the inability of the market to capture what is socially optimal), and in the case of SIBs it’s not clear that every good social welfare program will result in a net downstream cost savings for the government. Food bank programs, for example, may not result in any ultimate cost savings. Similarly, it’s commonly known that disease prevention programs may not really result in an economic cost savings–they can delay diseases better than preventing them altogether, and often result in the disease manifesting in older age (a social benefit, avoiding disability among younger people), but do not necessarily reduce ultimate medical costs to a significant degree (a subject that generated much contention during the recent US healthcare reform debates).
One additional concern is the idea that SIBs may undermine the role of democratically-elected government officials and their programs in favor of a privatized patchwork of non-government organization initiatives. While an ineffective government program at least has the theoretical possibility of being reformed (as elected officials can be voted out of office), an entirely private set of operations will likely be governed by whatever interests are most concerning to wealthy investors. This means that social welfare programs may be redirected towards only “popular” initiatives—e.g., children’s programs may receive many investors, but not the methadone program that public health officials determine necessary based on their higher degree of information about public health needs than private investor groups—or whatever programs can be most easily shown to produce dramatic cost savings, even if they are not the most socially-important programs in a given community.
Evidence to date
SIBs are a novel strategy, and therefore little data is available from which to evaluate them so far. One of the first SIB implementations took place in 2010 in the United Kingdom, where the Ministry of Justice created a SIB to reduce recidivism among prisoners. The Ministry allowed a financial intermediary called Social Finance to raise several million pounds of investments from private individuals and charities. These investments were directed to pay for interventions among incarcerated criminals serving prison sentences of less than 12 months at one of the country’s prisons. The Ministry agreed that if reconviction rates fell by at least 10% (compared to a matched control group, for each cohort of 1,000 criminals released from the prison), the Ministry of Justice would make a payment to the investors.
An initiation evaluation of the program by an independent research body (the RAND Corporation’s European branch) identified some key early lessons from the SIB, especially with regard to its effectiveness at reducing reconvictions, increasing efficiency, and balancing costs and benefits. The RAND research revealed that investors engaging in the SIB program found contractual relationships they entered into quite complex (an understandable issue given the novelty of the program, but a concern if transparent market mechanisms are intended to clearly direct investors towards valuable and important investments and away from likely-unworthy schemes). Despite this fact, the RAND researchers observed that many charities who invested in the SIB actually did so with their endowments, not by giving a grant. This suggests that the charities were viewing the SIB as a form of socially-responsible investment, not simply another donation. Furthermore, the SIB attracted a number of investors who hadn’t previously given money for any venture in the criminal justice realm, suggesting perhaps that the novelty of the instrument attracted new funds and interest in the sector.
But in terms of outcomes measures, results were more modest. It may be too early to tell, but UK program turned out to be too small to deliver substantial “cashable” savings (monetised benefits). The ability of the SIB model to ultimately lead to identifiable savings for the government has yet to be shown. And to maximize the ability of the project to detect an effect, there was much debate; the development of a methodologically robust outcome measure for the SIB, which had the confidence of all stakeholders, was a time-consuming and analytically complex process according to RAND’s technical report. It was difficult to establish that any statistically significant impact was achieved, and choosing a measure that could show statistically=significant changes would itself be time-consuming and potentially expensive to catalog (as those who participate in randomized controlled trials can attest).
A risk in the scheme is that providers of the service focus on members of the target group who are the easiest to help, i.e., cherry picking. In the case of the UK SIB, a comparison with a control group was required to determine how much impact the program had, but this cannot be rolled out nationally if the program is successful and meant to be available for everyone; hence future SIBs must other ways of measuring counterfactuals (for example, before-and-after measures) to determine how much investors should receive in return.
Lessons for other “social entrepreneurship” operations
The SIB is like several other “social entrepreneurship” projects—they all sound good in theory, especially using market forces to influence theoretical choices on the part of investors and implementors—but their implementation in practice is fraught with real obstacles and severe constraints. It’s not clear if we like these schemes simply because they sound like novel applications of capitalism—something other than traditional government. But if that’s the case, we may risk creating a patchwork of dozens or hundreds of different social welfare programs with competing overlapping priorities and disparate outcomes measures that are difficult to assess and harder to sustain investments over time. Like microcredit programs and “conditional cash transfer” schemes, we ultimately need to evaluate SIBs against their true counterfactual: to compare their outcomes against the outcomes from government social services that receive the same amount of money. Otherwise, we risk introducing a fad that ultimately undermines traditional approaches to reducing poverty, inequality or social problems–reducing government oversight and control while claiming to solve chronic problems faced by social welfare programs that may simply have been insufficiently funded.
It’s nevertheless worth conducting well-controlled evaluations of these projects to see which have merit. In the US, Social Impact Bonds have been called Pay for Success Bonds, and a report from the Center for American Progress has analyzed their potential. There are several other reports on the theory behind SIBs, such as Social Finance’s Technical Guide to developing these bonds, the Young Foundation’s proposal to expand these bonds, and Impact Economy’s review of the bonds in the context of global philanthropy. They all make interesting philosophical reading; we’re just waiting to see whether they can ultimately report real results…
This post first appeared on the blog EpiAnalysis.