I want to believe! Let me start with that confession. Our society needs pay-for-success schemes to work. We are spending too much on social programs that do not generate results, too much on high-cost treatments and not enough on lower cost and more humane prevention. With the fiscal cliff fast approaching, we must figure out how to do more with less. No wonder a growing array of technocrats, financial intermediaries, and more recently McKinsey and the New York Times are extolling "The Promise of Social Impact Bonds."
So why am I hesitating to jump on the bandwagon? In part because, despite the hype, to date there is exactly one social impact bond up and running, to combat recidivism among inmates coming out of the Peterborough Prison in the UK, and it is still two years away from generating even interim results. If the what works movement is about scaling proven solutions, there is considerable irony in so many of its participants calling for widespread adoption of a financial mechanism for doing so that is literally untested. I also question whether private capital will readily flow to underwrite the costs of supporting the most intensive users of our social safety net. Ultimately, though, my skepticism is grounded in the challenges that federal, state, and local government agencies have faced in trying to use performance-based contracts over the past two decades—and how social impact bonds may make several of these challenges worse.
Pay-for-success schemes are not new. Performance-based contracting, in which government pays not for activities (inputs and outputs) but rather for results (outcomes), was a central thrust of the reinventing government movement that began in the 1980s and was championed by Al Gore during the Clinton Administration. For all of the initial promise, performance-based contracting has not become pervasive and has really taken hold in only a few cities with strong management capacity (e.g., New York) or where legal challenges and scandals have cleared the way for it (e.g., child welfare in Illinois and Tennessee).
Social impact bonds could solve one problem that has constrained pay-for-success–the lag in time between when costs are incurred to deliver services, such as supporting and finding a home for a foster child, and when outcomes materialize, for instance the foster child remaining stably and safely housed over a given period. Government agencies seeking to pay only for results have had to fudge things by reimbursing nonprofits for select inputs and outputs along the way, enabling them to cover at least some of their costs, and then paying out the remaining portion of the contract after the outcomes have been realized. With social impact bonds, a third-party intermediary lines up the resources to pay nonprofits as they do the work and takes on the financial risk of delivering results. The intermediary is later reimbursed by the government at levels that provide it and its investors a sufficient rate of return–but only if and when the outcomes are realized. So what is not to like?
To start, all of this assumes that procurement functions in federal, state, and local government agencies will be ready, willing, and able to use social impact bonds. I doubt they will anytime soon. Another reason government contracting officials–typically not the most visionary and experimental of bureaucrats–remain wedded to using inputs and outputs as the basis for their contracts is that these things are much easier to confirm and count than outcomes. A 2009 survey of more than 600 government procurement officers working at all levels of government found that just 40-percent of them were using performance-based contracting for services. The most frequently cited barriers to increased use by respondents were "lack of trained procurement staff" (39-percent) and "lack of understanding on the part of top administrators" (28-percent). If these officials are challenged to implement basic performance-based, two-party contracts, how will they fare with social impact bonds given the third parties, extended time horizons, independent impact evaluations, and all-or-nothing payments involved?
Moreover, social impact bonds will exacerbate the government’s principal-agent problem. This problem exists even in traditional performance-based contracts, where the principal, i.e., the government agency issuing and managing the contract, has to rely from a distance on its agent, i.e., the nonprofit using taxpayer dollars appropriately (or not) to deliver services effectively (or not). Hence government’s temptation to micro-manage nonprofits, insisting that their work reflects certain inputs and outputs, even though the ostensible goal is to pay for results.
Social impact bonds will aggravate this problem in part because of the high stakes involved. As the McKinsey study points out, for social impact bonds to pay off, they need to offer enough savings from effectively delivering prevention to provide ample margins for all the intermediaries and investors involved—and so that government will find the more complicated arrangements worth the effort. This calls for engaging with high-cost users of government services such as the chronically homeless or ex-offenders. But these populations also present higher risks of harm, to themselves or others, and thus of headlines that can lead to one’s boss getting fired and one’s boss’s boss losing the next election.
The government’s principal-agent problem will be further compounded because social impact bonds create an additional layer of agents. The intermediary orchestrating the initial financing and delivery of services will now stand between government and the work it might ultimately be on the hook to pay for, not to mention the populations it has a particular interest in monitoring. The frictionless world imagined by social impact bond advocates, one in which government agencies defer to the decisions of intermediaries as they enlist and manage service providers and to the judgments of yet another third party on whether results have been realized and tax-payer dollars should be handed over will, I’d wager, prove to be just that—imaginary.
Finally, there will in almost all cases be not one but multiple government principals involved and wanting their policies and priorities reflected in a given social impact bond. High-cost populations that offer the financial upside needed to make social impact bonds feasible are typically the focus of several agencies and programs across the federal, state, and local levels of government. Most advocates of social impact bonds gloss over this complication. To its credit, the McKinsey report acknowledges that a social impact bond "will likely touch multiple programs and agencies, and requires coordination on terms and structure; agencies may resist giving up their power and autonomy," and that "repaying investors from realized cash savings may require aggregating [social impact bond] benefits across multiple agencies and programs as well as different levels of government. This could prove challenging." Indeed it could! McKinsey’s proposed solutions include a mayor or governor creating a "dedicated ‘super team’ with central decision-making authority" as well as "ensuring that government data systems are capable of tracking costs and service utilization at the client level." Especially in an otherwise detailed report, these suggestions come across as hand-waving in the face of the messy realities of American politics and government in most states and localities.
Oscar Wilde observed that "skepticism is the beginning of faith." If that is the case, I am clearly ready to believe in social impact bonds. I’d welcome feedback on what I may be overlooking or overcomplicating here. I will share what I am hearing from others on this point in future posts, as well as my take on circumstances in which skeptics might consider taking a leap of faith on social impact bonds.