It is very easy to look at pay for success as the next great innovation in public service. It’s also easy to look at it as the next great weapon of Wall Street greed. That’s the problem. There are still far too many unknowns about the real potential of these projects, and it is far too easy to write off early failures as proof that these concepts will never work. And while the concept is interesting, the current process of studying and marketing these opportunities is being executed in a wrong way.
Small number of deals
Although they are certainly gaining traction, the limited volume of pay-for-success deals in this country (and the world) is an issue. Since they were introduced in 2010, there have been roughly 10 social impact bonds created each year, and have now expanded to 15 countries. Each deal takes months of planning, feasibility studies, and collaboration among at least three different parties. Much of this planning goes around underwriting the actual services and how their results will lead to financial rewards—results based on very specific programs offered by very specific organizations. While appropriate for a piloting process for this mechanism, this is a massively inefficient way to build momentum for a scalable model.
Additionally, most of these deals take 3-7 years to provide any sort of finished program data. So even if all these initial deals become major successes, it would take years to prove that this mechanism should be scaled up (not to mention the potential disaster if even some of them fail). At this pace, it will take decades to build any substantial database around program outcomes that can be leveraged on a larger scale. Successful programs will submit their program data and performance metrics, as well as how they map their activities to outcomes, but no two service providers necessarily records results in a uniform manner. There is no GAAP for nonprofit activities, which will make any comparison among programs time consuming, if not impossible.
All results are not created equal
The most influential factor of a program’s success is the provider’s execution. But throw in any number of external variables and the numbers can be thrown off and outcomes not achieved—even if the results are still better than a baseline scenario. You never know when the economy is going to turn upside down—or when riots will erupt in the streets—or when a new virus spreads across the country. But when any of these things happens, they can influence results for any types of programs or activities.
Imagine a health initiative designed to decrease the number of hospital stays in a local community. The program performs as expected, closing in on the desired numbers. Suddenly a new virus sweeps across the community, wreaking havoc on the reported numbers. Despite the program’s real success (real being defined as net difference between reality and a baseline scenario, which in this case must be altered to include the virus), the deal will be designed to not reward investors.
Who gets the credit?
When pay-for-success initiatives succeed, the provider is often recognized as the reason. Future funders will look to fund this same provider because: A) they’re clearly effective at what they do, and B) they have experience measuring a very specific set of outcomes now (reducing the need to pilot more studies and set up another organization for similar data capture). This newly deemed “successful” provider essentially owns a monopoly over an outcome and will be able to scale up; but it gets no easier for similar organizations to do the same unless they copy the methods of the first one, which may divert them away from activities they traditionally excel at. Does program success merely lock in future funding to a specific set of providers?
Alternatively, when these initiatives fail, it is often an indictment of the flaws found in these types of financing mechanisms altogether. It is much easier to blame the mechanics of the deal than to blame the execution of the program or flaws in data analysis. This creates a discouraging environment to pursue these types of arrangements, further contributing to the long timeframe of putting together a solid base of data and understanding needed to scale these things up.
There must be a better way
Instead of waiting for some magical database to be populated decades from now, what if we started with a more generic outcomes database? The fundamental goals of most nonprofits (and governments for that matter) are similar in nature, and they all strive to achieve similar outcomes. There is no reason these outcomes can’t be defined in a way that maps them to specific activities. These activities, in turn, can be mapped to resources that can perform them. This simple map allows the alignment of activity performers to outcome providers.
The hard part is quantifying the relationship between activities and outcomes. What sort of graduation rate improvements can be achieved if 50% more kids went to pre-school? What sort of recidivism rate improvements can be achieved if 50% more prisoners went to job-training programs? And so on…
Basic government budget and audit data provide some baseline of resources expended on public activities. Other public records such as labor, census, public health, and crime data provide some baseline of outcomes. Connecting the dots between activities and outcomes becomes simpler when you look at it this way. Simple statistical analyses could possibly provide correlations between resources spent and outcomes, allowing for some sort of predictive calculation to be done (e.g., 10% more per-pupil spending leads to 10% better college enrollment rate).
Fund outcomes…not organizations
Assuming some database like this was created, nonprofits could be publicized on a more universal basis. Their activities will be on display and attached to outcomes without even providing any program data. Of course, as program data becomes more available, their inclusion in this sort of database would only add to the calculations’ accuracy, making it an even more powerful tool.
If a potential social investor wanted to fund a health initiative in his hometown, he could search this database for specific outcomes (life expectancy, quality of life, etc.) and a list of activities will show up (extracurricular education, preventive screenings, etc.). The investor could then select whichever activity fits his mission and a list of possible providers will appear. This process allows funding to flow through to outcomes rather than specific organizations—not the other way around, as is the case in setting up a traditional pay-for-success deal.
Looking ahead
Building a database is not easy. Sure, there are plenty of public data sources, but they are often inconsistent or outdated. An activities-to-outcomes map is very conceptual in nature and may require assumptions and arbitrary values in the beginning. But as programs pile up and more data become available for inclusion, the accuracy of the database will improve. And then it can truly be a sustainable and useful tool for providers, governments, and funders alike.