Public Safety Committee on June 5th, 2014

[The Chair, Conservative, Daryl Kramp]

Colleagues, we'll call meeting number 28 of the Standing Committee on Public Safety and National Security to order. Today we are following up with a further meeting and discussion with witnesses as we continue our study on social financing.

For the first hour of testimony and questioning today, we have two witnesses before us. We have, in person here, representing MDRC, David Butler, senior adviser. We also have by way of video conference from Cambridge, Massachusetts, from the Harvard Kennedy School, Mr. Jeffrey Liebman, director, social impact bond technical assistance lab.

Did I get that one correct, sir? Do we have you with us, loud and clear?

[Prof. Jeffrey Liebman, Director, Social Impact Bond Technical Assistance Lab, Harvard Kennedy School]

I'm here. Can you hear me?

[The Chair, Conservative, Daryl Kramp]

Absolutely. That's tremendous. Thank you very much. We do certainly appreciate your taking the effort over the long distance to communicate with us this way.

We will give each of you up to 10 minutes to make a statement. After that, for the balance of the hour, we will have questions from our committee members from all sides of the House here.

We will start with Mr. Butler.

You have the floor for up to 10 minutes, sir.

[David Butler, Senior Adviser, MDRC]

Do you mean I could have done this by video and stayed in my bed and hung out at home? You guys didn't tell me that.

[The Chair, Conservative, Daryl Kramp]

Well, let's put it this way. You're warming the chair now, so carry on.

[David Butler]

Okay.

Just quickly, because MDRC isn't a self-explanatory thing—although the letters are the name of the organization—we're a not-for-profit social policy research organization. We've been around for 35 years or so. We basically evaluate and do demonstration projects of interventions to help low-income individuals and families. Our projects involve either evaluating what someone else has developed—we're real believers in very rigorous evaluation, random assignment evaluations—or developing an intervention ourselves, with help, and then studying that. Those are the two arms of our business.

I'm here to talk to you today about the first social impact bond, in New York City, to be implemented in the United States. There were lots of discussions, and probably a number of them should have started earlier, but we were the first ones off the block. It's a project that's providing cognitive behavioural therapy to 16- to 18-year-olds in New York City's Rikers jail, which is the largest jail in North America, I believe, with a huge population, and a huge population of adolescents.

Before I get into a little more of the detail of the program, I'll make a few introductory remarks. I know there's been lots of hype in the debate around pay for success and the social impact bonds. In the States, when we first started the project and there was nothing else happening, that's when the debate was at its fiercest, because usually people debate when they know less about something. When there's less information, there's more to argue about.

In those early days, we heard two...and I took questions about both of these alternatives. There were those who were positively hyping this as a transformative strategy that would get new money from investors for preventive programs, increase government accountability, save money for government and taxpayers, and improve outcomes for at-risk populations. It was a “quadfecta” winner, so to speak, in horse-racing language.

On the other hand, I heard other things when I would talk about our project, that it was just a cynical strategy to privatize services, to polish the sorry reputation of banks while ripping off government and the taxpayer, and leave at-risk populations worse or no better off—a quadruple loser.

Now, I'm not an ideological type. I'm a program evaluator, so I try to be more like my man Friday on Dragnet, “Just the facts, ma'am.” But I had to deal with those questions early on. Fortunately, I think things have changed a little bit now that we've actually started to do some of these things. I think the consensus in the States more is “Let's see whether any of these things really can produce what they claim they can produce, and make our judgments there.” So I'm facing a little bit less both the more strident criticisms and the promotional statements, and facing more the “Well, let's wait and see, give this thing a chance, and see what happens.”

I have to say that we at MDRC, as evaluators, were also a little reluctant to get into this, to start. Evaluators look at programs with a very rigorous lens, and we don't see a lot of things that really work. There aren't a lot of programs out there that are very successful. So then how does one attract an investor to take the gamble on something that may not have a great likelihood of being successful? The only way they will is if they're getting a high rate of return on that investment and there's the ability, or the willingness, of government to pay high interest rates to banks.

It seemed to us that this probably was not going to happen, so we went in with a fair amount of skepticism but with the open mind of an evaluator of let's give it a chance and see what happens.

Now we're about three years into our project, which is up and running. The measures we use to see how we're doing indicate that we're on track to achieve the goals of the program, but I can't tell you that I know that this will happen. The goals of the program are to reduce recidivism by 10%, at a minimum, for our population. Recidivism means return to jail. In our case, it's compared with a historical comparison group of the same age. I don't have that data yet. All I have is the data about the program participation, and the program participation is on track to achieve those goals if the connections between those things play out as they're expected to do.

Our partners in the project are Goldman Sachs. They are providing a loan to pay the cost of this program, a loan that's secured, however, by the Bloomberg Family Foundation. We're the intermediary, so we are the folks who are responsible for kind of pulling the deal together, and more critically I think, for monitoring the ongoing performance and selecting the program intervention.

We have two well-regarded service providers in New York, Osborne Association and Friends of Island, which run the actual intervention. Then we have a partner in the City of New York, through both the department of correction and the mayor's office, which is also involved. It's a partnership with multiple players who have different perspectives on things. We've managed to work reasonably well together. The negotiations and getting the deal together were complex things. It was probably more costly for us than we had anticipated. There were transaction costs around working with a bank. We had to bring in our own Wall Street lawyer so we understood the language. We had to build trust. Everybody kind of had to build trust. But we got to a point where we got the thing up and running. We mounted it. We did a pilot stage. We've run it incrementally, and it looks as through it's going reasonably well.

It's a social impact bond, so if it is successful—and that means if it achieves at least a 10% reduction in recidivism for our population—then the investor gets paid back. If it goes beyond the 10%, investors can earn some interest, up to a rate 20%. That's the interest, and it's capped at that level. We'll see whether we actually get there. A third party evaluator, the Vera Institute, a very well-respected criminal justice evaluating organization, is evaluating the program. We're going to look at impacts on recidivism at one year and then at two years. An important thing, I would say, about social impact bonds, which may be somewhat different from prior pay-for-performance models and something that we think is very important, is that the measure of success is not outcome. It's not achieving a predetermined outcome. It's achieving a predetermined impact, meaning that the program does better than it would have done without the intervention, or the population does better. So that means you have to establish some point of comparison, either a control group if you're doing it really scientifically and rigorously or some legitimate comparison group.

To me that's a great development. For many years we had pay-for-performance contracts that were based on outcomes, but outcomes are easy to manipulate and easy to gain. If you serve a population you think is more likely to succeed, you will get those results, but they don't really tell you if you've made a difference or not. So the idea is that we have a net impact evaluation. I think most of the social impact bonds being proposed do have that kind of evaluation as well.

Another thing I would just say about lessons is that you face the challenge of having a restrictive contract that has some pretty clear terms. When you try to impose that on a program that has to operate in a very flexible changing environment, reconciling those two things can be difficult. In our case, one of the big challenges is that we had a predetermined number of folks we needed to serve, but at the end of the day, the number of people who are in the jail system is not within the control of the program or the control of the Department of Correction. So we built in all kinds of formulas for how to deal with reduced numbers, but it meant going back and renegotiating and rebudgeting. That was a complication. I think we figured out how to do it, and there certainly is some advantage to going in with something that says, “These are the goals you have to meet, and you have to stay on point on this”. But it is a tension, and it is a challenge in the program.

I think I can stop there. I'll just say that we are scheduled to have the interim results a year from August and then our final results in 2016. The payback will occur at the end of 2017. One of the tricky things about these projects is that you have to accumulate the success, the impact over time, before government will have achieved enough savings to be able to pay back the investor.

The last thing I want to say is that I think it's a mistake to think about social impact bonds or pay for success as limited to only projects that result in cost savings. So far, primarily that's been the focus to date, but there are a lot of other goals that in many cases are more important to help the population we care about and that may not lead to cost savings. So—

[The Chair, Conservative, Daryl Kramp]

Excuse me, Mr. Butler. Thank you very much. We appreciate that. Certainly I know that you have more to offer here, but I'm going to cut you off there, sir. You're a little over your time.

[David Butler]

That's fine.

[The Chair, Conservative, Daryl Kramp]

I know you'll have an opportunity to further your comments during the questioning from our members here.

We will now go to Mr. Liebman.

Thank you very much for your contribution here today as well, sir. You have the floor for up to 10 minutes.

[Prof. Jeffrey Liebman]

Thank you, Mr. Chair and members of the committee, for inviting me to join you remotely here today.

I know that you've heard several testimonies previously on social finance broadly, and also on the pay-for-success or social impact bond technique. What I have to offer is that for the last two years I've been leading a team that has been providing pro bono assistance on the government side of these projects to eight state governments and two city governments around the U.S.

Our role basically is to help these governments do proper cost-benefit analysis, think hard about what kinds of evaluations they want to set up, and, I guess more importantly, decide when it makes sense to do a pay-for-success project and when it doesn't. There are certainly as many projects that we've analyzed and we've recommended and that people have decided not to do as ones they've gone forward with.

The governments we're working with that are going forward on these projects are using this tool to adjust a wide range of policy areas, including early childhood education, homelessness, diabetes prevention, and most commonly, recidivism for ex-offenders. What I'd like to do with my few minutes here is to reflect on two things, really, that I think most of the U.S. governments using this tool are trying to accomplish.

The first is to do a better job of matching the right services to the right clients. The second is to generate evidence about which programs actually work and should be expanded, and which ones don't and need to be either reformed or replaced.

As you probably already know, under the most common social impact bond model, the government contracts for social services from a local service provider, or sometimes a team of providers, and the government pays entirely or almost entirely based on the results that are achieved, such as a 10% increase in employment, a 30% reduction in recidivism, or a 50% reduction in emergency room visits. The performance is rigorously evaluated, just as Mr. Butler said, by comparing the results of the people being served to the results of some sort of comparison group, so that impacts can be assessed.

If the program fails to meet the minimum performance targets, the government and the taxpayers don't pay, and then payments increase above the minimum threshold, up to some pre-agreed maximum. Sometimes these projects generate cost savings to the government in terms of reduced spending on remediation and paying for data outcomes. That can offset the full cost of these projects. Sometimes they offset some of the costs of the projects, but not all.

As an example that's related to this committee's work, consider a program that works with young men exiting the juvenile justice system. We know that without transitional support something like 60% of these young men end up back in prison within five years. In addition to the social cost of the new crimes they're committing, these high recidivism rates lead to big fiscal costs for the government, because the government ends up paying money to incarcerate the individuals.

Under a social impact bond model, the government contracts with the social service provider to help support these young men as they transition out of the juvenile justice system. The government then tracks the recidivism rate—the percentage of folks who were served in the program and who end up back and incarcerated—and compares that to a group of individuals who were not referred to the program. Then, at a predetermined time, they assess how the intervention did relative to the counterfactual and make payments based on the avoided bed-days of incarceration and perhaps on other things, such as the additional tax revenue that comes from increased employment.

Under this model, there's generally a lag of several years between when the services are delivered and when the results can be measured, and therefore, payments made to the service providers. The role of private investors in these projects is simply to bridge this gap. Private investors provide the operating funds to the social service providers so that they can deliver services up front and then wait to get repaid down the road after results have been demonstrated.

In the United States, there are four social impact bond projects that are actually delivering services already. Three of them are criminal justice programs working with ex-offenders to reduce recidivism and also to increase employment. The fourth is a preschool intervention. Let me talk in some further detail about one of the recidivism projects, the one in New York state.

The New York state project is targeting high-risk offenders coming out of the state prisons, and it's triaging and taking basically the half of the population coming out of prison that's at highest risk of reoffending. It's then connecting those folks, the highest-priority individuals, with the most intensive services. The particular program they are being connected with is delivered by the Center for Employment Opportunities, which delivers about four months of engagement with transitional and subsidized jobs, and then transitions people into permanent jobs. The basic model is to try to get someone into a job as soon as possible after they are released from prison.

The key thing here is that, as you know, in any sort of high-priority population, there's a range of different individuals who probably need a range of different services. The goal of the New York state project is to use data to try to figure out which individuals should be connected with which services, and in particular if there are some services that are the most expensive to provide and the most intensive interventions, you want to connect those to the highest-risk individuals.

This project is being piloted through a social impact bond project, but the hope is that, even beyond this project, this model of matching the right clients to the right services will get expanded, and it won't be just this one particular service. They will add other service providers and get better at figuring out which clients should be directed to which services over time.

What I think we've observed across the 10 governments we've been working with over the last couple of years is that there are three ways in which the social impact bond model is improving government performance.

The first thing it's doing is improving government decision-making. The SIB model brings important market discipline to government decisions about which of all the interventions out there should get scaled up and provided with more resources, because only interventions that have a strong enough track record to be able to attract the interest of private investors can get funded by this model. Projects that don't have a talented enough management team or don't really have evidence behind them are less likely to get resources.

The second thing I think we're seeing with this model is that it's helping governments shift resources from remediation towards prevention, so rather than waiting for bad outcomes down the road and incurring the budgetary costs for those, this model is allowing governments even in tight budget times to make additional investments, and preventive investments, which, in addition to providing better outcomes, will hopefully down the road yield some budget savings because we won't be paying for remediation.

The last thing this model seems to be doing—and I think this is actually the most important thing—is enabling governments to enter into much more effective multi-year collaborations with service providers to tackle tough social problems. It's really hard with traditional budgeting and contracting techniques to tackle a problem that takes a sustained effort over four to six years to solve. The political leadership that starts an initiative turns over. Annual budget cycles make it hard to make sustained commitments to providers, and just in general, it's hard to get everyone focused and headed in the same way on a sustained basis towards achieving a particular outcome.

These contracts seem to be enabling governments to do a more effective job of working on complex problems with private sector service providers, and to do so with a four-to-six-year time horizon, which is often how long it takes to actually make a dent in a social problem.

In conclusion, I want to say that we're really still experimenting with this technique. As I said, there are only four projects up and running in the U.S. We're working to develop another 12 right now with our partner governments.

It's important to realize that this tool is a very good fit for certain circumstances, but there are a bunch of places where it would also be a bad fit. So you have to be very careful to use it where it is a good fit and not where it's a bad fit, and you also have to design the projects carefully. It would be quite possible to design a project through which the wrong population was being served, for which the service provider was able to cream-skim and not target the highest-need individuals, and for which you didn't measure the right outcomes, and you could really, I think, do some damage by using this model if you didn't implement it well.

The key lesson I want to leave you with is that this tool, around the U.S., is allowing some innovative governors and mayors to tackle some problems that they don't think they can tackle with conventional mechanisms. But it's very much still an experimental tool that we are learning how to use, learning whether or not it truly will give us better results than the other tools within government tool kits.

Thank you. I'd be happy to take any questions.

[The Chair, Conservative, Daryl Kramp]

Thank you so much.

Mr. Liebman and Mr. Butler, we certainly thank you for your contributions today.

We'll now go to a round of questioning, and we'll start off with Mr. Richards, please.

[Blake Richards, Conservative, Wild Rose, AB]

Thank you.

Thank you to both of you for being here—or virtually here.

Most of my questions are for Mr. Butler, but I'd like to start with you, Mr. Liebman—or is that “Dr.” Liebman?

[Prof. Jeffrey Liebman]

Professor, Dr., Mr.; whatever you want to call me is fine.

[Blake Richards, Conservative, Wild Rose, AB]

Okay.

Dr. Liebman, you mentioned that you played a role in evaluating a number of potential projects for various levels of government. I assume that you've probably also done a certain amount of academic research, looking at this model as compared with more traditional models. Is that accurate? Is that something you've conducted in terms of academic research as well?

[Prof. Jeffrey Liebman]

Yes. The research is ongoing, because we're testing the model—

[Blake Richards, Conservative, Wild Rose, AB]

Fair enough.

[Prof. Jeffrey Liebman]

—but I'm definitely thinking about that question.

[Blake Richards, Conservative, Wild Rose, AB]

You mentioned the three areas in which you see this as being quite useful, or potentially quite useful. One of them was in prevention. The focus of our study here is looking at the idea of social financing in regard to crime prevention in particular, so I'd like to focus a little bit on that side of things, the prevention side.

You mentioned four projects that you're part of in some way and that are ongoing currently. Can you share with us any of the initial results from those particular projects or any of the research you've done in terms of comparisons with the more traditional models?

[Prof. Jeffrey Liebman]

That's a great question. The key thing is that all of these projects are pretty new, so in terms of their impact on the ultimate outcomes, we still have to be patient. You serve people for a couple a years, you wait a couple of years to see if they reoffend, and then you know what happened. We're still in that period.

I've learned three things about this model, though, that I didn't know six months ago. The first is that it is actually possible to put one of these projects together. Mr. Butler was describing how, when they started down this road, they just didn't know whether it was actually going to be possible to raise the investor dollars, given the risk. We've now seen that it actually is possible. If you start down this road and you put together a good project, you actually can get it off the ground and provide services. That's the first thing we learned.

The second thing we learned was that the governments that set up these projects decided to do more of them. One possibility was that they'd go through this process and say, “All right, we did one, but I'm never going to do another one again. This was too hard.” But that's not what happened. Both New York state and Massachusetts, when they started delivering services on their recidivism projects, immediately announced that they would do more of these projects. They found them to be tools that were allowing them to do things they just couldn't get done with conventional tools. Massachusetts said it was going to do a homelessness project and an adult literacy project, and New York state said, following on their recidivism project, they were going to do a diabetes project and an early childhood project. That's the second thing we learned.

I think the third thing we're seeing in the projects that are on the ground—I'd love to hear if Mr. Butler agrees—is that it really is the case that because there is this private attention to these projects, when you hit the bumps in the road that you hit in any project, you see an urgency in solving those problems, which, at least in my two stints in government, I didn't always see as being something that was easy to get the public sector to achieve.

So I do think we are seeing this focus on outcomes and the money at stake actually having an impact on how the services are being delivered on the ground.

[Blake Richards, Conservative, Wild Rose, AB]

Great. Thank you very much. That was very helpful.

Mr. Butler, I'd like to ask some questions of you as well. You talked about your specific project, the Rikers Island project. You talked a lot about what was involved in setting it up, etc. Can you give us a sense of the timeline involved in getting that all set up? I don't think I heard you mention how long it took.

If I could, then, I'd ask you to specifically mention the amount of time it took you to find investors and then find the service providers that those investors would.... Obviously they have to create some comfortableness with each other, and I'm wondering what timeframe was involved in that. Now that you've built that framework, will it be easier and quicker to set these up going forward?

[David Butler]

Before we got involved in the project, the City of New York was already thinking about doing this. There was some preliminary work before we got involved, which made our lives somewhat easier, at least faster. Nevertheless, I would say that from the time we shook hands with New York City, it was about a year before we got the thing up and running on a pilot stage. A lot had to go on in that year.

The most important thing, probably, was to figure out the costs and savings, the term sheet, which is the critical piece that you have to pull together for this thing. We didn't want to talk to investors until government had blessed what we had come up with. Fortunately New York City has good data about the costs of beds and the costs of recidivism, plus we did some pretty good diligence around what interventions were returning on investments that were cognitive behaviorally based.

By putting those two things together—but it still took probably two to three months—we came up with something that we could take to the market and say that we thought this was credible. It was definitely worth testing, and we thought we could possibly get these results. Once we started, we did a pilot for about four to six months in which we ran the project at half scale. I have to say that even beyond the pilot phase, we've still been tinkering. Unless you do something at scale, you can't really....

So, yes, there was a pretty long overture before things started.

[The Chair, Conservative, Daryl Kramp]

You have 30 seconds, Mr. Richards.

[Blake Richards, Conservative, Wild Rose, AB]

Do you feel that now that the framework or that template has been put in place, it will be easier and quicker to put these in place in the future, now that you have that experience?

[David Butler]

Yes, I think so, with one caveat. I'm also saying that these projects all look somewhat different. It depends on the kind of intervention you're doing, who the partners are, the mix of philanthropies and private employers, but I would say that in general, yes, we can do it more quickly.

[The Chair, Conservative, Daryl Kramp]

Thank you very much, Mr. Richards and Mr. Butler. That was interesting.

Now Mr. Garrison, you have seven minutes, please.