Can the Peterborough social impact bond be judged a success?

12 Aug 2014 Voices

Peterborough Prison, site of the first social impact bond

Last week the first social impact bond, at Peterborough prison, announced initial results. David Ainsworth examines the implications.

A couple of months ago I wrote a really very long assessment of the pros and cons of social impact bonds, and I wasn’t necessarily planning on writing another one. But given that we now have the results from the first tranche of prisoners from the pilot SIB, it seems sensible to look at whether it’s been a success.

First of all, a recap. Skip this part if you’ve read about the Peterborough SIB before.

What is the Peterborough SIB?

The basic model of a SIB is that the charity agrees with a public funder that it will try to solve a problem – in this case, reducing the number of crimes committed by short-stay offenders leaving Peterborough prison. If the charity succeeds in reducing the problem, the public body – in this case the Ministry of Justice – will make a payment. Otherwise, it pays out nothing.

But the charity doesn’t take the risk. That’s taken by social investors, who front the cash to deliver the service. In this case, 17 charitable foundations jointly put up the money, around £5m. They will make a profit if the intervention works. If the intervention fails, they will lose out.

Whether the intervention works or not, with a SIB the delivery organisation is financially secure. In this case, the delivery organisation is the One Service, a special purpose vehicle set up by Social Finance, the organisation which created the idea of the SIB, which works with a number of charities.

The SIB was intended to work with three tranches of 1,000 offenders, but will be halted early after two. This is because of a new national programme, Transforming Rehabilitation, which also uses payment by results, but based on different metrics.

What do we know now?

Okay, that’s out the way. So what do we know now we’ve seen the SIB results?

Well the headline findings of the recent results show that the SIB reduced reoffending by 8.4 per cent (from an eye-watering 155 reconvictions per 100 prisoners to 142 reconvictions), which isn’t enough to trigger a payment to the investors. That needed a 10 per cent reduction. But Social Finance is pretty confident it will generate greater reductions next time. Even if it stayed at the same level, that would be enough for investors would get paid next time because only a 7.5 per cent reduction is needed over all tranches to trigger a payment.

Of course the other news is that they’re stopping the SIB two thirds of the way through, which isn’t a great advert for the model. But I’m fairly convinced that it isn’t practical to run the SIB alongside TR, so while it’s unfortunate that it’s stopping, it’s also probably inevitable.

(Whether TR is a good idea is a very different question. My guess would be that it’s based on a good principle which is going to be badly applied, and it won’t work. But that’s another story.)

The fact that investors haven’t made any money yet looks like the key fact to take from the results, but I’m not sure it’s that important. It doesn’t tell you whether the SIB delivered useful outcomes, just whether they hit a pretty arbitrary target.

The pricing structure for the first SIB was inevitably pretty finger-in-the-air, because no one quite knew what could be achieved. In order to tempt government into putting up the money, the first SIB made some stretching promises. And even then the Big Lottery Fund had to promise to put up much of the cash to pay for it.

As far as I can see, future SIBs have been priced in a way that’s more generous to the investor, and indeed, several of them have already paid out.

So perhaps this SIB was priced ambitiously and government will have to adjust its pay scales. But there’s plenty of room between the money government saves and the price it paid this time. Government can clearly make SIBs work if it’s prepared to pay the right price.

Does the model work?

The question is not whether this particular SIB made a profit for investors – although I daresay it will – but what it tells us about the viability of the model as a whole.

I’ve said before that a SIB is actually two separate things – a service delivery model which gives charities unusual freedom and a funding model intended to move risk away from them. Can we now say that both of these have worked?

One thing to say is that it’s quite hard to measure whether this SIB’s succeeded. Numbers of convictions actually slosh about quite a lot naturally, it appears. This could be down to changes in the law or the economy, to changes in regional culture and demographics, or factors in the local criminal justice system such as the individual judges and magistrates.

Social Finance, the inventors of the SIB and the people running the Peterborough project, believe that Peterborough prison seems to have its own kind of reoffending microclimate which they don’t quite understand.

These changes can swamp the difference a charity intervention makes. All of this makes government quite reluctant to pay out without a high burden of proof.

Still, the stats show this SIB has reduced reoffending, compared to a control group. How reliable that is, with reference to the above, is not clear. The statisticians say there’s only a 90 per cent confidence interval. I’m not a statistician, so I don’t know exactly what that means, but I think we can probably say the delivery model, at least, is effective.

Of course this was the first SIB and as a result it’s probably quite atypical. This is the one where they made all the mistakes, but it’s also the one that got all the time and effort. It was designed to a very high specification. The outcomes to be measured were well-calibrated, and the public body sponsoring the programme put a lot of effort in. It had a lot of legal support.

All of this can’t be replicated every time. Some of the other SIBs in existence are not such high-quality work. I suspect it’s hard to get the right metric to measure success. I suspect it’s hard to stop the commissioner trying to direct the service. And I suspect it’s really hard to actually help the people you’re supposed to help. I think in some of the other SIBs, at least one of these things has not happened.

Still, this SIB offers evidence that getting good people working on the same project for a long period of time, with freedom to pick the best interventions, is an effective tactic. You’d hope this doesn’t come as too much of a shock, but it seems to have surprised some people.

Funding models

So... good delivery model, but I think everyone in the charity sector already liked long term delivery with secure funding before we had SIBs. The more contentious question is whether the SIB is also a good funding mechanism.

To be honest, there are problems. First, we don’t know if there are enough people willing to become social investors to make it work. Also it’s expensive to set up, and it needs a lot of expertise to make it work. Also, there are ethical concerns. It distorts behaviour, particularly if your metrics aren’t good, and can lead to creaming off the easy to help and parking the hard cases. Also, like it or not, it marketises disadvantaged people. I suspect if I was a disadvantaged person, I might put up with being marketised in exchange for some help, but I also find it pretty easy to understand why some people find it distasteful.

On the other hand, there are big pluses. First, it move risk away from charities towards professional risk takers. At least it does if done right. PBR has had mixed results so far in moving the risk to the right place. Just look at the Work Programme.

Also, it has the potential to bring in new types of capital. It’s already got some cash out of investments in the stock market and into charity, although the amount remains tiny compared to the world of investment, and it’s mostly from charity investors.

Most importantly, though, outcomes funding incentivises effective commissioning.

Now we know that the SIB’s holistic style of intervention works – long-term, connected to all stakeholders, focused on a broad outcome. In an earlier piece, I suggested maybe we should keep the delivery model and just grant fund it.

This met with a lot of support, but I don’t think government is going to do it. Outcomes finance – whether a SIB specifically or something similar – seems to trigger a willingness in government to then take a hands-off approach to bodies they fund. We’d like them to take the same approach with contracts or grants, but they won’t.

I don’t know that public bodies are completely wrong, either.

After all, they do need to protect their money, and it takes a lot of trust just to say to someone – ‘here’s some money, get on with it, we trust you’, which is effectively what they’d do if they grant-funded something like the Peterborough project. “Poor implementation always trumps a good model,” says Steve Goldberg, an American social investment consultant. Meaning that it doesn’t matter how good your intervention is, if your people are rubbish.

But if you only pay if the outcome is achieved, there’s no trust necessary. So long as you’ve designed your metric right – and that’s a big if – you either get what you want and save money, or you don’t get what you want and save nothing. Also, you pay when the saving’s achieved, not when you agree the deal. And in government, that’s important.

So whether the sector likes outcomes finance or not, it might be the price it has to pay to get a good delivery model.


More on