Pop Bonds

by on March 15, 2012 at 7:34 am in Data Source, Economics, Law, Political Science | Permalink

Pay on Performance Bonds incentivize private-sector creativity in the performance of public goals. One of the first Pop bonds (also called social improvement or social impact bonds, SIBs) was pioneered by the British government and the UK group Social Finance. The UK Pop bond is designed to reduce prisoner re-conviction rates. Social Finance raised about $8 million from investors to fund a variety of programs for released prisoners, helping them to find work, stay off alcohol and drugs, reintegrate with society and so forth. The programs are managed by a group of non-profits. The UK government has agreed to pay the investors a return but only if reconviction rates are 7.5% less than those of a control group. If reconviction rates fall below the target level, the investors will earn a good rate of return, 7.5-13%, depending on how far rates fall below the control, but they could also lose everything if rates do not fall. The Pop bond issued in 2010 and appears to be going well although no (potential) bond payments are scheduled until 2014.

A Pop bond puts little risk on governments, who pay nothing if the program does not work but who save money if the program does work. With less at risk government should be willing to experiment more and try new approaches to problems. By contracting out, the government also eliminates a public bureaucracy resistant to change. Most importantly, a Pop bond encourages creativity and innovation in social programs. Investors in a Pop bond have an incentive to monitor the groups implementing the programs and to ensure that they choose the very best, most cost-effective programs. The better the program works, the more the investors earn. If Pop bonds expand it may even pay investors to undertake their own experiments to see how best to maximize their returns.

For Pop bonds to work it is critical that outcomes be measured and marked to an appropriate, randomized, control group. If not carefully monitored, the private sector will also excel at innovative and creative gaming at the public expense (see the comments for some suggestions).

More Pop bonds are being planned in the UK and the idea is also catching on in the United States. The Department of Justice and the Department of Labor both have pilot programs in the works and Massachusetts has issued a request for proposals. By the way, Pop bonds are said to be a new idea but the U.S. bounty hunter and bail bond system which works very well is a clear precursor as is the system of privateering.

Pop bonds have the potential to produce public goods with private innovation; they are an idea worth watching.

Nick March 15, 2012 at 7:59 am

Won’t prisoners in the trial group expropriate a large portion of the gains?

This would seem to be the result game theory would suggest (Ultimatum game)

Not sure how repeated games work…

Cliff March 15, 2012 at 10:31 am

How would that work, exactly?

Neal March 15, 2012 at 8:03 am

You know, comparing pop bonds to bounty hunting and privateering may not produce the desired positive associations …

Rahul March 15, 2012 at 8:03 am

Ha, wonder if it might be interesting to track death rates in the group……..

BC March 15, 2012 at 8:28 am

“Ha, wonder if it might be interesting to track death rates in the group……..”

Too suspicious; instead, go short and pick off the control group.

Bernard Guerrero March 15, 2012 at 12:04 pm

Just what I was thinking….

Frederic Mari March 15, 2012 at 8:19 am

Yeah, my first reaction was – Hmmm, off-ing the guys in your group might allow you to drop the re-offending rate… Also, less drastic, what about sending/icentivising/forcing the ex-cons to foreign shores? They might re-offend but at least they won’t be screwing YOUR numbers…

And, of course, I can see bond traders suddenly looking at social papers for modelling re-offending rates and selecting a bunch of low probability re-offenders.

And, finally, Neal is right. Didn’t we stop bounty hunting and privateering for some pretty good reasons? I mean, cutting down the Defence budget by re-creating mercenary companies of Corp/Army size and equipped with planes and warships might have unintended consequences…

Urso March 15, 2012 at 10:14 am

Conspiracy theories about murder and forced emigration aside, can’t you solve the “pick and choose” problem by simply giving them a randomized group? Or by comparing their recidivism level not against some generic baseline but against the numbers for those specific types of offenders in the program? I’m sure recidivism rates can be parsed out by type of offense and priors.

Frederic Mari March 15, 2012 at 10:39 am

I am pretty sure you could. But randomising your selection would need significant numbers. After all, when you’re buying a bond, you’re usually not so keen on it being a lottery.

As to deliberately build specifically similar control groups, that’s also possible but then it easily becomes an administrative nightmare. CDOs and MBSs with mortgages behind are an example where, like with sausages, people didn’t want to have to delve too deeply into what was going into them. Just too complicated, trust the trustees, the auditors, the administrators, the bankers, the brokers, the custodians, trust everyone but do not do the work yourself. We know the end result.

JF March 15, 2012 at 11:48 am

“After all, when you’re buying a bond, you’re usually not so keen on it being a lottery.”

Then you shouldn’t buy a bond, because that’s what every bond is.

Urso March 15, 2012 at 11:54 am

Right, if the bondbuyer insists on a guaranteed return there’s no point in the program at all.

Frederic Mari March 15, 2012 at 12:18 pm

Right. And a meteor might crash the Earth, sure. But, no, bonds aren’t lotteries. The components are known, for example and not random. Furthermore, most people want bonds for the ‘secure’ cash flows, not for capital gains. You can move up and down the credit worthiness but here you’re trying to be overly cynical and end up mostly wrong.

msgkings March 15, 2012 at 1:01 pm

+1 to Frederic

Urso March 15, 2012 at 1:06 pm

Overly cynical? Hey, I’m not the one who brought up forced deportions.

It’s true that *most* people want bonds that are secure, but not everyone. And the kind of bonds we’re talking about here are far more akin to junk bonds than to sovreign debt.

Cliff March 15, 2012 at 10:32 am

We did not stop bounty hunting.

Frederic Mari March 15, 2012 at 10:40 am

It significantly evolved, though. And I don’t think you can go to your nearest post-office with the scalp or the ears of an Arab-looking guy and say “hey, s/he was AQ, I am pretty sure. Where’s my $100? reward?”

andrew fischer lees March 18, 2012 at 11:30 pm

One does not opt-in to being bounty hunted.
Here, prisoners would opt-in to receiving bridge services.

That is no petty difference.

Rahul March 15, 2012 at 8:27 am

$8 million and 13% returns are all fine but a crucial number missing is “$-per-prisoner”. If allotted $500,000 per person it may not be hard to get him to behave; if allotted $500 might be a tad difficult.

How much is the government willing to stake?

OTOH, maybe someone should run an experiment where prisoners themselves are told “Look, behave for 5 years and we’ll give you $100,000″ or some such. Isn’t economics all for eliminating the middleman?

albatross March 15, 2012 at 2:44 pm

If this worked, it would be worth it, except for the perverse incentives for the initial crime.

Ed March 15, 2012 at 8:31 am

“Social Finance raised about $8 million from investors to fund a variety of programs for released prisoners, helping them to find work, stay off alcohol and drugs, reintegrate with society and so forth.”

Is there any money for helping people who are not convicted criminals to find work, stay off alcohol and drugs, (re) integrate with society and so forth? This was my first reaction.

My second reaction was that this scheme seems highly gameable.

Only after reading the comments did I realize that this was a bounty hunting scheme. Like much public policy of these days, its a revival of old practices that were rejected for good reasons in the last two centuries, though now revived with extra jargon.

Cliff March 15, 2012 at 10:33 am

a) It has nothing to do with bounty hunting and b) bounty hunting has not been rejected because it works very well

Steven Kopits March 15, 2012 at 8:37 am

Pay-for-peformance: Good for criminals, not for politicians?

TallDave March 15, 2012 at 8:40 am

It’s like a charter school prison system. Maybe this can do for rehabilitation what charter schools are doing for education.

Neal March 15, 2012 at 9:50 am

Nothing?

dead serious March 15, 2012 at 10:46 am

+1

TallDave March 15, 2012 at 3:06 pm

Yeah, tell that to the parents attending lotteries to get their kids in.

Ben V March 15, 2012 at 7:12 pm

have you ever been to one of the lotteries? children all gathered around waiting in dread to find out whether or not theyre doomed to a terrible public school education. its just plain depressing.

Brandon Berg March 15, 2012 at 9:18 am

It would be awkward if they managed to pull this off by hiring better defense attorneys.

Rahul March 15, 2012 at 9:27 am

….or by entrapment of the control group.

RSaunders March 15, 2012 at 9:25 am

Effectively what Accountable Care Organizations in health care will be???

Bob Knaus March 15, 2012 at 10:01 am

I know that RCTs are the Latest Big Thing, but why are they necessary in this context? They add to overhead, increase the possibilities for gaming, and decrease the pool of program beneficiaries.

Why not just pay for the desired result? If the contractors achieve it, great. If not, let the contractors figure out what they did wrong.

andrew fischer lees March 18, 2012 at 11:35 pm

Agreed, we have enough data to decently predict recidivism rate +-2SD. If the result achieved is statistically significantly less than Mean – 2SD, then the program may have had an effect. “Clinically significant” is a different story, but that could be hashed out with the investors too. There’s no reason to have a control group.

Rahul March 15, 2012 at 10:14 am

Instead of an RCT you could also think of this as outsourcing-with-a-variable-bonus-clause.

Ryan March 15, 2012 at 11:00 am

1 – Risk sharing is an idea that comes from other marriages between big business and big government. Industry says, “If it doesn’t work, you pay nothing,” but of course the metrics are always rigged in favor of whatever the government expenditure is.

2 – There is no evidence suggesting public/private partnerships reduce the size of bureaucracy. In my admittedly anecdotal experience, the opposite is most assuredly true.

3 – I have my doubts about how much competition these things create. One of two scenarios unfold: (a) Only a handful of firms are equipped to meet the needs of the government, and these firms are already well-entrenched in the bureacracy, so the bond/partnership merely strengthens the entrenchment; Or, (b) a small firm is selected from the initial RFP, the firm grows as it suckles off the government contract, and by the time the program expires, the situation is already so solidified within the bureaucracy that it is impossible to select a new vendor as a matter of pure practicality.

Maybe I’m missing the point, though.

JWatts March 15, 2012 at 12:01 pm

“but of course the metrics are always rigged in favor of whatever the government expenditure is.”

If you mean that the businesses aren’t going to bid on a contract unless they are pretty sure they can make a profit, then yes they are ‘rigged’. Why else would anyone bid on the contract?

Ryan March 15, 2012 at 12:58 pm

No, what I mean is that the performance metrics are rigged. Here’s what I mean:

Let’s assume you are proposing to the government that you should be contracted to perform Service X for Cost A. You propose that Service X is a positive alternative to Services Y and Z, which the government already funds at Costs B and C, respectively.

To win the bid, you need only demonstrate that Cost A (which is hypothetical) is less than *either* Cost B or Cost C. Even if Y is cheaper than X, if you can demonstrate that Z is more expensive, then you present a cost analysis that looks favorable regardless of where optimality actually lies.

You win the bid, and you’re set for the next 10 years. I’ve seen it a million times.

TallDave March 15, 2012 at 4:25 pm

Good point, governments are notoriously bad at this sort of decision.

It would be much better if the gov’t simply refrained from criminalizing consensual behavior.

Joshua Miller March 15, 2012 at 11:15 am

What happens when it becomes obvious that the trial group isn’t going to beat the control group? Seems likely to lead to a perverse incentives. Analogously, no one even tries to claim the bounties on some criminals: there needs to be a mechanism to renegotiate the bond when difficulties arise, like increasing a bounty for particularly notorious fugitives. But then you’re back to cost-plus contracts by another name.

Hasdrubal March 15, 2012 at 12:06 pm

I don’t think these Pop bonds are set up as bounties, I think they’re sold as bonds whose face value is paid ontingent on the outcome of whatever it’s being sold to fund. It doesn’t have to be anything to do with criminals, it could be a bridge that pays off if traffic congestion is lowered a year after the bridge is completed, for example.

The reference to bounty hunting and privateering were just examples of payment that is contingent on success, rather than guaranteed payment. Think of the bounty hunter as a contractor that invests time and other resources up front, but only gets paid if he actually catches his fugitive. Compare that to a road construction company that gets paid up front (or continuously through the project) and then delivers their work product afterwards.

Hasdrubal March 15, 2012 at 11:59 am

Isn’t there a huge incentive for the government to sandbag? After all, a lot of bureaucracy and politics is about money spent, not results delivered. So, it’s a win for your local politician to say “Look at me! I created or saved 50 jobs with that 8 million dollars for a rehabilitation center,” and not clearly a loss when the follow up is “Well, the rehab center didn’t impact recidivism, but that just means the 8 million I spent won’t come out of your taxes and we still have 50 more jobs than we used to!” That can be spun as a pareto improvement: Crime isn’t any worse and now we have more jobs (and who talks about ongoing expenses in an election campaign?)

albatross March 15, 2012 at 2:49 pm

Why not exploit the tendency for cherry-picking? Give the organization with the performance bonds control over parole decisions, with their reward per prisoner assuming they make their low recidivism rate target about the same as the cost of keeping each prisoner in prison. Unlike parole boards, the performance bonds will give the organization making the parole decisions a strong incentive to correctly predict who will and won’t reoffend.

Alan March 15, 2012 at 3:50 pm

All this shows is that George Mason University should appoint an Inquisitor. Look at how easily The Evil has insinuated itself. Did you notice what Alex said? He said that the government will need to monitor the program becasue the privatre sector might behave dishonestly. It’s statism, I tell ya! Statism! Where is the solution that doesn’t require inspectors paid for with money extorted from taxpayers?

TallDave March 15, 2012 at 4:28 pm

A private function replacing a gov’t function might need gov’t monitoring? The horror!

What other dark secrets might an Inquisitor reveal? Revelations that Alex supports the existence of the SEC or FDA?

Roger Sweeny March 16, 2012 at 8:23 am

Of course, we have a system now where the government provides the service and itself monitors the result. Namely, the public schools. Every teacher knows that you can pretty much pass or fail as many people as you want by changing how hard the test is.

TallDave March 17, 2012 at 5:22 pm

And best of all, their unions elect the politicians that run the monitors and decide how much to pay them. It’s win-win for the unions and politicians, lose-lose for the students and taxpayers.

Comments on this entry are closed.

Previous post:

Next post: