Reducing Recidivism Through Incentives

Ryan Bradley, writing for CNNMoney, highlights an interesting policy experiment currently underway in New York City: a social impact bond geared at reducing recidivism:

They are called “social impact bonds.” The first, issued in 2012 by Goldman Sachs (GS), is underway in New York City for $9.6 million. The money is going toward a four-year program to reduce reincarceration of juveniles at Riker’s Island prison. Goldman Sachs has a vested interest in the success of this program. If participants stop returning to jail at a rate of 10% or greater, Goldman will earn $2.1 million. If the recidivism rate rises above 10% over four years, Goldman stands to lose $2.4 million. In a recent report, the Brennan Center for Justice at NYU School of Law calls this a “bet on success … instead of using the typical model of privatization, in which private prisons generally bet on failure (i.e. the more prisoners, the better).”

Bryan Stevenson, the founder and executive director of a nonprofit that, among other things, helps former convicts avoid reincarceration for minor parole violations, believes the idea could be “transformative.”  

“What if you personalize it?” he asks. “What if individual officers, and administrators, personally benefitted by reducing recidivism rates? If you created the right kind of bonus, something that was really meaningful, it would just be fascinating to see how quickly things would change. Guards would be desperate to get educational facilities, they’d want classes, drug treatment, safe environments. Their interests would align with the prisoners’ — they’d care if people came out. They’d care about who shouldn’t come out. It would have an impact on how they think about their work, their life. And you — we, taxpayers, the system — wouldn’t be spending new money, you’d just reallocate.


Perhaps we could apply the same concept to private schools AND public colleges. Tie funding to success rates.


That sounds a lot like the incentives that lead to the cheating teachers in Chicago.


Which leads straight to the core problem: any measurement used to reward people that can be gamed, will be gamed.

The question is: can recidivism statistics be gamed?


No way they can stop reincarnation. If I want to come back to this life as a carnation, that's my business!!

Voice of Reason

Why don't people give up the façade that incarceration is supposed to be correctional? It is, and has always been a deterrent against crime, and a way to keep the criminals off the street. You don't make people follow the straight and narrow path by putting them with other criminals.

nobody special

Later, clever investors start to put pressure on all parts of the criminal justice system, speeding up a case here, slowing down another there... all based on the timelines assigned to the bonds when they were issued.

I realize how cynical this sounds, but for 4.5 million dollars I'm sure someone can figure out a way to put their thumb on the scales.

Enter your name...

It'd be far simpler to change the drug laws to prefer treatment over incarceration. Give everyone a single opportunity to try treatment, regardless of previous conviction status. Voila, a short-term reduction in reincarceration rates.

Similarly, they almost certainly won't track kids out of state, so recommend that any teen with an out-of-state parent be moved out of state.


It's great to give incentives to institutions, but the prisoners' choices are the only ones which will determine whether they are re-incarcerated or not. Are there any incentives aimed specifically towards the prisoners? Other than taking their freedom away. Because, you know, the incentive to remain free didn't seem to work the first time around...
So many facets to be considered, but still, an interesting thought.

Enter your name...

Most prisoners are there because of impulsive choices they made. However, there are relevant choices you can make that have a useful effect. For example, you could "choose" to teach inmates how to make better choices and to manage their emotions, so that they will be less prone to making the kind impulsive choice that results in criminal charges.


You could add another level of market incentives by having prisons bid for prisoners. Whichever bids the highest reincarceration penalty (or smallest non-reincarceration bonus) gets the prisoner.

You'd probably want to to have the state bias the bidding process somewhat in favour of putting the prisoner in commuting distance of friends and family to facilitate visits.

(I'm just pointing out this idea, not advocating for it. I'm actually quite unhappy about prisons-for-profit.)


I am extremely intrigued by the second proposition of tying guard’s benefits/compensation to recidivism and am a proponent of trying this in jails. Unlike prisons that hold longer term sentenced populations, jails are primarily county run facilities housing pretrial defendants, relatively short-term felony offenders, or those awaiting transfer to prison (about 20% actually continue to prison). Because it’s operationally difficult and politically challenging to offer programming in jails, most time is spent watching TV and interacting with guards. Guards, who also spend their long days all year in the same depressing facility we hope deters return visits, quickly get calloused and often inject little confidence and hope into those cycling through. Since these are all government pension employees, it seems very possible to create an incentive based approach tied to the recidivism rate. Hopefully, the nearly 34,000 released daily from jail will not return at the current ~50% rate.


Marian Kechlibar

What about removing the "permanent record" element? In the US, when you are convicted of a felony, you remain a felon for life, with many civic disabilities. Hell, even a simple arrest will stay on your permanent record and show up on many occassions. The sex offender registry is even worse, as the original intent mutated into something monstrous.

I think that adopting the principle of some foreign systems, where most criminal records (not murder) are removed from your record once certain reasonably long time elapses, could help. At least the people would be employable again, could vote, own guns etc.

steve cebalt

The proposed measurements on Figure 5 of Page 29 of the PDF report is work a look before criticizing this. The proposed goals make much more sense than the current ones, and are just as easy to measure.


If the parole officer has an incentive to reduce recidivism rates, wouldn't he avoid arresting his parolees when they reoffend?