The Great California Prison Experiment

I published an academic paper back in 1996 that tried to measure the impact that changes in the prison population have on the crime rate. It turns out that this is a hard question.

You can’t just look across states and compare incarceration rates and crime rates, because the places with the worst crime problem will also tend to have the most prisoners.

Holding constant the punishments, the more crime that occurs, the greater the prison population. So correlations just are not very useful for determining whether prisons are effective or not.

“While those crime numbers sound bad, according to my estimates, letting out the prisoners is more or less a wash from a societal cost-benefit perspective.”

What you need is something more like a randomized experiment in which, for extraneous reasons, a whole bunch of prisoners get let out (or a whole bunch of extra criminals get locked up). For obvious reasons, the National Science Foundation would never give me a research grant to do that myself.

Indirectly, however, the A.C.L.U. (or in the California case, another prisoners’ rights group) does the next best thing. It brings lawsuits against state prison systems arguing that the prisons are so overcrowded that they represent “cruel and unusual” punishment. The A.C.L.U. virtually always wins these suits. The state appeals, and roughly a decade after the suit is filed, the court’s initial decision is upheld and the A.C.L.U. is victorious.

As I report in my paper, these lawsuits have a large impact on the prison populations in the affected states. After the suits are filed, but before any court decisions are handed down, prison populations grow more slowly in the litigation states.

The preliminary court decision doesn’t have much of an effect. But when the final verdict is handed down, prison populations shrink by about 15 percent relative to the rest of the country over the next three years.

Yesterday, a prisoners’ rights group won a preliminary decision against the state of California’s prison system. Consistent with my earlier results, the lawsuit already seems to have had some impact on California’s prison population. For instance, in 2007 California’s prison population shrank by about 1 percent, whereas the overall U.S. prison population grew by nearly 2 percent. It will take a few years before a final court decision is handed down, but the likely outcome is that five or six years from now there will be 25,000 fewer inmates than there otherwise would have been.

What does this mean for crime? If my estimates are correct, ultimately violent crime will be roughly 6 percent higher in California than it would have been absent the lawsuit. That is roughly 150 extra homicides a year, 500 additional rapes, and 4,500 more robberies.

While those crime numbers sound bad, according to my estimates, letting out the prisoners is more or less a wash from a societal cost-benefit perspective. The money we save from freeing the prisoners is on the same order of magnitude as the pain and suffering associated with the extra crime.

I do have one very specific policy recommendation to the state of California. If they do a mass release of prisoners, it should be done with strings attached. Namely, if the released prisoner gets convicted of a crime again in the future, his sentence the next time around should be whatever it normally would be plus all of the time that he should have served on his current sentence that gets cut short because of the early release.

This rule would strengthen the incentives for the ex-cons to stay straight. Italy enforced such a policy after a mass release, and it appears to have been quite effective.


Fred T.

"The money we save from freeing the prisoners is on the same order of magnitude as the pain and suffering associated with the extra crime."

I'm pretty sure those extra 150 homicide victims would disagree with you.

Tom Anderson

Do we get any explanation whatsoever for the opinion that there will be more homicides, rapes, and whatever? Suppose the state of California decides to release 57,000 marijuana smokers. Do they suddenly turn into violent criminals? I think I'm missing the genius in this post.

Robert

"The money we save from freeing the prisoners is on the same order of magnitude as the pain and suffering associated with the extra crime."

Hi Steve,I've found this statement quite surprising and can't help but feel very curious about how "pain and suffering" can be measured against economic benefit. Would you mind expanding on it a bit more? Or did you mean that savings are proportional to pain and suffering?

Thanks!

misterb

Dr. Levitt,
Are you assuming that CA releases prisoners randomly? With the huge numbers of non-violent offenders incarcerated for drug offenses, it would seem that CA could comply with the court orders without releasing any murderers, rapists or armed robbers. If that's the case, why would these non-violent offenders turn violent just because they've been released? Or do they get released to become victims? If that's the case, they were probably subject to more criminality in jail than they would be out of it, but I doubt that your statistics include crimes committed on prisoners.

oddTodd

If it is a wash from a societal perspective, but clearly a big loss for the individuals affected by the crime, society should just build more prisons. The societal cost is zero, and we have fewer individual victims.

mrem

anyone interested in why the U.S. prison population is booming and what our nation is reaping as a result should check out the forthcoming book "Do Prisons Make Us Safer? The Benefits and Costs of the Prison Boom" It is think with facts, but absolutely essential reading. https://www.russellsage.org/publications/books/080912.222969

Mark Brucker

I'm wondering if SL's analysis might be incorrect for 2 reasons. One is that the proposal is that some of the prisoners would be shifted to county jails, not released. Second, those released are intended to be the least dangerous. Actually, there's a third factor. Part of what they are talking about is increasing incentives for good behavior. I would think it's possible that could have positive effects on their future behavior in prison and outside. So unless the first 2 were factored in, I'd suspect his estimates are overly pessimistic. I don't know if there's any way to assess the third factor....

Rob

but what if you only release non-violent criminals as a part of the release? can't you optimize in that respect?

Glen

1) "I've found this statement quite surprising and can't help but feel very curious about how "pain and suffering" can be measured against economic benefit. Would you mind expanding on it a bit more"

Its relatively easy, Courts do it all the time in wrongful death suits

2) If it is a wash from a societal perspective, but clearly a big loss for the individuals affected by the crime, society should just build more prisons. The societal cost is zero, and we have fewer individual victims

How on earth is the societal cost zero? You realize this would (typically) involve the State spending millions of tax payer dollars to build, maintain, fund, and staff these prisons, right?

travis ormsby

@ #5, oddTodd,

The societal cost of building new prisons is hardly zero. Either taxes will have to go up, or resources will have to be shifted from other programs. The people of California will pay a high price for more prisons.

Nick F.

You mention 500 additional rapes. Seeing as prisoner on prisoner rape goes largely unreported, I wonder how many rapes are avoided by decreasing the prison population.

Justin James

"The money we save from freeing the prisoners is on the same order of magnitude as the pain and suffering associated with the extra crime."

What in the world is wrong with your thinking? Are you really this insensitive? In the field of psychology, this kind of thinking (equating financial cost with rape and murder to balance the two) is considered a severe mental illness.

In fact,it is *precisely* this kind of thinking which creates crimes like murder and rape in the first place. "I want that car more than he deserves to live" and the like.

J.Ja

Jason Goodman

"I'm pretty sure those extra 150 homicide victims would disagree with you."

I'm pretty sure they wouldn't. Their next of kin might, but the dead aren't talking.

Regarding the "selection effect", in which prisons choose to only release the least dangerous criminals: Unpunished criminals tend to get worse over time: today's back-alley mugger may become tomorrow's hit man. Levitt's data based on observations of real prison releases: one can presume that if the prisons he's studying did *not* select their releasees carefully, his prediction would be even more dire.

Quill

Those 5,150 victims and their family members would probably rather pay higher taxes than be assaulted or robbed. Unless they're Ron Paul supporters, of course.

Thomas B.

"This rule would strengthen the incentives for the ex-cons to stay straight."

This is how prosecution already works. You build a history on a criminal, escalating punishments and reducing your willingness to cut deals all the time. The criminals know that when a Prosecutor lets you walk with time served, it's so next time they can remind the judge of how they gave you a second chance and request the statutory maximum.

Daniel C.

Some of the commenters think there might not be an increase in violent crime if California only releases non-violent offenders. Dr. Levitt's estimate is based on his paper studying 12 other states that had court-imposed prison population limits. Presumably those other states released non-violent offenders first, so unless California has a much higher ratio of non-violent to violent offenders in its prison population, we probably can't expect a better result.

Also, the observed increase in violent crime following court-mandated prison population caps is a correlation, but not necessarily a direct causation; i.e., that the released prisoners are themselves committing the additional crimes. There may be secondary causes. For example, less fear of incarceration for drug possession may encourage drug use, meaning there's more money in drug dealing, and more incentive for the violent crimes associated with high level drug dealing.

Read more...

asmart

This is an important, perhaps overlooked, consideration:

Steven's paper investigates crime rate as a function of prison population, and (controlling for external factors) seeks to describe this relationship in terms of an elasticity. I have no qualms with the "method" of control. However, I would argue that the crime rate-prison population relationship may be best described by two elastic coefficients (one for prison population increases and one for reductions) rather than one.

It may be a mistake to view prison population as a "state" variable, as is apparently the convention, whereby the crime rate function increases and decreases along a single curve (such that an increase in crime rate due to doubling the old prison rate can be effectively reversed by halving the new prison rate). When this is the case, a study investigating prison rate reduction -- e.g. Steven's study -- is essentially interchangeable with one investigating increase.

I would argue, however, that the process is not reversible -- an ex-con would presumably find it more difficult to live a normal life post-prison than pre-prison. Aside from the fact that prison may have acted to harden the criminal, he would also likely find greater difficulty getting a job and acceptance into mainstream society. It would then seem likely that the marginal "crime costs" of reducing a prison population would be less than the benefit of increasing the prison population, and the elasticity of increase would be less than the elasticity of reduction. Although I'm not completely familiar with the paper, this may help to explain why Steven found a larger elasticity than in previous studies.

The situation at the Guantanamo prison is perhaps a relevant model, and an interesting case study, for this concept.

Read more...

asmart

mistake in my previous post. Should have read: "...likely that the marginal 'crime costs' of reducing a prison population would be LARGER than the benefit of increasing the population..."

Eric M. Jones

I think it should be mentioned that the more money put into child welfare, head-start, schooling, and various programs that send kids in a low-crime direction, the less we have to pay to house adult prisoners. I suspect that putting money into the former is 1000000X more cost-effective, given that prisoners usually don't pay taxes.

Bucik Batard

With the unemployment situation, this program is setting up many people for failure. With no jobs, what are they to do? Many will turn back to drugs which would be much less likely to happen were they all able to find employment.

I agree with the program, but the timing is terrible.