« What is a Drug Crime? Part I | Main | Lose the Wig »

Monday, April 29, 2013

Graham and the Problem of Predicting Lifecourse Offenders

Adam Liptak has a depressing piece up in today's New York Times about how states are respodning to the Supreme Court's decision in Graham v Florida, which declared life without parole sentences for minors unconstitutional. As Liptak explains, there are two contradictory ways to interpret Graham: under what I will call the Literal interpretation, LWOP is impermissible, but sentences that act like LWOP sentences in everything but name are okay; while under the Spirit (of the ruling) interpretation, ultra-long sentences are impermissible, whether we call them "LWOP" or something else.

Liptak's point is that many states are following the literal interpretation, giving minors covered by Graham sentences such as 95 or 100 years. Technically these are not life sentences, since the defendants could live for a long time (putting aside the fact that lifespans are shortened in prison and the current life expectancy for a 15-year old male is only 76.3). I feel like I am back in my 1L property class, learning about the Rule Against Perpetuities and debating about fertile octegenarians.

The following graph--I love graphs--which I have stolen from Robert Sampson and John Laub, shows the problem with the Literal interpretation:

SampsonLaub

This graph shows the likelihood of arrest by age over a person's life for several types of offenders. This comes from a longitudinal study of nearly 500 Boston youth that two sociologists, Sheldon and Eleanor Glueck, started in the 1940s. Other studies using different populations return similar patterns of lifecourse offending. The only departure from what I'm showing here was in a study (paywalled, alas) of 5,000 people in Holland, in which about 2% remained persistent, high-level offenders (at least with respect to property crimes) into their 70s.

This graph is a strong argument against persistently long sentences for juvenile offenders, since a vast percentage of people age out of crime.1 Why send an 18-year old to prison for 60 years if he is likely to quit offending in 12 or 15 years? Looking at that graph, even a 10-year sentence seems to run the risk of imposing more punishment on average than is needed.

The obvious rebuttal is "selective incapacitation": if we are good at predicting who will reoffend, we should lock those people up for long periods of time. In fact, the argument could go, the reason we see the declining age profile of offending may not less because people age out of crime, but because we are good at locking up persistent offenders, so as a cohort ages its offending propensity drops due to the removal of high-risk offenders to prison.

But this simply isn't the case. Empiricists cannot predict when someone is young which trajectory he will follow. So there is no reason to assume that courts, prosecutors, legislatures, etc., are getting it right.

Take Sampson and Laub's experience. They have data on offending for the each of the 500 men in the Glueck sample over his entire life; they look just at each man's juvenile data, try to create an predictive model to say which youth will follow which path, and then compare the actual outcome to the predicted. And two of the best scholars of the age-profile path of offending repeatedly fail. We simply aren't skilled enough yet to make these kinds of life-course predictions.

This failure isn't surprising. As the Dutch study indicates, only a small fraction of men are persistent, high-lifetime offenders. In such low-baserate situations, the risks of false positives are great. Here is a good summary of the base-rate fallacy, which basically warns that when baserates are low, then even highly-accurate tests--and we don't even have that here--are often wrong quite a lot more often than they are right.

Thus the long-term locking up of juveniles, given the general trends in desistance and our inability to identify persistent offenders in advance, seems to be a woefully inefficeint approach.

Of course, "woefully inefficient" isn't a Constitutional standard, so this is not to say that the Supreme Court should necessarily save the states from their own mistakes (although I personally think the brain-development argument banning long sentences for juveniles is compelling). But it is nonetheless depressing to see states continue to impose harsh sentences whose long-term deterrent/incapacitative effect is surely dwarfed by the long-run lock-up costs being imposed on future taxpayers.

 

1I'm making a utilitarian argument here. If your position is that from a retributive perspective, regardless of current brain development or future offending risk/behavioral norms, a murderous child deserves lifetime incarceration, then there is nothing I can say to dissuade you.

Posted by John Pfaff on April 29, 2013 at 02:01 PM | Permalink

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c6a7953ef017eeaad7fff970d

Listed below are links to weblogs that reference Graham and the Problem of Predicting Lifecourse Offenders:

Comments

What, exactly, do you think the "brain science" has shown and how much weight do you think is appropriate?

Posted by: flop | Apr 29, 2013 8:50:34 PM

I know you're focusing on the life expectancy issue, but the more important issue is that these "de facto life sentences" typically arise from multiple counts. Liptak ignores this problem, even though it's clear from the decision in the case he discusses.

The offender committed multiple counts of rape. He also kidnapped and robbed. It's not as if a judge gave a flat 100 years. The judge gave multiple sentences of 10 years or less.

The alternative is that a juvenile has a maximum, perhaps 40 years. If he rapes 20 women, he can only get two years for each. If he keeps a woman locked up and rapes her 20 times, same thing.

I strongly doubt that's a better alternative. We should at least be talking about this as an issue as important as life span.

Posted by: passing by | Apr 30, 2013 10:45:01 AM

Post a comment