« Fourth Circuit to reconsider en banc its Whiteside ruling concerning reconsideration of guideline errors in 2255 | Main | Careful examination of California's "mixed" record with realignment »

July 14, 2014

"Moneyball Sentencing"

The title of this post is the great title of this interesting-looking new article by Dawinder Sidhu now available via SSRN. Here is the abstract:

Sentencing is a backward- and forward-looking enterprise. That is, sentencing is informed by an individual’s past conduct as well as by the criminal justice system’s prediction of the individual’s future criminal conduct.  Increasingly, the criminal justice system is making these predictions on an actuarial basis, computing the individual’s risk of recidivism according to the rates of recidivism for people possessing the same group characteristics (e.g., race, sex, socio-economic status, education).  The sentencing community is drawn to this statistical technique because it purportedly distinguishes with greater accuracy the high-risk from the low-risk, and thereby allows for a more efficient allocation of sentencing resources, reserving incarceration for the truly dangerous and saving the low-risk from needless penal attention.

Despite these asserted benefits, risk-assessment tools are exogenous to the theories of punishment, the very foundation for sentencing in Anglo-American jurisprudence.  This Article reviews the legality and propriety of actuarial predictive instruments, using these theories and governing constitutional and statutory law as the touchstone for this analysis.  This Article then applies these normative and legal principles to seventeen major characteristics that may comprise an offender’s composite risk profile.  It argues that risk-assessment instruments are problematic for three reasons: they include characteristics that are prohibited by constitutional and statutory law; subject the individual to punishment for characteristics over which the individual has no meaningful control; and presume that the individual is a static entity predisposed, if not predetermined, to recidivate, thereby undermining individual agency and betting against the individual’s ability to beat the odds.

July 14, 2014 at 03:29 PM | Permalink


TrackBack URL for this entry:

Listed below are links to weblogs that reference "Moneyball Sentencing":


So with a recidivism rate of 2 to 3%, registered sex offenders should have virtually no jail time added for prospective likelihood. Somehow, I don't think that's how it works...

Posted by: Eric Knight | Jul 14, 2014 11:05:41 PM


One is punishing for past crimes, not trying to predict the future. One is preventing recidivism in the most reliable way, by death. Start the count at the real age of adulthood, 14. No violent repeat offender makes it to 18, and the start of their busy period of crime. Any deterrence is incidental and an added benefit.

Because each conviction has to stand in for an average of 10 other crimes, by the time one gets to 3, one knows the defendant is a bad guy. His innocence of the third crime is not a problem, because ha has committed many hundreds of crimes.

Posted by: Supremacy Claus | Jul 15, 2014 1:58:05 AM

Think of it this way. The core of a correction plan has two parts. One is determinate and the other is indeterminate. Its determinate part is designed to penalize and punish. This is the stuff we call a sentence. Its indeterminate part is designed to control the offender's risk of committing another crime. The first part is based on the offender's past conduct, which does not change once established. The second part of a correction plan is based on the person's risk of committing another crime. This risk does change, sometimes for the better, sometimes for the worse.

Risk is a conclusion. We do not punish people because they constitute a risk. We manage risk.

The same kinds of strategies are used to punish and control risk; namely restraints, requirements and takings. They can be structured so that the deprivations used to punish include those used for risk management, or vice versa. In other words, the most restrictive deprivations used for either purpose should control at any given point in time. I have a couple of graphics that explain this process, which I would be glad to send to anyone who is interested.

Posted by: Tom McGee | Jul 15, 2014 1:15:35 PM

Post a comment

In the body of your email, please indicate if you are a professor, student, prosecutor, defense attorney, etc. so I can gain a sense of who is reading my blog. Thank you, DAB