Opinion, Berkeley Blogs

Scientific risk assessment in sentencing may beat the alternative

By Jennifer Skeem

This blog post is coauthored by Christopher Slobogin, who holds Vanderbilt Law School's Milton R. Underwood Chair in Law and is one of the 10 most cited criminal law/procedure professors in the nation, according to the Leiter Report.

Recently in the New York Times, Professor Sonja Starr defended Attorney General Eric Holder’s expressed hesitation about relying on risk assessment instruments to assist at sentencing, arguing that sentences based on such instruments are probably unconstitutional and likely to be discriminatory.  Her principal claim is that, in assessing an individual’s risk of future re-offending, these instruments not only rely on criminal history (which she thinks is legitimate), but also illegitimately focus on marital history, employment status, education, neighborhood, and financial background, which can be proxies for minority race and poverty — and in any event are unrelated to moral culpability for past crime.

We think caution in using these instruments is appropriate, but putting the brakes on their use, which Professor Starr seems to prefer, would be a myopic policy.  In our view [1], risk assessment should be considered in sentencing within bounds set by moral concerns about the offender’s culpability for the past crime.  For example, if the morally appropriate sentence is in the range of 5 to 9 years, then risk assessment can be used to sentence the highest-risk offender to 9 years and the lowest risk offender to 5.  Offenders should never be sent to prison for more time than they deserve simply because they received a high score on a risk assessment instrument.

Applied judiciously, risk assessment has much to offer the criminal justice system — particularly in reducing bloated prison populations without at the same time increasing the crime rate.  For example, Virginia has been able to reduce its prison population by 25% with little impact on public safety through intelligent use of risk assessment techniques. Under Virginia’s system, nonviolent offenders bound for incarceration under the state’s sentencing guidelines are assessed for risk of future offending.  Those who represent a low risk are recommended for alternative punishment like probation or community service; offenders with higher scores proceed with their sentence recommendations unchanged.  Risk assessment instruments can also be used by correctional professionals to develop individualized supervision and treatment plans that (when implemented well) help reduce an offender’s risk of recidivism.

If Professor Starr’s goal is not to limit — but rather to eliminate — risk assessment as a sentencing factor, she is bucking a long tradition that includes diversion, probation, parole, and other accelerated release practices designed to use correctional resources efficiently while protecting public safety. Professor Starr is most worried about the use of risk profiles, based on statistical analysis.  As Starr acknowledges, however, judges have always considered risk of future crime “informally.”  Without statistically validated instruments, judges must make seat-of-the-pants judgments that, because they are opaque, are virtually impossible to challenge, even when they are influenced by an offender’s race or class.  At least when an instrument is used, the criteria are transparent, consistent, and can be examined for patterns.  Furthermore, research consistently shows that predictions based on well-validated risk assessment instruments are more accurate than intuitive judgments and judgments based solely on criminal history.

Professor Starr is also concerned that reliance on risk assessment instruments will result in longer sentences for the poor and minority groups, based on factors over which they have little or no control.  That concern is well-taken.  But it leaves out half of the story. The alternative to a regime that relies judiciously on risk assessment is one based solely on the crimes the offender has committed, acts that can be just as closely related to the offender’s class and racial status as a risk assessment (see Frase, 2008 on how racial disparities in imprisonment can be attributed to sentencing guidelines' emphasis on criminal history).  Over 95% of the people currently in American prisons are poor.  A regime based partly on risk at least gives people who succeed at rehabilitation programs or who can in some other way show they do not need to be incarcerated a chance at avoiding imprisonment for the full length of their sentence.

Despite Professor Starr’s suggestion to the contrary, the Supreme Court has never directly addressed how risk should be assessed.  The caselaw she cites merely stands for the proposition that inability to pay a fine should not, by itself, lead to imprisonment.  It says nothing about whether a person considered to be dangerous because of factors that may be related to poverty can have their sentence lengthened on that ground.

From a practical point of view, the “compared to what?” question for risk-assessment informed sentencing is existing practice.  Whether the use of risk assessment instruments to inform sentencing decisions exacerbates, ameliorates, or has no effect on existing racial and socioeconomic disparities is an open question — one that we can and should address with research.  We predict that the answer to this question will vary by jurisdiction, as a function of what practices are being replaced … and how competently risk assessment is applied to specific sentencing questions.

Science does have something to contribute to justice.  Properly validated, judiciously applied risk assessment instruments can enhance both fairness and the efficient use of scarce correctional resources.  It would be unwise to discourage their proper development and use.

[1] Monahan & Skeem (2014).  Risk redux:  The remarkable resurgence of risk assessment in the criminal justice system.  Federal Sentencing Reporter, 26, 158-166.