Sentencing, by the Numbers

New York Times

ANN ARBOR, Mich. — IN a recent letter to the United States Sentencing Commission, Attorney General Eric H. Holder Jr. sharply criticized the growing trend of evidence-based sentencing, in which courts use data-driven predictions of defendants’ future crime risk to shape sentences. Mr. Holder is swimming against a powerful current. At least 20 states have implemented this practice, including some that require risk scores to be considered in every sentencing decision. Many more are considering it, as is Congress, in pending sentencing-reform bills.

Risk-assessment advocates say it’s a no-brainer: Who could oppose “smarter” sentencing? But Mr. Holder is right to pick this fight. As currently used, the practice is deeply unfair, and almost certainly unconstitutional. It contravenes the principle that punishment should depend on what a defendant did, not on who he is or how much money he has.

The basic problem is that the risk scores are not based on the defendant’s crime. They are primarily or wholly based on prior characteristics: criminal history (a legitimate criterion), but also factors unrelated to conduct. Specifics vary across states, but common factors include unemployment, marital status, age, education, finances, neighborhood, and family background, including family members’ criminal history.

Such factors are usually considered inappropriate for sentencing; if anything, some might be mitigating circumstances. But in the new, profiling-based sentencing regimen, markers of socioeconomic disadvantage increase a defendant’s risk score, and most likely his sentence.

Advocates of punishment profiling argue that it gives sentencing a scientific foundation, allowing better tailoring to crime-prevention goals. Many hope it can reduce incarceration by helping judges identify offenders who can safely be diverted from prison. 

While well intentioned, this approach is misguided. The United States inarguably has a mass-incarceration crisis, but it is poor people and minorities who bear its brunt. Punishment profiling will exacerbate these disparities — including racial disparities — because the risk assessments include many race-correlated variables. Profiling sends the toxic message that the state considers certain groups of people dangerous based on their identity. It also confirms the widespread impression that the criminal justice system is rigged against the poor. 

It is naïve to assume judges will use the scores only to reduce sentences. Judges, especially elected ones, will face pressure to harshly sentence those labeled “high risk.” And even if risk scores were used only for diversion from prison, it would still be wrong to base them on wealth and demographics, reserving diversion for the relatively privileged.

Evidence-based sentencing also raises serious constitutional concerns. The Supreme Court has consistently held that otherwise-impermissible discrimination cannot be justified by statistical generalizations about groups, even if those generalizations are on average accurate. People have a right to be treated as individuals, and individuals often do not conform to group averages.

For example, in its 1983 decision in Bearden v. Georgia, the court unanimously rejected the state’s contention that a defendant could have his probation revoked because his recent job loss increased his crime risk. The court held that “lumping him together with other poor persons and thereby classifying him as dangerous ... would be little more than punishing a person for his poverty.”

Litigation has been slow in coming, however. The risk-prediction instruments are not very transparent (some are proprietary corporate products), and defendants may not understand the role of poverty and personal characteristics. But challenges could be on the horizon. For example, I recently participated in training the Michigan defense counsel on constitutional objections to evidence-based sentencing, in preparation for the state’s impending implementation. 

Of course, judges have always considered future crime risk informally, and it’s worth considering whether actuarial methods can help make those predictions more accurate. The problem isn’t risk assessment per se; it’s basing scores on demographics and socioeconomics. Instead, scores could be based on past and present conduct, and perhaps other factors within the defendant’s control. 

Data-driven predictions grounded in legitimate factors might be about as accurate as current profiling schemes. There is no persuasive evidence that the current troubling variables add much predictive value, once criminal conduct is already taken into account. But even if they do improve accuracy, this gain doesn’t justify sacrificing fairness. 

Criminal justice policy should be informed by data, but we should never allow the sterile language of science to obscure questions of justice. I doubt many policy makers would publicly defend the claim that people should be imprisoned longer because they are poor, for instance. Such judgments are less transparent when they are embedded in a risk score. But they are no more defensible.