New risk-forecasting tool reinforces racial disparities and emphasizes future risk in criminal sentencing.
This past July, Pennsylvania courts began using a controversial risk-assessment instrument at criminal sentencing hearings. Although this procedural milestone followed a lengthy, iterative, and highly contentious development process, the final step garnered little media attention. Nevertheless, Pennsylvania’s new sentencing process marks an alarming development in the wrong direction for sentencing reform, further toward “evidence-based” risk management and away from due process.
Put simply, the state’s sentencing instrument forecasts a defendant’s relative risk of “recidivism,” or re-conviction for a new offense, by relying on a set of static factors including the defendant’s gender, age at sentencing, and any and all adult and juvenile criminal history. At sentencing, Pennsylvania judges are instructed to review the instrument’s forecast of a defendant’s risk classification and are advised to gather more information about lower-risk defendants when determining what, if any, sentence to impose.
The instrument relies heavily on analysis of a large set of administrative criminal history collected by state courts to profile defendants, namely by inferring that defendants will follow a similar trajectory as individuals with similar demographic profiles and criminal histories.
Even though the instrument is authorized for only limited use and is not explicitly factored into sentencing outcomes, Pennsylvanians should be wary of this development. The instrument adds a step to the sentencing process that exclusively conceives of a defendant as a future danger to society, a determination based on the criminal trajectories of other individuals. And the instrument informs judges of the defendant’s risk classification before sentencing, which may affect the ultimate sentencing determination, regardless of the instrument’s quite limited information-gathering purpose.
It has been 10 years since the Pennsylvania General Assembly passed Act 2010-95 mandating that the Pennsylvania Commission on Sentencing develop and adopt a risk-assessment tool such as the recently adopted instrument. Over the past decade, the Commission proposed several versions of the instrument.
In line with Pennsylvania’s statutory process for the adoption or revision of sentencing guidelines, the Commission published each proposal with a solicitation for public comment. Considered in light of the development and adoption of sentencing risk-assessment instruments in other states, the Pennsylvania Commission’s deliberate approach was remarkably transparent and responsive to public concerns. The Commission conducted 19 public hearings in jurisdictions across the state, gathering feedback from concerned citizens and advocacy groups. At these hearings, concerned citizens and activist organizations lobbied the Commission to revise the instrument and to remove racial bias in particular. Critics raised a variety of concerns about the proposed instrument’s indicators for recidivism and reliance on criminal history data.
In response to concerns, the Commission substantively revised the instrument several times, explaining the reasoning behind its revisions and publishing analyses. For instance, an earlier version of the instrument used arrest data as the key indicator for recidivating or reoffending. In response to concerns about due process and potential undue racial disparity in arrest data, the Commission adjusted the instrument to consider convictions when measuring recidivism instead of arrests. Although some commentators ultimately criticized the Commission for adopting the revised instrument, critics nevertheless credited the Commission with its unusually transparent process and revision of the instrument.
Yet despite this transparent process, the instrument is not guaranteed to be fair or unbiased. One reason for concern is that Black Pennsylvanians are greatly overrepresented in the data used to develop the instrument—and this overrepresentation can be explained by reasons other than a greater propensity to reoffend.
For instance, earlier proposed iterations of the instrument factored in a defendant’s county of residence as an indicator for risk of recidivism. Black people in Pennsylvania are much more likely to live in Pennsylvania’s urban hubs, Philadelphia and Allegheny County. Both of these areas have high concentrations of law enforcement officers, increasing the likelihood that residents may have contact with the criminal justice system.
While Pennsylvania’s Black population is concentrated in heavily policed neighborhoods and urban municipalities, white Pennsylvanians, by contrast, are much more likely to live in rural counties that do not have a full-time dedicated police force, instead relying on the Pennsylvania State Police for law enforcement. By design, state police do not have the capacity, focus, or presence in local communities in the ways a municipal law enforcement agency would, and state police do not even enforce local ordinances.
Disparate policing in counties with larger Black populations results in the disproportionate arrest and incarceration of Black people. In 2019, Black people accounted for 46 percent of Pennsylvania’s prison population, compared to an estimated 12 percent of residents in Pennsylvania generally. And this overrepresentation in prisons is not a new phenomenon.
Ultimately, it is impossible to know how much the uneven distribution of law enforcement in the Commonwealth contributes to racial disparities in administrative data. But in response to concerns about potential racial bias stemming from the county indicator, the Commission removed that factor from the instrument’s formula.
Even though the process of developing the instrument stands as an exemplar of thoughtful and responsive bureaucratic practice, the Commission’s adjustments cannot fully address concerns about skewed data. And no adjustment to the formula can address the critical flaw of the instrument, which is an object lesson in well-intentioned but fundamentally misguided policy reform.
The most insidious problem with the adoption of the instrument is that it prioritizes a problematic, secondary focus of Pennsylvania sentencing: risk-based punishment of individuals, expressly because they are predicted to be dangerous in the future. Although risk-based punishment is not a new phenomenon, it undermines the irreducibly retributive logic of the criminal process. As a general principle, Pennsylvania cannot just take people to criminal court and punish them for posing a risk of future criminality. The determination that a defendant is guilty of some codified criminal act is a necessary precondition of criminal sentencing proceedings in the first place. And—in a well-functioning system, at least—that determination is reached only after prosecutors have developed compelling evidence that the defendant’s actions satisfied all statutory elements of the specific criminal violation beyond a reasonable doubt.
Furthermore, Pennsylvania’s sentencing scheme is expressly designed with “a primary focus on retribution.” Other common considerations during sentencing, such as rehabilitation, deterrence, and incapacitation, are secondary. Sentencing courts in Pennsylvania are explicitly instructed to focus on a defendant’s blameworthiness to “provide sanctions proportionate to the severity of the crime and the severity of the offender’s prior conviction record.”
Yet actuarial instruments that forecast future criminal behavior or dangerousness are not built to do retributive work. The state’s new risk-assessment instrument was developed using norms and statistical methods traditionally used for forecasting the classification and treatment of individuals who are at risk of adverse outcomes based on factors outside of their present control, such as age and gender. Typically, such instruments are based on the premise that outcomes are more a matter of demographic destiny than individual deviance.
Pennsylvania’s sentencing risk-assessment instrument is no different in this respect. This fundamental fact makes such a tool an inappropriate decision-making aid for those determining how to punish individuals adjudicated guilty of specific criminal acts. Although the instrument may be valuable in an appropriate, non-punitive context, lawmakers should reconsider the statute mandating its use at criminal sentencing.
This essay is part of a series entitled Racism, Regulation, and the Administrative State.