Main content

Just the Facts: Post-Conviction Supervision and Recidivism

In the federal judicial system, recidivism among people on post-conviction supervision is down, and signs of rehabilitation in that population are increasing.

Background

On any given day, more than 130,000 people convicted of federal offenses are serving terms of probation, parole, or supervised release. Federal probation officers are required to monitor and help improve the conduct of those people by using a variety of social work, treatment, and law enforcement strategies that are consistent with the conditions of supervision imposed by the court. The desired outcome is that the person under supervision desists from criminal and anti-social behavior. The opposite is recidivism, when the person under supervision commits new crimes or otherwise violates the conditions of supervision. 

Recidivism can be measured in different ways, but for the purposes of this report it is defined as a person under supervision incurring: (1) a felony-level new charge or (2) a revocation of supervision by the court for noncompliance. A conviction is not required for the new charge because disposition data is not always available. Only felonies are considered because misdemeanors and infractions are not uniformly reported, and what constitutes such an offense in one jurisdiction may not be a crime in another -- for example, possession of marijuana for personal use.

This report focuses on trends in recidivism among persons who entered federal supervision from fiscal years 2005 through 2016, and follows them for three years, which is about the average length of a federal supervision term.

Facts and Figures:

• The reasons why a person may return to crime while under supervision are varied and complex. Research has shown that certain factors are more correlated with recidivism than others. Leveraging that research, the Administrative Office of the U.S. Courts has developed an actuarial tool called the Post-Conviction Risk Assessment (PCRA) to help officers identify and respond to specific risk factors of persons under supervision.¹
• The PCRA takes into account 15 discrete static and dynamic risk factors, aggregated into five areas or “domains.” The domains are criminal history, education/employment, substance abuse, social networks, and attitude/cognition.  
• The initial administration of the PCRA categorizes offenders into levels. From lowest risk to highest risk, these are labeled minimum, basic, elevated, or intense. Reassessment at a later time shows whether the factors that tend to produce criminal behavior are changing over time and in which direction, and how this change relates to subsequent reoffending behavior.²
• As shown in Chart 1, most offenders entering post-conviction supervision fall within the two lowest PCRA risk categories: minimum risk (39.93 percent) and basic risk (36.62 percent).
• Chart 2 shows the decline in rearrest rates -- the percentage of people under supervision who are rearrested for a non-minor charge within three years of entry into supervision. The data are adjusted to control for the risk levels of the persons under supervision.
• However, there is still an elevated risk of reoffending in even the lowest risk group, which is evident when that group is compared to the general U.S. population. Chart 3 shows that the average rearrest rate of the lowest risk group remains nearly three times higher than the average arrest rate of the U.S. population between 2009 and 2013.
• At the same time, as Chart 3 shows, both national arrest rates of U.S. citizens and rearrest rates of those under supervision are trending downward.

The trends in recidivism are the result of the federal probation system’s consistent effort to apply evidence-based principles to policy and practice decisions.

• Officers routinely refer offenders to education programs, employment programs, substance abuse and mental health treatment, and other services that address factors associated with offending behavior.
• At the same time, officers work directly with offenders to increase awareness of problematic thinking patterns and reinforce changes in thinking and new behaviors that indicate improved decision-making skills.
• The PCRA is one measure of progress while under supervision. Chart 4 shows that for those with no change in any of the dynamic PCRA domains when reassessed, 31.5 percent are predicted to reoffend within 12 months. In general, a higher percentage of those showing increases on any of the PCRA domains when reassessed are rearrested. For those whose PCRA domains decreased, a lower percentage were rearrested within 12 months of the PCRA reassessment.
• Positive changes in behavior as reflected by a decrease in the PCRA upon reassessment are associated with lower levels of risk for rearrest and may indicate the positive effects of the supervisory intervention.

Chart 1

Source: Post-Conviction Chiefs and Deputy Chiefs Administrative Meeting (CDAM) Reports for Persons Under Supervision FY 2004 – FY 2016.

Chart 2

Source: Post-Conviction Chiefs and Deputy Chiefs Administrative Meeting (CDAM) Reports for Persons Under Supervision FY 2005 – FY 2013.

Chart 3

Source: Post-Conviction Chiefs and Deputy Chiefs Administrative Meeting (CDAM) Reports for Persons Under Supervision FY 2005 – FY 2013, and Crime in the United States, FBI, CY 2005-CY 2016, https://www.fbi.gov/, accessed online 7/9/2018.

Chart 4

Source: Cohen, Thomas H., Christopher T. Lowenkamp, and Scott W. VanBenschoten. Examining Changes in Offender Risk Characteristics and Recidivism Outcomes: A Research Summary. 80 Federal Probation 2:57-65.³

---

¹ Johnson, James, Christopher T. Lowenkamp, Scott W. VanBenschoten, Charles R. Robinson (2011), The Construction and Validation of the Federal Post Conviction Risk Assessment (PCRA) 75 Federal Probation 2 (P. 3).

² Cohen, Thomas H., Christopher T. Lowenkamp, Scott W. VanBenschoten (2016), Examining Changes in Offender Risk Characteristics and Recidivism Outcomes: A Research Summary 80 Federal Probation 2 (P. 57).

³ This is a simplified version of a more complex model that can be seen in Cohen, et al.