500 Word Discussion

52 FEDERAL PROBATION Volume 81 Number 1

Comparison of Recidivism Studies: AOUSC, USSC, and BJS

James L. Johnson Probation Administrator, Administrative Office of the U.S. Courts

RECIDIVISM IS ARGUABLY one of the greatest challenges facing the criminal justice system today. Reoffending not only has rele - vance for public safety, but has resource and cost implications related to incarceration and other criminal justice costs (Urban Institute, 2009).

For these and other reasons, recidivism rates are often used by those examining the effectiveness of criminal justice policies, evaluating program performance, and mea - suring the success of community supervision (Lowenkamp, VanNostrand, & Holsinger, 2015; Urban Institute, 2009). Detailed recidivism data can help distinguish which defendants/offenders have the most interac - tions with law enforcement and correctional agencies, which types of offenses are com - mitted by those recidivating, and the timing of reoffending (Urban Institute, 2009). But in order to accomplish all these measurement and analytic objectives through recidivism data, we must first define recidivism in a man - ner that allows it to be effectively measured. Recidivism is commonly defined as reengaging in criminal behavior after receiving a sanction or undergoing an intervention for a previous crime (Elderbroom & King, 2014; National Institute of Justice, 2014). As a conceptual definition, this is relatively straightforward; however, as an operational definition—one that permits measurement—it is not so simple (National Advisory Commission, 1973: 512).

Recidivism can be measured in a variety of ways, with the various measures setting differ - ent criteria for labeling a person a recidivist. Recidivism is generally calculated as a rate or percentage of people in a specified group who meet certain criteria in a defined span of time.

How recidivism is defined can vary simply by changing the group, the criteria, or the amount of time for which recidivism is cal - culated (Ruggero, 2015). Most experts agree that rearrests, reconvictions, and returns to incarceration during a specified period of time are the primary ways to measure recidivism (Maltz, 2001; Armstrong, 2013; Elderbroom & King, 2014; Urban Institute, 2009). Because each measure captures a recidivist at a dif - ferent point in the criminal justice system, they require different definitions. If we use rearrest as a measure, a person is defined as a recidivist if he or she has been arrested for a new crime after being released directly into the community on probation or after serv - ing a term of imprisonment. Rearrests may also include arrests for alleged violations of supervised release, probation, or parole (Hunt & Dumville, 2016). The reconviction measure defines a person as a recidivist if an arrest resulted in a subsequent court conviction. Violations and revocations of supervision are not included in reconvictions, since no formal prosecution occurred. Returns to incarceration define a person as a recidivist if a conviction or revocation results in a prison or jail sentence (Hunt & Dumville, 2016). The various definitional differences can cre - ate discrepancies among reported recidivism statistics. For example, two agencies that use reconviction to measure recidivism will pro - duce different recidivism rates if one agency includes only reconvictions for felony offenses and the other agency limits reconvictions to the same type of offense as the instant offense (Armstrong, 2013). Because varied measures are used to determine recidivism, it is difficult

to compare recidivism rates between agencies or amongst states.

Defining Recidivism Since 2010, the Administrative Office of the U.S. Courts (AOUSC) has produced annual recidivism statistics on offenders placed on probation and supervised release. Consistent with AOUSC’s systems strategy to measure and report on results of mission-critical work, AOUSC periodically publishes articles in Federal Probation that describe the most recent recidivism statistics and changes over time. Because Federal Probation is available to those outside the judiciary, the statistics published are available to both internal and external stakeholders.

Recently, other government agencies have reported on federal recidivism, describing recidivism rates higher than those reported by AOUSC in past Federal Probation articles. In 2016, both the United States Sentencing Commission (USSC) and the Bureau of Justice Statistics (BJS) released reports on federal recidivism that conveyed recidivism rates differing from those reported by AOUSC in Federal Probation articles. The USSC report examined the most serious post-release recidi - vistic event for a cohort of offenders released in 2005 (Hunt & Dumville, 2016). The BJS report examined the extent to which offend - ers placed on federal community supervision were arrested by federal and nonfederal (i.e., state and local) law enforcement agencies prior to and following their placement on community supervision for a cohort of per - sons released from prison in fiscal year 2005 (Markman, Durose, Rantala, & Tiedt, 2016).

June 2017 COMPARISON OF RECIDIVISM STUDIES 53

In order to better compare the outcomes of the studies, each study must clearly state their definition of recidivism as well as the methods used to measure that definition (Ruggero, 2015). AOUSC has routinely defined recidi - vism as a return to crime by those who have either served a term of supervised release or probation. The USSC has used the term recidivism to refer to a person’s relapse into criminal behavior, often after the person receives sanctions or undergoes intervention for a previous crime (Hunt & Dumville, 2016). BJS did not provide a definition of recidivism for their study.

Measuring Recidivism AOUSC measures recidivism by the first rearrest for new criminal activity that occurs during and after an offender’s term of super - vision. However, only the first rearrest for a serious criminal offense is counted as a recidi - vistic event in AOUSC’s recidivism statistics. In other words, the focus is on whether the person who is or had been under supervision recidivated, rather than on the number of times a new rearrest occurred for a given per - son who was or had been under supervision.

In addition, because states vary in their practices regarding the extent to which mis - demeanor and petty offenses are reported to their state repositories, AOUSC excludes offenses against public peace, invasion of pri - vacy and prostitution, obstruction of justice, liquor law violations, and traffic offenses. By focusing on major offenses, AOUSC is able to compare recidivism rates across districts and over time, because the statistics are much less influenced by changes in state reporting practices. (AOUSC does tabulate recidivism rates for minor offenses and can report those statistics as well; however, excluding minor offenses does not materially understate its arrest statistics.)

In addition to minor offenses, arrests resulting from violations of the conditions of supervision are also excluded from AOUSC’s recidivism statistics. Arrests for technical vio - lations are not indicative of new criminal behavior, but rather reflect an offender’s failure to comply with certain conditions of his or her supervision, such as testing positive for illegal drugs, failing to complete substance abuse treatment, or traveling outside of the area with - out prior permission. The USSC study, on the other hand, considered all recidivism events (including felonies, misdemeanors, and techni - cal violations of the conditions of supervision) except minor traffic offenses when measuring

recidivism. The offenses were ranked in order of seriousness (Hunt & Dumville, 2016). The BJS study used the first arrest, including arrests for technical violations, as a recidivistic event, but also reported recidivism rates for multiple arrests. The most serious offense charge was used to characterize the arrest offense type (Markman et al., 2016).

Compared to offenders who began a term of supervision a decade ago, the current federal offender is at an increased risk to recidivate, as measured by federal risk assess - ment instruments. In an effort to account for changes in risk, AOUSC has begun to use statistical techniques to adjust for risk in their recidivism statistics. Statistics that adjust for risk provide standardized comparisons over time and among districts, thus mak - ing comparison analyses more meaningful. Moreover, recidivism rates that are adjusted for risk of the population demonstrate that, despite a steady increase in supervisee risk profile, recidivism defined by rearrest, revoca - tion, or a combination of the two measures is decreasing. This result is highly encouraging for stakeholders and policymakers alike, as it suggests that recent advances in federal super - vision practices are producing more favorable outcomes. AOUSC studies of recidivism sta - tistics, unlike those of the other studies, report adjusted rates that control for person-level characteristics, including age, race, sex, risk level, and instant offense type.

Study Cohort and Follow-up Time A major difference between the AOUSC study and the USSC and BJS studies is the size of the population being studied. AOUSC’s study cohort included a total of 454,223 persons serving active supervision terms of proba - tion and supervised release that commenced between October 1, 2004, and September 30, 2014. The USSC report only examined 25,431 offenders who were released from federal prison after serving a sentence of imprisonment or were placed on probation in calendar year 2005. Although larger than the USSC study, the BJS study, which is based on 42,977 offenders placed on federal community supervision during fiscal year 2005, is still relatively small in comparison to AOUSC’s study. All three studies focused on U.S. citizens. An area AOUSC did not explore that the other two studies did explore was offender demographics. As part of its offender demographics, USSC examined race/ethnic - ity, gender, and education level. USSC also

looked at recidivism rates by criminal history score and sentence originally imposed. In addition to offender demographics (race/eth - nicity, sex, and age), BJS examined recidivism rates by number of prior arrests.

Although AOUSC is capable of tracking its earliest cohort of offenders for 10 years or more, statistics published for external con - sumption focus on five-year rearrest rates while under supervision and three-year rear - rest rates after completing supervision. The USSC study uses an 8-year follow-up period and the BJS study uses a five-year follow-up window. Neither the USSC nor the BJS studies distinguish between arrests that occur during supervision and those that occur after super - vision. Not surprisingly, studies with longer follow-up periods tend to report higher rates of recidivism. In this case, one would expect the BJS study to yield the highest recidivism rates, and it does (43.0 percent compared to 42.1 per - cent for USSC and 27.7 percent for AOUSC).

Recidivism Rates All three studies report cumulative rearrest rates over the follow-up periods. For example, if an offender who was sentenced to two years of supervision is rearrested after six months, that arrest will be included in both the one-year and two-year recidivism statis - tics. However, if an offender was sentenced to 12 months of supervision and was arrested after six months, the arrest is only included in the 12-month rearrest statistics but not in the two-year statistic. AOUSC reported that, within the first year of starting supervision, 9.3 percent of federal offenders were rearrested for a serious offense. In comparison, USSC reported a one-year recidivism rate of 16.6 percent and BJS reported a rate of 18.2 percent. All three studies indicate that the majority of reoffending occurs within the first two years of starting supervision (see Table 1). These find - ings suggest that offenders who have recently re-entered the community are the most vul - nerable and the most likely to reoffend.

On average, most federal offenders receive between 36-60 months of community supervi - sion. After three years of supervision, AOUSC reports a recidivism rate of 20.8 percent, which is 12.9 percentage points lower than USSC’s reported rate (33.7 percent) and 14.2 percent - age points lower than BJS’s reported rate (35 percent). The five-year recidivism rate is argu - ably the most significant performance marker in these studies, because it represents the end of the average supervision term. Moreover, in terms of public safety, the expectation is that

54 FEDERAL PROBATION Volume 81 Number 1

community supervision will have a positive effect on reducing criminal behavior. AOUSC reports a five-year recidivism rate of 27.7 per - cent, while USSC reports a rate of 42.1 percent, and BJS reports 43 percent (see Table 1).

Among those who are only aware of the different recidivism rates reported by USSC and BJS, without any further context on varia - tions in defining and measuring recidivism, the differences can arouse confusion and perhaps even doubts about the accuracy of AOUSC’s published recidivism rates. In an effort to eliminate the confusion, outlined in Table 1 below is a summary of the major dis - crepancies among the three studies.

Conclusion Recent recidivism studies by the AOUSC, USSC, and BJS have brought attention to the importance of understanding the scope of reoffending in the federal probation and pre - trial services system. These studies have also brought to light how difficult it is to compare

TABLE 1. Comparison of Key Findings

recidivism rates across agencies. Even when using similar data, discrepancies can exist based on definitional and methodological dif - ferences. No study is without error, and any definition will underestimate the “true” recidi - vism rate, because rates are based on official criminal record data that only show crimes for which people have been arrested or con - victed (Blumstein & Larson, 1971). However, when reviewing various recidivism studies, it is important to keep in mind how recidivism is measured and, more importantly, what is excluded or included in the analysis (e.g., technical violations and traffic offenses).

When examining recidivism it is also important to look at more than just the overall rate. One must also consider the risk associ - ated with the offenders. Not all offenders share the same levels of risk and therefore do not reoffend at the same rate. Only the AOUSC study controls for risk; as a result, it provides a more accurate and nuanced reflection of recidivism among federal offenders.

AOUSC USSC BJS

Study cohort size

Study population

Probation

Term of supervised release (TSR)

U.S. citizen

Recidivism rates (cumulative)

1 year

2 years

3 years

4 years

5 years

8 years

Recidivism Rates by Most Serious Offense (5-yrs)

Drug

Violence

Property

Public Order

Length of follow-up period

Separate during and after supervision rates

Restricted to first arrest only

454,223 25,431 42,977

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

9.3% 16.6% 18.2%

15.6% 27.1% 28.3%

20.8% 33.7% 35.0%

24.5% 38.4% 39.0%

27.7% 42.1% 43.0%

- 49.3% -

29.4% 21.5% 16.1%

24.5% 32.3% 14.5%

25.5% 18.2% 14.8%

- 28.0% 54.5%

5 years 8 years 5 years

Yes No No

Yes No No

References Armstrong, B. (2013). Notes on recidivism.

Alaska Justice Forum, 28(4)-29(1): 4-5 (Winter/Spring).

Blumstein, A., & Larson, R. C. (1971). Problems in modeling and measuring recidivism. Journal of Research in Crime and Delinquen ­ cy, 8(2), 124-132.

Hunt, K. S., & Dumville, R. (2016). Recidivism among federal offenders: A comprehensive overview. United States Sentencing Commis ­ sion. Washington, DC.

King, R., & Elderbroom, B. (2014, Octo - ber). Improving recidivism as a perfor - mance measure. Urban Institute. http:// www.urban.org/sites/default/files/ publication/23031/413247-Improving-Re - cidivism-as-a-Performance-Measure.PDF.

Lowenkamp, C. T., VanNostrand, M., & Holsinger, A. (2013). Investigating the impact of pretrial detention on sentencing outcomes. The Arnold Foundation, http:// www.arnoldfoundation.org/wp-content/ uploads/2014/02/LJAF_Report_state-sen- tencing_FNL.pdf.

Lyman, M., & LoBuglio, S. (2006). “Whys” and “hows” of measuring jail recidivism. Urban Institute. Paper submitted for Jail Reentry Roundtable June 27-28, 2006.

Maltz, M. D. (2001). Recidivism. Orlando, FL: Academic Press, Inc. https://books.google. com/books?id=VxLb2pHetakC&printsec=f rontcover#v=onepage&q&f=false

Markman, J. A., Durose, M. R., Rantala, R. R., & Tiedt, A. D. (2016). Recidivism of offenders placed on federal community supervision in 2005: Patterns from 2005 to 2010. Bureau of Justice Statistics, Washington DC: US Department of Justice, June. NCF 249743.

National Institute of Justice (2014). Recidivism. https://www.nij.gov/topics/corrections/re - cidivism/Pages/welcome.aspx.

Ruggero, T. (2015). Measuring recidivism: Definitions, errors, and data sources. Center for Public Safety Initiatives. Working Paper #2015-03 (February). https://www.rit.edu/ cla/criminaljustice/sites/rit.edu.cla.crimi - naljustice/files/images/2015-03%20-%20 Measuring%20Recidivism%20-%20Defi - nitions,%20Errors,%20and%20Data%20 Sources.pdf.

Urban Institute (2009). Measuring recidivism at the local level: A quick guide. Urban Institute. http://www.urban.org/sites/default/files/ recidivism-measures_final-for-website.pdf.

Minor offenses included No Yes Yes

Technical violations included No Yes Yes

Adjustment for risk Yes No No

Explicit definition of recidivism Yes Yes No