JUNE 2016
a journal of correctional philosophy and practice
The Supervision of Low-risk Federal Offenders: How the Low-risk Policy Has Changed Federal Supervision Practices without Compromising Community Safety By Thomas H. Cohen, David Cook, Christopher T. Lowenkamp
Examining Overrides of Risk Classifications for Offenders on Federal Supervision By Thomas H. Cohen, Bailey Pendergast, Scott W. VanBenschoten
Trademarks, Press Releases, and Policy: Will Rigorous Research Get in the Way? By Kristin Bechtel, Anthony W. Flores, Alexander M. Holsinger, Christopher T. Lowenkamp
A Case Study of the Implementation of Staff Training Aimed at Reducing Rearrest (STARR) By Tammatha A. Clodfelter, Jefferson E. Holcomb, Melissa A. Alexander, Catherine D. Marcum, Tara N. Richards
Street-level Discretion and Organizational Effectiveness in Probation Services By Alese Wooditch, Lauren Duhaime, Kimberly Meyer
Extralegal Factors and the Imposition of Lifetime Supervised Release for Child Pornography Offenders By Niquita M. Vinyard
Trends in the Criminality and Victimization of the Elderly By Peter C. Kratcoski, Maximilian Edelbacher
Juvenile Focus By Alvin W. Cohn ADVISORY COMMITTEE
m e m b e r s
a journal of correctional Dan Richard Beto philosophy and practice National Association of Probation Executives Huntsville, Texas
PUBLISHED BY James Byrne The Administrative Office of the U.S. Courts University of Massachusetts, Lowell Lowell, Massachusetts James C. Duff, Director Honorable James G. Carr Matthew G. Rowland, Chief United States District Court Probation and Pretrial Services Office Toledo, Ohio
Federal Probation ISSN 0014-9128 is dedicated to informing its readers about current Alvin W. Cohn thought, research, and practice in corrections and criminal justice. The journal welcomes the Administration of Justice Services, Inc. contributions of persons who work with or study defendants and offenders and invites authors Rockville, Maryland to submit articles describing experience or significant findings regarding the prevention and control of crime and delinquency. A style sheet is available from the editor. Ronald P. Corbett, Jr. University of Massachusetts, Lowell Federal Probation is published three times yearly—in June, September (on a special topic), Lowell, Massachusetts and December. Permission to quote is granted on the condition that appropriate credit is given the author and Federal Probation. For information about reprinting articles, please contact Thomas Henry the editor. Seton Hall University Subscriptions to Federal Probation are available from the Superintendent of Documents at an South Orange, New Jersey annual rate of $16.50 ($22.40 foreign). Please see the subscription order form on the last page of this issue for more information. Magdeline Jensen CEO, YWCA of Greater Pittsburgh Pittsburgh, Pennsylvania
Tim Murray Pretrial Justice Institute Washington, DC
EDITORIAL STAFF Honorable David D. Noce United States District Court Nancy Beatty Gregoire, Executive Editor St. Louis, Missouri Ellen Wilson Fielding, Editor Daniel B. Ryan Justice Solutions Group Federal Probation Lakeville, Minnesota Administrative Office of the U.S. Courts Washington, DC 20544 Faye Taxman telephone: 202-502-1651 George Mason University fax: 202-502-1677 Fairfax, Virginia email: [email protected] Marie VanNostrand Senior Consultant, Luminosity, Inc. Postmaster: Please send address changes to St. Petersburg, Florida the editor at the address above. June 2016 1
THIS ISSUE IN BRIEF The Supervision of Low-risk Offenders: How the Low-risk Policy Has Changed Federal Supervision Practices without Compromising Community Safety 3 The authors present a preliminary analysis of the federal probation system’s recently adopted evidence-based supervision policy for low- risk offenders, asking 1) Have the number of officer/offender interactions changed after implementation of the low-risk policy, 2) What are the recidivism patterns of low-risk offenders supervised by officers before and after the low-risk policy went into effect, and 3) Has the collection of imposed fines and restitution changed following adoption of the low-risk policy? By Thomas H. Cohen, David Cook, Christopher T. Lowenkamp Examining Overrides of Risk Classifications for Offenders on Federal Supervision 12 The federal probation system’s Post Conviction Risk Assessment instrument (PCRA) has been empirically shown to effectively predict the likelihood that an offender will recidivate during his or her supervision period. Judicial policy allows officers the option to override the PCRA’s risk classification of an offender for certain policy-related reasons (sex offender, persistently violent, mental health issues, or serious youthful offender) or as “discretionary overrides.” In this article the authors first examine the overall prevalence of overrides for offenders under federal supervision, the types of overrides, and the rationales. They then explore whether offenders with overrides are supervised differently compared to those without and whether they recidivate at rates similar to their original or reclassified risk levels. By Thomas H. Cohen, Bailey Pendergast, Scott W. VanBenschoten Trademarks, Press Releases, and Policy: Will Rigorous Research Get in the Way? 22 It is no secret that reports regarding new practices or concepts can be broadly branded, trademarked, marketed, and distributed when the information has not been subjected to replication and peer review. What are the implications when policies and practices are adopted and substantial funding is allocated without an adequate level of empirical support? By Kristin Bechtel, Anthony W. Flores, Alexander M. Holsinger, Christopher T. Lowenkamp A Case Study of the Implementation of Staff Training Aimed at Reducing Rearrest (STARR) 30 Research on evidence-based correctional practices notes the critical importance of program implementation in assessing program effectiveness. In this article, the authors describe the implementation of a training program to improve officer-offender interactions in the federal probation system. They use several sources of information to assess implementation strategies and the success of those efforts within a single federal district. By Tammatha A. Clodfelter, Jefferson E. Holcomb, Melissa A. Alexander, Catherine D. Marcum, Tara N. Richards Street-level Discretion and Organizational Effectiveness in Probation Services 39 The authors provide an overview of street-level discretion among probation officers (and in probation departments more generally) and discuss how discretion supports the attainment of goals at the organizational level. They also examine the limitations of the street-level bureaucrat approach and conclude with recommendations for how future practices and scholarship may overcome these limitations. By Alese Wooditch, Lauren Duhaime, Kimberly Meyer Extralegal Factors and the Imposition of Lifetime Supervised Release for Child Pornography Offenders 45 For child pornography offenders, the supervised release term is particularly important because of the statutory override found in 18 U.S.C. 3583(k) enhancing the length of the term from a maximum of five years to a minimum of five years to life and providing for the revocation of supervised release resulting in the incarceration of the defendant for the remainder of the period. The author examines the effects of legal and extralegal factors on supervised release outcomes, specifically lifetime supervised release for child pornography offenders. By Niquita M. Vinyard 2 FEDERAL PROBATION Volume 80 Number 1
Trends in the Criminality and Victimization of the Elderly 58 Among the relatively reliable variables predictive of amount and types of crime is age: The proportion of violent crimes will decline as the ages of the offenders increases. On the other hand, the proportion of the elderly who are victimized by crime, particularly theft, financial fraud, and physical abuse, exceeds that of other age groups. The authors focus on changes in the amount and types of crime committed by the elderly and also on the various methods used to victimize the elderly. By Peter C. Kratcoski, Maximilian Edelbacher
D E P A R T M E N T S
Juvenile Focus 64
Contributors to This Issue 71
The articles and reviews that appear in Federal Probation express the points of view of the persons who wrote them and not necessarily the points of view of the agencies and organizations with which these persons are affiliated. Moreover, Federal Probation’s publication of the articles and reviews is not to be taken as an endorsement of the material by the editors, the Administrative Office of the U.S. Courts, or the Federal Probation and Pretrial Services System. June 2016 3 The Supervision of Low-Risk Federal Offenders: How the Low-risk Policy Has Changed Federal Supervision Practices without Compromising Community Safety
Thomas H. Cohen David Cook Christopher T. Lowenkamp Probation and Pretrial Services Office Administrative Office of the U.S. Courts
SINCE THE EARLY 2000s, the federal low-risk offenders being influenced by their the number of officer/offender interactions probation and pretrial services system higher-risk counterparts. In addition, placing changed after implementation of the low-risk has adopted an approach that emphasizes low-risk offenders into intensive monitor- policy? (2) What are the recidivism patterns using evidence-based practices to reduce ing regimes could potentially disrupt their of low-risk offenders supervised by officers the risk and recurrence of recidivism prosocial networks, including their ability to before and after the low-risk policy went into (Alexander & VanBenschoten, 2008; Cohen maintain long-term employment or remain in effect? and (3) Has the collection of court- & VanBenschoten, 2014; Hughes, 2008). As stable relationships with non-criminal peers imposed fines and restitution changed since part of that approach, the federal probation (Lowenkamp & Latessa, 2004). the low-risk policy was adopted? As we will system adopted the risk, needs, and responsiv- The federal probation system has devel- show, we find evidence that low-risk offenders ity (RNR) model of correctional supervision oped and implemented a risk assessment are being supervised less intensively by federal (Alexander & VanBenschoten, 2008; Andrews, instrument (the Post Conviction Risk probation officers and that this change in Bonta, & Hoge, 1990; Andrews & Bonta, Assessment or PCRA) that identifies those offender management has not compromised 2010). One of the key tenets of the RNR model offenders at lowest risk of recidivism; the community safety nor impeded the collection is that officers should focus on high-risk system has also promulgated policies to guide of court-imposed fines. The collection of res- offenders, while spending minimal time and officers on the supervision of low-risk offend- titution obligations, however, declined during resources on offenders at low risk to reoffend. ers (Guide to Judiciary Policy, 2014; Johnson, the period in which the low-risk policy was The risk principle is a core component Lowenkamp, VanBenschoten, & Robinson, implemented. Future studies can assess the of the RNR model. Specifically, research has 2011; Lowenkamp, Johnson, VanBenschoten, influence of the low-risk policy over longer shown that focusing time, attention, and Robinson, & Holsinger, 2013). Policy guid- periods and examine whether the negative resources on low-risk offenders has negligible ance on the supervision of low-risk offenders effect on restitution collections is offset by the impacts on recidivism. In fact, intensive super- was put into place on or about June 2012 benefits of this policy. vision of these offenders can produce negative when the Criminal Law Committee of the We note that this work represents one of consequences: Low-risk offenders supervised Judicial Conference endorsed this policy and the first efforts to investigate the potential at higher levels are more likely to reoff- recommended its ultimate adoption by the impacts of the low-risk supervision model end compared to low-risk offenders who are Judicial Conference of the U.S. in September on a system-wide basis. We are unaware of placed under supervision programs involving 2012. Now that the low-risk policy has been any efforts by other organizations analyzing minimal levels of contacts, treatment, moni- in effect for a few years, we seek to under- the effects of instituting this core component toring, etc. (Andrews, Bonta & Hoge, 1990; stand whether the policy of minimizing the of the RNR model for an entire correctional Lowenkamp & Latessa, 2004; Lowenkamp, resources expended on these offenders has agency. The few empirical assessments of low- Holsinger, & Latessa, 2006; Lowenkamp, succeeded without compromising community risk supervision practices tended to involve Flores, Holsinger, Makarios, & Latessa, 2010). safety or impeding the collection of the court- smaller field experiments or pilot studies. The reason is that intense supervision typically imposed financial obligations of fines and Hence, this research addresses this gap in the results in the intermixing of low- and high- restitution. community corrections literature. risk offenders. Placing low- and high-risk This research is a preliminary analysis of offenders in the same program can poten- the implications of the low-risk policy that tially result in negative social learning, with addresses the following questions: (1) Have 4 FEDERAL PROBATION Volume 80 Number 1
The Low-risk Supervision Policy and require written justification by the officer from the lower to higher risk categories during and approval by the supervisor. the period examined. The low-risk policy became an integral part For those low-risk offenders not reclas- For this section of the article, we used the of post-conviction supervision in June 2012 sified to higher supervision levels, judicial PCRA rather than the RPI to examine contact when it was endorsed by the Criminal Law policy provides additional details on appropri- patterns by risk level over the period of policy Committee of the Judicial Conference, which ate reporting requirements and monitoring. implementation. The PCRA served as the recommended that it be adopted as offi- Of particular importance are the judicial pol- basis for risk differentiation because we are cial policy by the Judicial Conference of the icy’s instructions that after completion of the analyzing officer behavior towards offend- United States. Details about this policy are initial case plan, subsequent contact should ers rather than outcomes. Using the PCRA provided in the Guide to Judiciary Policy (judi- be minimized unless circumstances warrant allowed us to standardize the measure of risk cial policy). This policy states that offenders further intensive supervision. In addition, the and driver of officer behavior over the period classified as low risk by either the PCRA or policy recommends that officers forgo sub- in which the low-risk policy was integrated the Risk Prediction Index (RPI) actuarial tool sequent case plans and reassessments unless into the federal supervision system. Moreover, are eligible for supervision under the low- the officer suspects or has been informed of since the low-risk policy was being promul- risk policy. The PCRA is a fourth generation a negative change in the offender’s conduct gated during the same time that the PCRA was risk assessment instrument currently used by or conditions. Instead, officers are to rely being deployed, officers tended to associate federal probation officers to classify offend- on notification from law enforcement data- the low-risk policy more with the PCRA than ers into one of the four following recidivism bases and other sources to learn if a low-risk with the RPI. We believe this is because the risk categories: low, low/moderate, moder- offender has returned to crime. Judicial policy PCRA training included heavy reinforcement ate, and high. Prior to its implementation, also informs officers to consider petitioning of the risk principle to officers. federal probation officers relied on the RPI, the court to remove or suspend any unnec- We extracted officer/offender contact a second-generation risk assessment tool that essary special conditions imposed on these information from the Probation and Pretrial classified offenders into high, moderate, or offenders. By stating that limited resources Services Automated Case Tracking System low recidivism risk categories.1 The low-risk should be expended on lower risk offenders, (PACTS), the case management system used policy references the earlier RPI as well as the the low-risk policy allows officers to conserve by federal officers. In this analysis, the aver- PCRA because the PCRA was deployed into their time so that they can focus on offenders age and median number of monthly total, the federal system gradually, thus overlap- at the higher end of the risk continuum. In personal, and collateral officer/offender con- ping with the earlier risk instrument that it fact, the low-risk policy provides the frame- tacts was calculated during an offender’s first was replacing. work in which officers can concentrate most six months of supervision.3 Personal con- The low-risk policy states that offenders of their time, resources, and services on the tacts are direct interactions between officers classified as low risk by either the PCRA or highest risk offenders. and offenders and include interactions taking RPI are predicted to reoffend at relatively low place in the probation office, the offender’s rates. Hence, judicial policy instructs officers Low-risk Policy and Officer/ home, the offender’s place of employment, to limit their supervision activities for low- Offender Contacts or elsewhere in the community. Personal risk offenders to “monitoring compliance We analyzed the relationship between officer/ contacts can also include electronic commu- with the conditions of release, if applicable, offender contacts and the low-risk policy by nications between the officer and offender and responding appropriately to any changes calculating the median and average num- such as telephone contacts, voice mail, or text in circumstances.” Although judicial policy ber of monthly contacts for offenders with messaging.4 Collateral contacts are officer recommends applying minimal levels of PCRA assessments received into supervi- interactions with third parties familiar with supervision to low-risk offenders, there are sion both before and after implementation the offender such as treatment providers, law important exceptions to this general rule. In of the low-risk policy. The pre-policy period enforcement officers, employers, and family particular, judicial policy provides officers covers offenders received into supervision members. These contacts can also be made with discretion to place low-risk offenders between June 28, 2009, and June 26, 2012,2 electronically (through telephone, voice mail, into a higher supervision level when the offi- while the post-policy periods covers offend- and text messaging). cer determines through his or her professional ers received into supervision between June judgment that the offender’s proclivity to 27, 2012, and August 12, 2015. In addition 3 reoffend is underestimated. This reclassifica- In this analysis, supervision encompasses both to examining officer/offender contacts for offenders placed on terms of supervised release tion process is known as the professional or low-risk offenders, we calculated changes in (TSR) and those on straight probation. TSR refers supervision override and applies where the monthly contacts for the other PCRA risk to offenders serving a term of supervision after officer determines that the offender has met being released from federal prison, while proba- categories (i.e., low/moderate, moderate, and one of the following policy-related criteria: tion refers to a court-imposed sentence involving high risk). By analyzing trends in monthly being classified as a sex offender, manifest- community monitoring without an incarceration contact data for all risk levels, we explore sentence. See 18 USC § 3583 & § 3563. ing persistently violent behavior, evidencing whether there was a redistribution of contacts 4 This definition of personal contacts differs from severe mental health issues, or being consid- that used in internal Probation and Pretrial Services ered a serious youthful offender. Changes in 2 Although the low-risk policy was not officially reports, which do not count electronic communica- supervision level occurring for non-policy implemented until September 2012, we used the tions between officers and offenders as personal reasons are labeled discretionary overrides June 2012 date, because that is when this policy contacts. An examination of this more restricted was adopted by the Criminal Law Committee of the version of personal contacts revealed patterns simi- 1 For more information about the PCRA and RPI, Judicial Conference. lar to those reported in this paper. see AOUSC, 2011. June 2016 SUPERVISION OF LOW-RISK OFFENDERS 5
We focused on the first six months of of supervision. This study population was into place.6 For example, the median number supervision because that allowed us to exam- further divided into a pre and post low-risk of total monthly officer/offender contacts for ine three years of post-policy contact patterns policy group. The pre low-risk policy group low-risk offenders decreased by 23 percent covering fiscal years 2013 through 2015. included offenders who started their supervi- from over 2 contacts per month prior to the Moreover, officer/offender contacts tend to sion terms prior to the low-risk policy (i.e., low-risk policy to slightly fewer than 2 con- be more intense during the first six months before June 2012), while the post policy cohort tacts per month after implementation of the of supervision and are often driven by an included offenders placed on federal supervi- policy. The median number of total monthly offender’s supervision conditions, such as sion after the low-risk policy went into effect officer/offender contacts also declined by 15 the requirement to undergo mandatory drug (i.e., after June 2012) (see Table 1). Because percent for low/moderate-risk offenders. testing.5 All offenders supervised for less than the PCRA was deployed gradually starting in Apparently, declines in total officer/ 6 months were excluded from this analysis. fiscal year 2009, there is not an even number offender contacts for lower risk offenders were Whether trends reported in this paper hold of offenders in each group. Forty-three per- not commensurate with increases in contacts true or become more pronounced beyond the cent of offenders in the study population were for higher risk offenders. High-risk offenders first six months of supervision considered in received into supervision before promulgation saw no changes in their total median monthly this study is left for future inquiry. of the low-risk policy, while 57 percent had contacts between the pre and post policy peri- In total, the study population included their supervision terms commence after the ods, while moderate-risk offenders witnessed 229,919 offenders with PCRA assessments low-risk policy was put into place. a 9 percent reduction in their total median whose monthly contacts with officers could Table 2 shows the average and median monthly contacts. We note that high-risk be calculated during their first six months number of monthly officer/offender contacts offenders constitute an increasing proportion for low, low/moderate, moderate, and high of the federal supervision population, and 5 The mandatory conditions of 18 U.S.C. § 3563(a) (5) and (e), 3583(d), and 4209(a) are outlined in the risk offenders during their first six months that officers’ caseloads have risen over the Guide to Judiciary Policy. Offenders are required to under supervision both before and after last few years (Baber, 2015). Consequently, refrain from any unlawful use of a controlled sub- implementation of the low-risk policy. As it is possible that absent the low-risk policy, stance and submit to one drug test within 15 days expected, the median number of total officer/ resources dedicated to higher risk offenders of release when on probation or supervised release offender contacts has declined the most for could potentially have declined rather than and at least two periodic drug tests thereafter, unless this condition was suspended by the court low-risk offenders since this policy was put remain unchanged. after a determination that the offender presents a Comparing changes in median personal low risk of future substance abuse. contacts pre and post policy shows the median number of monthly personal contacts declin- TABLE 1. ing by 17 percent for low/moderate risk Study Population of Federally Supervised Offenders for Low-Risk Contacts Analysis offenders. The low and moderate risk offend- Risk Policy & PCRA Risk Levels Number Percent ers witnessed similar decreases in median monthly personal contacts (13 percent and 10 All offenders 229,919 100% percent, respectively). Conversely, high-risk Pre-low risk/* 98,044 43% offenders saw no changes in their median Low 37,633 16% monthly personal contacts during the study Low/Moderate 39,036 17% time frame. Moderate 16,583 7% In terms of collateral contacts, low-risk High 4,792 2% offenders saw their median monthly collat- eral contacts decline by 40 percent, from .5 Post-low risk/* 131,875 57% contacts per month before the low-risk policy Low 48,836 21% to .3 contacts per month after the low-risk Low/Moderate 49,615 22% policy came into effect. The median monthly Moderate 25,531 11% collateral contacts for the other PCRA risk High 7,893 3% categories remained unchanged. Note: Includes offenders with actual PCRA assessments received onto federal supervision between fiscal years While an examination of contacts for 2009–2015. offenders received into supervision between */Refers to whether offenders were received onto federal supervision before or after enactment of the low-risk policy. the pre and post low-risk policy supports that officers are contacting low-risk offend- ers less frequently, the analysis presented in Table 2 can mask important trends. For example, the practice of supervising lower risk offenders less intensively might have
6 The median is the number separating the higher half of the data from the lower half. In this report, median contacts can be more useful than average contacts because averages can be disproportion- ately influenced by the small number of offenders with exceptionally high contact rates. 6 FEDERAL PROBATION Volume 80 Number 1
TABLE 2. Mean and Median Number of Monthly Officer/Offender Contacts Prior to and After Implementation of the Low-Risk Policy
Median Contacts Per Month Mean Contacts Per Month Contact Types & PCRA Risk Levels Pre-Low Risk Post-Low Risk Percent Change Pre-Low Risk Post-Low Risk Percent Change Total contacts Low 2.2 1.7 -23% 2.6 2.3 -14% Low/moderate 2.7 2.3 -15% 3.2 2.8 -10% Moderate 3.3 3.0 -9% 3.9 3.7 -5% High 4.0 4.0 0% 4.8 4.8 0% Person Low 1.5 1.3 -13% 1.9 1.6 -13% Low/moderate 1.8 1.5 -17% 2.1 1.9 -10% Moderate 2.0 1.8 -10% 2.4 2.2 -6% High 2.3 2.3 0% 2.6 2.6 0% Collateral Low 0.5 0.3 -40% 0.8 0.7 -15% Low/moderate 0.7 0.7 0% 1.1 1.0 -10% Moderate 1.0 1.0 0% 1.6 1.5 -4% High 1.5 1.5 0% 2.1 2.2 3% Note: Includes offenders with actual PCRA assessments received onto federal supervision between fiscal years 2009-2015. The pre and post low risk terms refers to whether offenders were received onto federal supervision before or after enactment of the low-risk policy. Contacts based on initial six months under supervision. gradually permeated the federal probation those shown for total contacts. For exam- comparison, the median number of monthly system, meaning that the patterns of spending ple, the median monthly personal contacts collateral contacts increased by 31 percent for less time with lower risk offenders may not be declined by 29 percent for low-risk offenders high-risk offenders from about 1 contact per apparent without examining yearly monthly from nearly 2 contacts per month for fiscal month in 2010 to nearly 2 contacts per month contact trends. Table 3 examines the median year 2010 to about 1 contact per month for in 2015. monthly officer/offender contact rates on an fiscal year 2015. Median monthly contacts Figures 1 and 2 show the distribution of annual basis for fiscal years 2010 through also decreased 17 percent for low/moderate monthly total contacts on an annual basis for 2015. The contacts were examined separately risk offenders. In comparison, moderate-risk high- and low-risk offenders. These figures by PCRA risk levels and contact types (e.g., offenders saw their median monthly personal show the percentage of high- and low-risk total, personal, and collateral). contacts decline by 10 percent and high-risk offenders received into supervision from fiscal offenders saw no changes in their median years 2010 through 2015 who were contacted Total monthly contacts monthly personal contacts in the period span- less than once per month, 1-1.9 times per The median monthly total contacts decreased ning 2010 to 2015. About half of the high-risk month, 2-2.9 times per month, 3-3.9 times the most for lower risk offenders, while offenders were contacted 2 or more times per per month, and 4 or more times per month. offenders on the higher end of the PCRA risk month during the study period. An analysis of changes in the distribution continuum witnessed either smaller declines of contacts provides another way of show- or slight increases in their median contacts. Collateral monthly contacts ing whether officers are contacting low-risk For example, the median monthly total con- Unlike personal contacts, collateral contacts offenders less frequently over time compared tacts declined by 26 percent for low-risk and manifested patterns more in alignment with to their high-risk counterparts. 19 percent for low/moderate-risk offenders the risk principle. Specifically, collateral con- Over the past six fiscal years, increas- from 2010 to 2015. In comparison, moderate- tacts declined the most for low-risk offenders, ingly higher percentages of low-risk offenders risk offenders saw their median monthly total while they increased substantially for offend- were contacted less than once per month. contacts decline by only 6 percent, while high- ers classified in the highest risk category. For For instance, 7 percent of low-risk offenders risk offenders witnessed their total contacts instance, the median number of monthly placed on supervision in fiscal year 2010 were increase by 11 percent, from 3.8 contacts per collateral contacts declined by 40 percent contacted less than once per month, while 18 month to 4.2 contacts per month during the for low-risk offenders from 0.5 contacts per percent placed on supervision during fiscal 2010 to 2015 period. month in 2010 to 0.3 contacts per month year 2015 were contacted less than once per in 2015. In addition, median monthly col- month. Among high-risk offenders, a slightly Personal monthly contacts lateral contacts decreased by 13 percent for higher percentage were contacted 4 times or An examination of trends in personal con- low/moderate-risk offenders and exhibited more per month in 2015 (53 percent) com- tacts reveals somewhat similar patterns to no changes for moderate-risk offenders. In pared to 2010 (48 percent). June 2016 SUPERVISION OF LOW-RISK OFFENDERS 7
TABLE 3. Median Number of Monthly Officer/Offender Contacts by PCRA Risk Levels, Fiscal Years 2010–2015 Median Contacts Per Month Median Person Contacts Per Month Median Collateral Contacts Per Month Low/ Low/ Low/ Fiscal Year Low Moderate Moderate High Low Moderate Moderate High Low Moderate Moderate High FY-2010 2.3 2.7 3.2 3.8 1.7 1.8 2.0 2.3 0.5 0.8 1.0 1.3 FY-2011 2.2 2.7 3.3 4.3 1.5 1.8 2.2 2.4 0.5 0.7 1.2 1.7 FY-2012 1.8 2.5 3.2 4.0 1.3 1.7 2.0 2.2 0.5 0.7 1.0 1.7 FY-2013 1.8 2.3 3.0 4.0 1.3 1.5 1.8 2.2 0.3 0.7 1.0 1.5 FY-2014 1.7 2.2 3.0 4.0 1.3 1.5 1.8 2.3 0.3 0.7 1.0 1.5 FY-2015 1.7 2.2 3.0 4.2 1.2 1.5 1.8 2.3 0.3 0.7 1.0 1.7 Percent change -26% -19% -6% 11% -29% -17% -10% 0% -40% -13% 0% 31% Note: Includes offenders with actual PCRA assessments received onto federal supervision between fiscal years 2010-2015. Offenders received into supervision during fiscal year 2009 excluded as there were relatively few PCRA assessments during that fiscal year. Contacts based on initial six months under supervision. Bold denotes the year that the low-risk policy was implemented.
Low-risk Supervision Policy and FIGURE 1. Offender Recidivism Distribution of Monthly Officer/Offender Total Contacts for Low-Risk Next we examined recidivism of low-risk PCRA Offenders, Fiscal Years 2010–2015 offenders before and after implementation of the low-risk policy. Three groups of offend- 010 7% 0 32% 27% 15% 19% ers were analyzed. The first were offenders who started and ended their supervision 011 9% 34% 25% 14% 18% terms before the beginning of the low-risk 01 13% 38% 23% 12% 14% policy (i.e., before June 2012). The second group comprised offenders whose supervi- 01 16% 39% 22% 10% 13% sion terms started after the low-risk policy was instituted (i.e., after June 2012). The third 01 17% 40% 20% 10% 13% group includes offenders who started their 01 18% 39% 20% 10% 13% supervision terms before the low-risk policy but ended their terms after the policy was 0 10 0 0 0 0 0 0 80 0 100 instituted. We labeled this third group “split er e o o e er cases.” Last, we used the RPI rather than the e 1 o 1 0 1 o 0 o 0 o or more o PCRA for the recidivism analysis because our primary focus in these sections is on outcomes rather than officer behavior. Offender recidi- vism outcomes and payment patterns are FIGURE 2. unlikely to be influenced by the risk instru- Distribution of Monthly Officer/Offender Total Contacts for High-Risk ment that officers use. Offenders, Fiscal Years 2010–2015 Table 5 shows the recidivism rates for offenders placed under supervision by risk 010 10% 21% 18% 48% level before and after the low-risk policy was placed into effect, as well as for split cases. 011 10% 16% 17% 55% Recidivism rates were calculated within a 01 12-month period after the supervision start 10% 19% 19% 51% date and include arrests for any felony or mis- 01 11% 19% 18% 51% demeanor offense. In addition, the RPI risk classifications were used because most pre-pol- 01 10% 18% 18% 52% icy offenders did not have PCRA assessments. 01 This analysis shows low-risk offenders 11% 17% 17% 53% recidivating at nearly identical rates regardless 10 0 0 0 0 0 0 80 0 100 of whether they were supervised before or er e o o e er after implementation of the low-risk policy or e 1 o 1 0 1 o 0 o 0 o or more o whether their supervision terms spanned the No e er e e o o or o e er le 1 o er mo e e r e rom 1 8 FEDERAL PROBATION Volume 80 Number 1
TABLE 4. years of supervision, and 3) offenders with 2-3 Descriptive Statistics for the Low-Risk Recidivism Anaylsis years of supervision. Table 6 shows the average amount assessed Supervision Period Number Percent by the courts in fines and restitution by Pre Low-Risk Policy 208,595 53% offender risk level and the year they started Post Low-Risk Policy 64,102 16% supervision. Fines are monetary payments Split Low-Risk Policy 121,134 31% incurred as part of the sentence and are based RPI Category on an offender’s ability to pay, while restitution Low Risk 150,685 38% refer to monetary payments that seek to com- pensate victims for their losses. Restitution Moderate Risk 176,524 45% obligations can be assessed against individu- High Risk 66,622 17% als or, in cases of joint and several liability, 12 Month Rearrest 393,831 13% against multiple parties for the same monetary amount. Joint and several liability makes each of multiple defendants liable for the entirety pre- and post-policy periods. For instance, the from the Clerk’s Office Civil and Criminal of a victim’s loss irrespective of each defen- recidivism rates for low-risk offenders were Accounting Module (CCAM). The CCAM dant’s degree of fault (WilsonElser, 2013). The essentially unchanged between the pre-policy tracks the criminal monetary penalties owed median amounts owed in single restitution group (6 percent arrest rate) and post-policy and payments made by offenders, as well as and joint and several restitution more than group (4 percent arrest rate). Offenders in the funds disbursed and monies owed to victims doubled for low-risk offenders between fiscal 8 split category had recidivism rates of 4 percent. of crime. years 2009 and 2015.9 The median restitu- Including splits in the pre-policy group pro- We examined the repayment patterns for tion and joint and several restitution amount duced relatively similar recidivism patterns for two types of low-risk offenders. The first were imposed also increased for moderate- and the pre (5 percent) and post (4 percent) policy offenders who finished supervision before the high-risk offenders. groups. When low-risk splits were merged low-risk policy started (i.e., before June 2012), Given the large increases in court-imposed into the post-policy cohort, the recidivism and the second were offenders who entered amounts over the last seven years, we exam- rates were 6 percent for the pre-policy group supervision after the policy was implemented ined assessments by offense type to see if 7 to 4 percent for the post-policy group. (i.e., after June 2012). Since the pre-policy a particular offense type was driving these group was under supervision for longer peri- Low-risk Supervision Policy and results. For low-risk offenders, fraud cases ods of time and hence, had more time to comprised over 60 percent of the joint and the Payment of Court-imposed make repayments than the post-policy group, Financial Obligations several restitution amounts and 40 percent we standardized the repayment follow-up of the single restitution obligations. The joint Last, we examined the relationship between periods for both study groups. Specifically, we and several restitution and single restitution the low-risk policy and the payment of divided the pre and post policy groups into amounts imposed by the courts for fraud court-ordered fines and restitution. As with three subgroups based on their time under cases nearly doubled from 2009 through 2015, the recidivism analysis, we used the RPI federal supervision: 1) offenders with less than approximating the overall increases for these rather than the PCRA because the focus was one year of supervision, 2) offenders with 1-2 obligation types (data not shown in report). on outcomes rather than officer behavior. In future studies, we will explore continuing We combined data from PACTS with data 8 Please note that different datasets were used to to investigate possible drivers of the increase. 7 It should also be noted that the recidivism examine officer/offender contacts, recidivism pat- rates declined slightly for moderate- and high-risk terns, and information on the collection of fees/ 9 When examining amounts owed, we used median offenders. fines. amounts because large values skew the means.
TABLE 5. Recidivism Rates for Offenders Prior to and After Implementation of the Low-Risk Supervision Policy, by RPI Groups
Low Moderate High All Supervision Policy Percent Percent Percent Percent Group Number Arrested Number Arrested Number Arrested Number Arrested All 150,685 5% 176,524 14% 66,622 27% 393,831 13% Pre Policy 82,046 6% 92,940 17% 33,609 30% 208,595 15% Post Policy 23,580 4% 27,907 13% 12,615 26% 64,102 12% Split 45,059 4% 55,677 11% 20,398 21% 121,134 10% Pre (Including Splits) 127,105 5% 148,617 15% 54,007 27% 329,729 13% Post 23,580 4% 27,907 13% 12,615 26% 64,102 12% Pre 82,046 6% 92,940 17% 33,609 30% 208,595 15% Post (Including Splits) 68,639 4% 83,854 12% 33,013 23% 185,236 11% Note: Risk classifications based on the RPI. Arrest rates calculated within a 12-month period. June 2016 SUPERVISION OF LOW-RISK OFFENDERS 9
Table 7 shows the payment rates and aver- court-imposed restitution penalties because the recidivism rates had changed after enact- age amounts paid for offenders supervised fines are assessed based on an offender’s abil- ment of this policy, and analyzed whether the before and after implementation of the low- ity to pay. Restitution penalties, on the other collection of court-imposed financial penalties risk policy. Overall, low-risk offenders who hand, are based on the actual financial dam- differed after the low-risk policy took effect. were supervised before institution of the ages caused by offenders. In general, findings are supportive of the low-risk policy paid a greater percentage of low-risk policy. This research shows that all their financial obligations than those under Conclusion and Implications low- and low/moderate risk offenders in the supervision after the implementation of this The low-risk supervision policy institution- post policy group have fewer officer/offender policy. This is especially pronounced with alized that officers should expend minimal contacts compared to their pre-policy coun- regards to restitution, both individual and amounts of time and resources on low-risk terparts. This finding suggests that the joint and several obligations. On average, pre- offenders, while placing most of their efforts low-risk policy is influencing officer behavior policy low-risk offenders supervised for two on offenders classified into the higher risk by encouraging federal officers to engage in years or less paid about half their individual categories. Low-risk offenders should be fewer interactions with offenders on the lower restitution ordered, while post-policy low- provided with minimal supervision services, end of the risk continuum. Importantly, the risk offenders on supervision for similar time because research has shown that correctional policy of supervising low-risk offenders less periods paid between 28 percent-37 percent of interventions aimed at reducing recidivism intensively has not compromised community their restitution obligations. The differences in for these offenders tend to be ineffective safety. Post-policy low-risk offenders were no payment rates among low-risk offenders were and can actually produce higher recidivism more likely to recidivate compared to their less marked between the pre and post policy rates for this risk group (Andrews, Bonta & pre-policy counterparts. This finding indi- groups for court-imposed fines. On average, Hoge, 1990; Lowenkamp & Latessa, 2004; cates that federal officers can spend less time the fine payment rates were about 90 per- Lowenkamp, Holsinger, & Latessa, 2006). and resources on low-risk offenders without cent for the pre-policy group and 84 percent The current research examined the relation- an accompanying rise in their recidivism rates. for the post-policy group. Court-imposed ship between the low-risk policy and officer/ For the most part, federal probation fines were probably paid at higher rates than offender contact patterns, explored whether officers continued to successfully monitor
TABLE 6. Court-Assigned Fines, Restitution, and Joint and Several Restitution by Supervision Year
Fine Restitution Joint/Several Restitution Fiscal year Number Mean Median Number Mean Median Number Mean Median Low Risk 2009 2,905 $20,680 $1,500 3,957 $418,932 $35,435 3,229 $649,643 $40,786 2010 3,065 $10,720 $1,500 4,113 $1,356,499 $41,627 3,793 $766,517 $52,397 2011 3,132 $24,899 $1,500 4,233 $600,112 $44,855 4,246 $1,407,249 $63,394 2012 2,950 $50,046 $2,000 4,335 $548,136 $55,800 4,294 $780,935 $68,545 2013 2,999 $16,137 $2,000 4,386 $649,946 $61,114 4,889 $828,933 $64,404 2014 3,062 $30,830 $2,000 4,461 $554,189 $76,894 4,210 $2,914,723 $78,457 2015 1,993 $17,204 $2,000 2,929 $982,220 $85,090 3,246 $1,040,297 $90,950 Moderate Risk 2009 2,105 $4,540 $1,000 1,962 $56,956 $8,338 2,118 $77,903 $6,750 2010 2,078 $2,582 $1,000 1,960 $67,390 $8,243 2,088 $118,963 $11,861 2011 2,277 $2,976 $1,000 1,970 $152,472 $8,501 2,540 $99,636 $10,809 2012 2,035 $2,943 $1,000 1,821 $94,423 $10,000 1,937 $165,627 $10,491 2013 1,952 $6,191 $1,000 1,846 $95,683 $9,369 2,014 $164,676 $11,925 2014 1,831 $2,810 $1,000 1,781 $131,154 $9,800 2,026 $225,532 $14,143 2015 1,305 $2,906 $1,000 1,277 $167,494 $10,978 1,556 $254,819 $15,253 High Risk 2009 1,217 $2,337 $1,000 1,627 $31,875 $4,930 1,652 $30,772 $6,300 2010 1,215 $5,651 $1,000 1,671 $30,805 $4,600 1,656 $28,958 $5,027 2011 1,316 $2,129 $1,000 1,736 $29,727 $4,976 1,718 $45,776 $6,198 2012 1,235 $3,082 $1,000 1,746 $50,457 $5,312 1,660 $44,930 $7,000 2013 1,215 $2,845 $1,000 1,731 $39,692 $5,001 1,564 $50,342 $6,001 2014 1,255 $2,045 $1,000 1,631 $37,152 $4,763 1,784 $33,049 $5,060 2015 878 $2,056 $1,000 1,262 $40,939 $4,834 1,262 $59,908 $5,818 10 FEDERAL PROBATION Volume 80 Number 1
TABLE 7. Repayment Ratio and Amount While Under Supervision Pre and Post Policy
Pre-Low Risk Policy Post-Low Risk Policy Payment Rate Amount Paid Payment Rate Amount Paid Payment types Number Mean Median Mean Median Number Mean Median Mean Median Low Risk Fines Under One Year of Supervision 2,305 93% 100% $4,114 $500 2,193 91% 100% $5,388 $500 One to Two Years of Supervision 1,148 91% 100% $9,356 $1,318 2,457 83% 100% $29,870 $1,858 Two to Three Years of Supervision 448 92% 100% $11,250 $2,000 1,131 79% 100% $7,694 $2,200 Joint/Several Restitution Under One Year of Supervision 312 19% 3% $19,625 $2,102 569 15% 1% $20,451 $1,400 One to Two Years of Supervision 440 24% 7% $34,441 $3,368 1,948 14% 2% $46,231 $2,913 Two to Three Years of Supervision 351 19% 4% $35,155 $3,820 1,553 13% 2% $27,418 $4,680 Restitution Under One Year of Supervision 1,131 49% 38% $35,682 $1,276 1,667 37% 7% $18,493 $1,260 One to Two Years of Supervision 1,341 49% 28% $26,601 $2,316 4,429 28% 5% $17,438 $2,366 Two to Three Years of Supervision 784 40% 15% $32,426 $3,622 2,944 23% 4% $18,270 $3,428 Moderate Risk Fines Under One Year of Supervision 1,505 83% 100% $838 $375 1,143 74% 100% $681 $300 One to Two Years of Supervision 674 80% 100% $1,468 $532 1,345 63% 76% $1,272 $540 Two to Three Years of Supervision 310 85% 100% $2,172 $1,000 826 66% 78% $1,469 $950 Joint/Several Restitution Under One Year of Supervision 326 24% 9% $5,830 $949 326 15% 3% $3,591 $511 One to Two Years of Supervision 378 24% 7% $6,132 $873 988 18% 4% $8,285 $888 Two to Three Years of Supervision 229 27% 8% $4,408 $1,375 660 19% 6% $4,633 $1,578 Restitution Under One Year of Supervision 682 37% 13% $2,223 $589 610 30% 7% $1,467 $400 One to Two Years of Supervision 591 42% 20% $2,670 $830 1,532 28% 7% $4,966 $751 Two to Three Years of Supervision 369 43% 23% $3,546 $1,350 1,039 31% 9% $3,348 $1,240 High Risk Fines Under One Year of Supervision 638 61% 70% $809 $328 1,143 74% 100% $681 $300 One to Two Years of Supervision 466 67% 96% $1,071 $500 1,345 63% 76% $1,272 $540 Two to Three Years of Supervision 204 74% 100% $1,061 $793 826 66% 78% $1,469 $950 Joint/Several Restitution Under One Year of Supervision 395 19% 5% $2,140 $403 326 15% 3% $3,591 $511 One to Two Years of Supervision 266 23% 10% $2,408 $692 988 18% 4% $8,285 $888 Two to Three Years of Supervision 129 27% 8% $2,379 $918 660 19% 6% $4,633 $1,578 Restitution Under One Year of Supervision 824 34% 13% $1,399 $404 610 30% 7% $1,467 $400 One to Two Years of Supervision 595 37% 15% $1,514 $580 1,532 28% 7% $4,966 $751 Two to Three Years of Supervision 255 46% 25% $1,929 $879 1,039 31% 9% $3,348 $1,240 June 2016 SUPERVISION OF LOW-RISK OFFENDERS 11 the collection of court-owed fines despite There are limitations in this study that Baber, L. (2015). Inroads to reducing federal the fact that less time and resources were could be addressed by additional research. In recidivism. Federal Probation, 79(3) 3-8. being expended on the low-risk population. sum, this is a descriptive analysis that exam- Cohen, T., & VanBenschoten, S. (2014). Does Restitution payments, however, did decrease ines how the low-risk policy is related to the the risk of recidivism for supervised offend- noticeably under the low-risk policy. This contact patterns and the recidivism rates for ers improve over time?: Examining changes in the dynamic risk characteristics for finding could indicate that low-risk offenders offenders classified as low risk. It suggests that offenders under federal supervision. Federal implementation of the low-risk policy resulted are less amenable to paying restitution under a Probation, 78(2), 15-21. policy that minimizes their contacts with offi- in less intense supervision practices and that Hughes, J. (2008). Results-based management cers. It’s important to note that restitution has the minimization of contacts with low-risk in federal probation and pretrial services. historically been paid at lower rates than fines, offenders has not jeopardized public safety. Federal Probation, 72(2), 4-14. and that this was true prior to the low-risk More rigorous research approaches could be Johnson, J., Lowenkamp, C., VanBenschoten, S., policy. Hence, while payment of restitution used to further understand how officers are & Robinson, C. (2011). The construction has decreased among post-policy low-risk modifying their supervision strategies under and validation of the Federal Post Convic- offenders, the payment rate has always been the low-risk policy and the effect of this tion Risk Assessment (PCRA). Federal less than for fines. policy on supervision outcomes of arrests and Probation, 75(2), 16-29. Among higher risk offenders, the expected revocations. Specifically, methods including Lowenkamp, C., Johnson, J., VanBenschoten, S., Robinson, C., & Holsinger, A. (2013). The increase in officer/offender contacts did not propensity score matching and regression Federal Post Conviction Risk Assessment discontinuity could be applied to introduce completely occur. Specifically, moderate-risk (PCRA): A construction and validation offenders recorded slight decreases in their statistical controls. Also, subsequent research study. Psychological Services, 10(1), 87-96. median contacts while high-risk offenders saw could examine other aspects of supervision Lowenkamp, C., Flores, A., Holsinger, A., Ma- some rise in their contact activity. Changes practices, including whether the provision of karios, M., & Latessa, E. (2010). Intensive in collateral contacts accounted for most of substance abuse and mental health treatment supervision programs: Does program the increases in contact activity for high- services and monitoring services such as drug philosophy and the principles of effective risk offenders. Conversely, personal contacts testing have declined for low-risk offenders. interventions matter? Journal of Criminal remained essentially unchanged for offenders Justice, 38(4): 368-375. in the highest risk category. It is crucial to References Lowenkamp, C., Latessa, E., & Holsinger, A. note that the number of offenders per officer Administrative Office of the U.S. Courts. (2006). The risk principle in action: What have we learned from 13,676 offenders increased by an average of 15 offenders dur- (AOUSC) (2011). An Overview of the federal Post Conviction Risk Assessment. Wash- and 97 correctional programs? Crime and ing the time period covered by this study. Delinquency, 51(1): 1-17. Moreover, the federal system has received ington, D.C.: Administrative Office of the U.S. Courts. Lowenkamp, C., & Latessa, E. (2004). Increasing an increase in the proportion of higher risk Alexander, M., & VanBenschoten, S. (2008). the effectiveness of correctional program- offenders. According to a recently published The evolution of supervision in the federal ming through the risk principle: Identifying report, the average PCRA scores rose from probation system. Federal Probation, 72(2), offenders for residential placement.Crimi - 5.09 to 6.55 between 2005 and 2011 (Baber, 15-21. nology and Public Policy, 4(2): 263-290. 2015). The increase in the number of offend- American Bar Association. (2016). Federal court The Administrative Office of the U.S. Courts. ers being supervised per officer, combined funding at http://www.americanbar.org/ (2012). Guide to judiciary policy: Volume with a rise in the risk level, may explain why advocacy/governmental_legislative_work/ 8, probation and pretrial services. Washing- median contacts have remained unchanged priorities_policy/independence_of_the_ju- ton, DC: Administrative Office of the U.S. Courts. for these high-risk offenders. Lastly, some of diciary/federal-court-funding.html Andrews, D. R., Bonta, J., & Hoge, R. D. (1990). WilsonElser. (2013). Joint and Several Li- the declines in contact activity for low-, low/ ability: 50 State Survey. Washington, DC: moderate-, and moderate-risk offenders may Classification for effective rehabilitation: Rediscovering psychology. Criminal Justice WilsonElser. be explained by budget sequestration, which and Behavior, 17, 19-52. resulted in cutbacks in officer services for all Andrews, D. R., & Bonta, J. (2010). The psychol- levels of offenders. ogy of criminal conduct (5th edition). Cincin- nati, OH: Anderson Publishing. 12 FEDERAL PROBATION Volume 80 Number 1 ArticleExamining Title Overrides of Risk Classifications for Offenders on Federal Supervision
Thomas H. Cohen1 Bailey Pendergast Scott W. VanBenschoten Probation and Pretrial Services Office Administrative Office of the U.S. Courts
THE FEDERAL PROBATION and pretrial an offender’s learning styles and potential score underrepresents his or her risk to reof- services system employs approaches grounded treatment barriers (the responsivity principle) fend. This component of the risk classification on evidence-based practices to ensure (AOUSC, 2011).2 process is referred to as professional discre- community safety and reduce recidivism The PCRA has been empirically shown to tion or supervision override and is one of (Alexander & VanBenschoten, 2008; Cohen & effectively predict the likelihood that an offender the major principles of effective evidence- VanBenschoten, 2014; Hughes, 2008). In order will recidivate during his or her supervision based supervision practices. The rationale to meet these objectives, federal probation period (Johnson, Lowenkamp, VanBenschoten, for allowing overrides in risk assessment has adopted the risk, needs, and responsiv- & Robinson, 2011; Lowenkamp, Johnson, mechanisms is that actuarial scores cannot ity (RNR) model of correctional supervision VanBenschoten, Robinson, & Holsinger, 2013; always capture the unique characteristics of practices (Alexander & VanBenschoten, 2008; Lowenkamp, Holsinger, & Cohen, 2015). individuals that officers can identify through Andrews, Bonta, & Hoge, 1990; Andrews Moreover, several studies have shown the various investigation techniques (Schmidt, & Bonta, 2010). One of the key events in PCRA’s efficacy at measuring change in an Sinclair, & Thomasdottir, 2016). Professional federal probation’s embrace of the RNR offender’s recidivism risk factors over time and overrides, hence, allow officers to depart model was the decision to construct, develop, the relationship between change in actuarial from the actuarial score when the totality of and implement the Post Conviction Risk risk and arrest outcomes (Lowenkamp et al., an offender’s characteristics suggests that the Assessment instrument (PCRA). The PCRA 2013; Cohen & VanBenschoten, 2014; Cohen, offender should be supervised at a level that is a dynamic actuarial risk assessment instru- Lowenkamp, & VanBenschoten, 2016). diverges from the risk classification. The over- ment developed for federal probation officers It is crucial to note that officers do not ride function is woven not only into the PCRA that incorporates most of the aspects of the have to supervise offenders according to their but into many risk classification instruments RNR model into federal supervision. Through original PCRA risk designations. Specifically, (Andrews et al., 1990; McCafferty, 2015). the PCRA, officers can classify offenders into judicial policy allows officers the option of As we will subsequently discuss, judicial different risk levels and identify those who are departing from the PCRA’s risk classification policy allows overrides for reasons we call most likely to recidivate (the risk principle), scheme by changing the risk level originally policy-related if the offender meets the follow- ascertain dynamic criminogenic characteris- assigned to the offender (Guide to Judiciary ing specified criteria: sex offender, persistently tics that if addressed could reduce reoffending Policy, 2014). For example, offenders placed violent, mental health issues, or serious youth- behavior (the need principle) and tailor inter- in the low-risk category by the PCRA could ful offender. Conversely, all other overrides are ventions and treatments that take into account be overridden to a higher risk level for super- labeled “discretionary overrides.” Although vision purposes should the officer, upon judicial policy gives officers the discretion to 1 Correspondence concerning this article should reviewing the offender’s profile, feel that in override, there have been few empirical efforts be addressed to Thomas H. Cohen, Administrative his or her professional judgment the PCRA to examine officer overrides in the federal Office of the U.S. Courts, One Columbus Circle, supervision system. This research will exam- NE, Washington, DC 20544. Email: Thomas_ [email protected]. The authors would like to 2 See Johnson, Lowenkamp, VanBenschoten, ine several key issues, including the overall thank our colleagues at the Administrative Office and Robinson (2011) and Lowenkamp, Johnson, prevalence of overrides for offenders under of the U.S. Courts, including Laura Baber, Nancy VanBenschoten, Robinson, and Holsinger (2013) federal supervision, the types of overrides (i.e., Beatty, and Matthew Rowland, for their helpful for information about the construction, validation, policy or discretionary) used by officers, and suggestions and comments. This publication also and implementation of the PCRA in the federal benefited from the careful editing of Ellen Fielding. supervision system. June 2016 EXAMINING OVERRIDES OF RISK CLASSIFICATIONS 13 the rationales provided by officers when using cases where the officer believes that the PCRA supervision override data were not electroni- discretionary overrides. has not adequately assessed an offender’s cally available until August 31, 2012 (n lost In this article we will also explore the risk of recidivism. The low-risk PCRA clas- = 90,585). In addition, offenders with PCRA adjustments in risk levels that occur as a result sification of an offender, for example, can be assessments occurring after 2013 were removed of overrides and whether offenders with over- overridden to a higher risk level (e.g., moder- because our recidivism follow-up period ended rides are supervised differently in terms of ate or high risk) if the officer thinks that the on December 31, 2014 (n lost = 33,818). Since their monthly officer/offender contacts and offender’s likelihood of recidivism is being we wanted to track offender recidivism patterns treatment services compared to offenders with- underestimated by the PCRA. Conversely, for at least 12 months, we excluded offenders out overrides. Last, we will examine whether officers can override the classifications of who received their PCRA assessments with offenders with overrides are recidivating at higher-risk offenders into lower risk levels if fewer than 12 months of recidivism follow-up. rates similar to their original or reclassified risk they deem that the initial PCRA calculation Despite the omission of these offenders, the levels. In other words, are the recidivism rates overstated the offender’s recidivism risk. study population mirrors that of larger popula- for low-risk offenders reclassified for supervi- The Guide also states that overrides should tions analyzed for other PCRA studies in terms sion purposes into high-risk levels similar to be relatively rare and that officers should use of their overall risk factors and demographic that of offenders initially classified as high risk, overrides for only certain case types (in the characteristics (see Lowenkamp et al., 2015). or are the arrest rates for these offenders more case of policy overrides) or supply rationales for Moreover, the percentage of offenders receiving similar to those of low-risk offenders? employing overrides (in the case of discretion- professional overrides has been relatively stable In the next section, we provide an overview ary overrides). Policy overrides involve instances over the past several fiscal years. Hence, the of the federal supervision system’s override where officers move offenders into higher or findings gleaned from these 58,500 offenders policy. Afterwards, we detail the methodologi- lower supervision levels because the offenders should be generalizable to the larger population cal framework used for this study. meet one or more of the following criteria: (1) of offenders currently under supervision in the they are classified as sex offenders, (2) they evi- federal system.5 Policy on Supervision Overrides dence patterns of persistently violent behavior, The risk and demographic characteristics of According to Volume 8E, §440, of the Guide (3) they manifest past or current indications of offenders in the study population are provided to Judiciary Policy (the Guide), officers may severe mental illness, or (4) they are youthful in Table 1. According to the PCRA, 75 percent diverge from the PCRA’s risk classification offenders with extensive criminal histories. In of offenders assessed within the study period scheme by placing offenders into different— addition to policy overrides, the Guide provides were initially classified as either low (35 percent) higher or lower—risk levels. To understand officers with latitude to issue overrides for other or low/moderate (40 percent) risk, while the how officers employ supervision overrides, we reasons; in this case they are discretionary over- remaining 25 percent fell into the moderate (19 first detail the PCRA’s risk classification mech- rides. A comprehensive justification is required percent) or high risk (6 percent) classification anism. The PCRA assesses an offender’s risk of whenever the officer decides to override an categories. Interestingly, the risk distribution recidivism through a process in which federal offender for discretionary reasons. Regardless changes somewhat once supervision overrides probation officers score offenders on 15 static of whether an officer overrides for policy or dis- are taken into account. After accounting for and dynamic risk predictors related to an cretionary reasons, any override request must be override adjustments, the percentage of offend- offender’s criminal history, education/employ- reviewed and approved by a supervising officer ers classified as low risk decreases from 35 ment, substance abuse, social networks, and (AOUSC, 2011; Guide to Judiciary Policy, 2014). percent to 31 percent, while the percentage supervision attitude characteristics.3 Officers placed in the highest risk category increases use these 15 predictors to generate a raw Data and Methods from 6 percent to 11 percent. Additional details PCRA score ranging from 0 to 18, which Participants on override adjustments will be provided in the translates into the following four risk cat- Data for this study were obtained from 94 findings section of this paper. egories: low (0-5 points), low/moderate (6-9 U.S. federal judicial districts and comprised Regarding the study population’s demo- points), moderate (10-12 points), or high (13 58,524 initial PCRA assessments conducted graphic characteristics, 57 percent were white or more points). These risk categories provide between August 31, 2012, and December 30, and 37 percent were black. Hispanics com- crucial information about an offender’s likeli- 2013. These assessments were drawn from a prised 24 percent of the sample. Over four hood of recidivism and inform officers about larger dataset containing 182,927 initial PCRA fifths (84 percent) of these offenders were the appropriate levels of supervision inten- assessments conducted within the time frame male and the average age was 39 years. Last, sity that should be allocated (AOUSC, 2011; spanning August 1, 2010, through December 85 percent were placed on supervised release, Johnson et al., 2011; Lowenkamp et al., 2013). 31, 2014.4 PCRA assessments prior to 2012 while the remainder had been directly sen- 6 The Guide incorporates the principle were excluded from this study because the tenced to a term of probation. that officers should be able to use profes- sional judgment when determining the most 5 4 We used the initial PCRA assessment date rather According to federal probation’s internal report- suitable levels of supervision intensity by than the actual supervision start date to anchor ing systems, a total of 135,468 offenders were on providing guidance on when officers should this study because when the PCRA was deployed, federal supervision as of 9/30/2015. exercise their discretion to override offend- PCRAs were done on offenders who might have 6 Supervised release refers to offenders sentenced ers (Andrews et al., 1990). According to the been well into their supervision term. Since our to a term of community supervision following a focus was on examining supervision overrides for Guide, supervision overrides should occur in period of imprisonment within the Federal Bureau all offenders receiving PCRA assessments, we were of Prisons (18 U.S.C. §3583). Probation refers to not concerned with restricting our study population offenders sentenced to a period of supervision 3 See AOUSC (2011) for a detailed discussion of to offenders with short time periods between their without any imposed incarceration sentence (18 the PCRA’s recidivism predictors. supervision start and PCRA assessment dates. U.S.C. §3561). 14 FEDERAL PROBATION Volume 80 Number 1
TABLE 1. For consistency purposes, we recoded these comparable to those offenders classified at Characteristics of Federal Offenders discretionary departures into the appropriate their original or adjusted risk levels. in Study Sample policy override categories. Hence the percent- age of offenders with policy and discretionary Findings Descriptive Offender Characteristics Information overrides reported in this study will differ Overall Prevalence of from that shown in federal probation’s internal Professional Overrides Original PCRA Risk Levels reporting systems.7 Low 34.9% We initially focus on the prevalence of over- Finally, in certain sections of this paper rides for offenders under federal supervision. Low/Moderate 40.3% we combined the policy overrides involv- Overall, 9 percent of the 58,524 offenders in Moderate 19.0% ing history of persistently violent behavior, our study population received supervision High 5.7% evidence of severe mental illness, or youthful overrides (see Table 2). Among offenders with Adjusted Supervision Levels offenders with extensive criminal histories overrides, officers overrode about two-thirds into an “other” policy override category. We Low 30.7% (68 percent) for policy reasons, while discre- combined these override types into one cat- tionary overrides accounted for the remainder Low/Moderate 38.5% egory because, as will be shown, there were of supervision adjustments. Examining the Moderate 19.4% relatively few offenders overridden for these relationship between officer overrides and High 11.4% specific policy types. initial PCRA risk levels shows that over- Supervised Release 85.0% Offender Recidivism Outcomes rides occurred more frequently for low- than Male Offender 84.3% high-risk offenders. For instance, 13 percent Recidivism is defined in this study as the Race of low-risk offenders were overridden to arrest of an offender for either a felony or another risk level compared to 9 percent of White 57.2% misdemeanor offense (excluding arrests for low/moderate and 8 percent of moderate-risk Black 36.6% technical violations) within one year after the offenders. Less than 1 percent of offenders Other 6.2% PCRA reassessment date. In addition to mea- initially classified in the high-risk category Hispanic Offender 23.9% suring any arrests, we also identified arrests were overridden to a lower supervision level. for violent offenses committed within one Mean Age 39.3 yrs. A combination of sex offender policy and dis- year after the initial PCRA assessment. Violent cretionary overrides drove the override rates Number of Offenders 58,524 arrests were defined using the definitions for lower-risk offenders. Interestingly, other- Note: Includes offenders with PCRA assessments that from the National Crime Information Center occurred between August 31, 2012, and December policy overrides were slightly more frequent 31, 2013. (NCIC), which included homicide and related for moderate (3 percent) than for low-risk (1 offenses, kidnapping, rape and sexual assault, percent) offenders. robbery, and assault. The recidivism data were Table 3 shows override rates by an offend- Measuring Discretionary gathered through the NCIC and Access to Law er’s most serious conviction offense and Supervision Overrides 8 Enforcement System databases (ATLAS). demographic characteristics. The override In this study, we explored the rationales pro- Analytical Plan rate was highest for sex offenders; over three- vided by officers for discretionary overrides in fourths of these offenders (77 percent) were greater detail. This research presented several The current study uses descriptive statistics placed into supervision levels that differed challenges in that officers can, and often do, to explore overrides for offenders on federal from their initial PCRA risk classifications. provide extensive written rationales in the text supervision. It examines the overall frequency In addition, the override rates for offenders fields when justifying a discretionary override. of overrides and investigates the types of over- convicted of firearms (12 percent) and violent While these text fields provide rich rides (e.g., policy or discretionary) employed offenses (11 percent) were slightly higher than information about an offender’s risk char- by officers, with specific inquiries into the the 9 percent baseline override rate. The fact acteristics, they do not lend themselves to rationales used for discretionary overrides. that most sex offenders were overridden to quantifiable analysis. We addressed this issue This research also explores the adjustments higher risk levels is not surprising, since policy by using text-mining techniques to categorize in risk levels that result from overrides and provides officers with discretion to adjust the these rationales into broader groups such whether officers deliver supervision services supervision levels for these offenders upwards as substance abuse problems, evidence of commensurate with the reclassified risk level. at the beginning of supervision while the offi- noncompliant behavior, electronic monitor- The research then analyzes whether the recidi- cer thoroughly assesses the offender (Guide to ing, and gang activity, which could be used vism rates for offenders with overrides are Judiciary Policy, 2014). Offenders convicted for analytical purposes. Ultimately, we were 7 See Decision Support Systems (DSS) report of firearms and violent offenses also gar- able to successfully classify 90 percent of the #1193 on policy and discretionary override rates. nered overrides at higher rates, because they 3,121 discretionary overrides into broader 8 ATLAS is a software program used by the are more likely to have characteristics that categories. Interestingly, 45 percent of the Administrative Office of the U.S. Courts that pro- would justify policy overrides for persistently 3,121 discretionary overrides were identified vides an interface for performing criminal record violent behavior. through this text-mining process as having checks through a systematic search of official state and federal rap sheets. It is widely used by probation occurred for policy reasons. In other words, and pretrial services officers to perform criminal officers had provided policy justifications record checks on defendants and offenders for (e.g., sex offender; offender has serious mental supervision and investigation purposes (Baber, health issues) for the discretionary overrides. 2010). June 2016 EXAMINING OVERRIDES OF RISK CLASSIFICATIONS 15
TABLE 2. Percent of Federal Offenders with Any, Policy, or Discretionary Overrides
Policy Overrides Initial PCRA risk Number of offenders All overrides Any Sex offender Other Discretionary Override Total 58,524 9.4% 6.5% 4.6% 1.8% 2.9% Low 20,439 12.5% 8.2% 6.8% 1.4% 4.3% Low/Moderate 23,599 8.5% 5.9% 3.8% 2.1% 2.6% Moderate 11,130 8.4% 6.5% 3.8% 2.7% 1.9% High 3,356 0.2% 0.0% 0.0% 0.0% 0.2% Note: Other policy includes mental health, persistently violent, and youthful offender overrides.
TABLE 3. Types of Policy and Percent of Federal Offenders with Policy or Discretionary Overrides, by Offense Discretionary Overrides Type and Demographic Characteristics Next, we examine the types of policy and dis- Percentage of Offenders with cretionary overrides used by officers. Figure Offense & Number of Any 1 focuses on policy overrides. This figure Demographics Offenders Overrride Policy Discretionary includes overrides that were originally sub- Conviction offense mitted as discretionary before being re-coded into policy overrides. Nearly three-fourths (72 Sex Offense 2,268 76.6% 75.9% 0.7% percent) of policy overrides were for offend- Firearms 8,667 12.4% 9.9% 2.5% ers who met the sex offender criteria. The Violence 2,809 10.5% 7.2% 3.3% remainder of policy overrides involved severe Other 2,829 8.7% 4.7% 4.0% mental illness (16 percent) and persistently White Collar 10,963 6.1% 2.5% 3.6% violent behavior (12 percent). Figure 2 displays the most common discre- Drug 26,865 5.0% 2.0% 3.0% tionary overrides. In over a third (35 percent) Immigration 2,581 3.4% 1.6% 1.8% of discretionary overrides, the officer cited Public Order 1,417 3.5% 1.1% 2.5% the offender’s substance abuse problems as Gender a reason for adjusting the supervision level. Male 49,326 10.1% 7.2% 2.9% Evidence of noncompliant behavior accounted Female 9,198 5.6% 2.5% 3.1% for 17 percent of discretionary overrides, while 10 percent or more occurred because Race the officer indicated that issues related to loca- Other 3,600 11.8% 9.7% 2.1% tion monitoring (13 percent), employment White 33,382 11.1% 8.0% 3.1% (12 percent), criminal history (11 percent), or Black 21,385 6.4% 3.6% 2.8% financial penalties (10 percent) necessitated a 9 Ethnicity supervision adjustment. Not Hispanic 43,887 10.7% 7.6% 3.1% It is notable that some of the justifications provided by officers for discretionary over- Hispanic 13,749 5.3% 2.9% 2.4% rides are already being measured through Note: Offense types excludes offenders convicted of escape/obstruction, technical violations, and other offenses. the PCRA. For example, evidence of non- compliant behavior and criminal history are currently measured in the PCRA’s criminal history section, while employment issues and substance abuse problems are scored in its education/employment and substance abuse domains. Moreover, other rationales such as location monitoring and the collection of financial penalties suggest that issues related to workload and case activity might be driving the override decision rather than enhanced recidivism risk. Substantial amounts of officer
9 Because officers write their rationales for over- rides, multiple reasons could be attributed to one offender. 16 FEDERAL PROBATION Volume 80 Number 1
was adjusted by one or multiple levels. Sex FIGURE 1. offender policy overrides, for example, almost Override Types for Federal Offenders with Policy Overrides always resulted in offenders being placed in the highest risk category, irrespective of e o e er l o their initial PCRA risk designation. Over 80 or urre m om percent of low risk and nearly 90 percent of 1 o e ere me l ll e low/moderate risk offenders with sex offender