MEASURING CHANGE FROM RATES OF RECIDIVISM TO MARKERS OF DESISTANCE

Author(s): CECELIA KLINGELE

Source: The Journal of Criminal Law and Criminology (1973-), Vol. 109, No. 4 (2019), pp.
769-817
Published by: Northwestern University Pritzker School of Law
Stable URL: https://www.jstor.org/stable/48572943
Accessed: 10-11-2024 23:59 UTC

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/48572943?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Northwestern University Pritzker School of Law is collaborating with JSTOR to digitize,

preserve and extend access to The Journal of Criminal Law and Criminology (1973-)

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
0091-4169/19/10904-0769
THE JOURNAL OF CRIMINAL LAW & CRIMINOLOGY Vol. 109, No. 4
Copyright © 2019 by Cecelia Klingele Printed in U.S.A.

MEASURING CHANGE: FROM RATES OF

RECIDIVISM TO MARKERS OF
DESISTANCE

CECELIA KLINGELE*
Reducing the incidence of crime is a primary task of the criminal justice
system and one for which it rightly should be held accountable. The system’s
success is frequently judged by the recidivism rates of those who are subject
to various criminal justice interventions, from treatment programs to
imprisonment. This Article suggests that, however popular, recidivism alone
is a poor metric for gauging the success of criminal justice interventions or
of those who participate in them. This is true primarily because recidivism
is a binary measure, and behavioral change is a multi-faceted process.
Accepting recidivism as a valid, stand-alone metric imposes on the criminal
justice system a responsibility beyond its capacity, demanding that its
success turn on transforming even the most serious and intractable of
offenders into fully law-abiding citizens. Instead of measuring success by
simple rates of recidivism, policymakers should seek more nuanced metrics.
One such alternative is readily available: markers of desistance. Desistance,
which in this context means the process by which individuals move from a
life that is crime-involved to one that is not, is evidenced not just by whether
a person re-offends but also by whether there are increasing intervals
between offenses and patterns of de-escalating behavior. These easily
obtainable metrics, which are already widely relied on by criminologists, can
yield more nuanced information about the degree to which criminal justice
interventions correlate with positive (or negative) life changes. They also
resemble more closely the ways in which other fields that address behavioral
change such as education attempt to measure change over time. Measuring

* Associate Professor of Law, University of Wisconsin Law School. Many thanks to

Shannon Toole and Andrew Lorenz for their diligent assistance, and to Steven Wright, Keith
Findley, Jason Yackee, Miriam Seifter, Rob Yablon, Heinz Klug, Yaron Nili, Susannah Tahk,
Rabea Benhalim, Andrew Turner, and Margaret Raymond for their discussion and comments
on early drafts. Special thanks to Wendy Henderson and Bryn Martyna for their inspiring work
promoting desistance among young people, and for encouraging me to reduce to paper the
ideas discussed in this Article. +JMJ

769

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
770 KLINGELE [Vol. 109

the success of criminal justice interventions by reference to their effects on

desistance would mean seeking evidence of progress, not perfection. Such
an approach would allow criminal justice agencies to be held accountable
for promoting positive change without asking them to do the impossible,
thereby creating new pathways by which the criminal justice system could be
recognized for achieving real and measurable progress in crime reduction.

INTRODUCTION ............................................................................. 770

I. DEFINING SUCCESS IN CRIMINAL JUSTICE ........................ 777
II. RECIDIVISM AS A SUCCESS METRIC ................................... 783
A. Defining Recidivism ....................................................... 783
B. Reporting Recidivism ...................................................... 790
III. BEHAVIORAL CHANGE AND DESISTANCE FROM CRIME
................................................................................................ 794
A. Understanding Change .................................................... 794
B. Measuring Progress ......................................................... 801
IV. THE LIMITS OF RECIDIVISM DATA..................................... 806
A. Data Gaming ................................................................... 807
B. Premature Termination of Promising Programs.............. 813
C. Risk Aversion .................................................................. 815
CONCLUSION: TOWARD MARKERS OF DESISTANCE .......... 816

INTRODUCTION
The U.S. criminal justice system operates on an estimated annual budget
of more than 180 billion dollars.1 At any given time, the system employs
more than 870,000 police officers,2 32,800 prosecutors,3 an untallyable

1 John F. Pfaff, Criminal Punishment and the Politics of Place, 45 FORDHAM URB. L. J.
571, 579 (2018); TRACEY KYCKELHAHN, U.S. BUREAU OF JUST. STAT., JUSTICE EXPENDITURE
AND EMPLOYMENT EXTRACTS, 2012-PRELIMINARY (2015), https://www.bjs.gov/index.cfm?ty
=pbdetail&iid=5239 [https://perma.cc/2T8H-ER4D].
2 DUREN BANKS ET AL., N ATIONAL SOURCES OF LAW ENFORCEMENT E MPLOYMENT DATA,
BUREAU OF JUST. STAT. 2, T.1 (2016), https://www.bjs.gov/content/pub/pdf/nsleed.pdf
[https://perma.cc/EK8V-8742] (reporting on number of sworn state law enforcement officers);
BRIAN A. REAVES, FEDERAL LAW ENFORCEMENT OFFICERS, 2008, BUREAU OF JUST. STAT. 1
(2012), https://www.bjs.gov/content/pub/pdf/fleo08.pdf [https://perma.cc/P62R-YKUX]
(reporting 120,000 full-time federal law enforcement officers in 2008).
3 STEVEN W. PERRY & DUREN BANKS, PROSECUTORS IN STATE COURTS, 2007—
STATISTICAL TABLES, BUREAU OF JUST. STAT. 4, T. 3 (2011), https://www.bjs.gov/content/pub/
pdf/psc07st.pdf [https://perma.cc/ZFJ2-JKXC] (reporting that 35% of the 77,927 people
employed full-time by state prosecutors’ offices were chief or assistant prosecuting attorneys);
Daniel Richman, Political Control of Federal Prosecutions: Looking Back and Looking
Forward, 58 DUKE L.J. 2087, 2088 (2009).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 771

number of criminal defense attorneys,4 and over 502,000 correctional

officials.5 It prosecutes more than fifteen million cases each year6, and
imprisons or supervises more than 6.6 million people.7 Given the size of the
criminal justice infrastructure, it is not surprising that policymakers and the
public at large have pressed for more transparency about the effectiveness—
or ineffectiveness—of various criminal justice practices and programs.8 The

4 There is no direct estimate of the number of criminal defense lawyers in the United
States, though the National Association of Criminal Defense Attorneys—a completely
voluntary organization for private and public defense lawyers, and other interested
stakeholders—boasts over 40,000 members. Nat’l Assoc. Crim. Def. Lawyers, About NACDL,
https://www.nacdl.org/about.aspx [https://perma.cc/22QX-8KYP]. Estimates suggest that in
2008, state and county indigent defense agencies employed more than 15,000 criminal defense
attorneys. LYNN LANGTON & DONALD FAROLE, STATE PUBLIC DEFENDER PROGRAMS, 2007,
BUREAU OF JUST. STAT. 3 T.1 (2010), https://www.bjs.gov/content/pub/pdf/spdp07.pdf
[https://perma.cc/R9Q5-CSSZ]. The nation’s 94 federal judicial districts employ more than
12,000 panel attorneys under the Criminal Justice Act and maintain a combined lawyer and
support staff of more than 3,700 people. DEFENDER SERVICES, U.S. COURTS, http://www.
uscourts.gov/services-forms/defender-services [https://perma.cc/4RPV-FUV4].
5 BUREAU OF LABOR STATISTICS, CORRECTIONAL OFFICERS AND JAILERS, MAY 2018, https:
//www.bls.gov/oes/current/oes333012.htm#(1) [https://perma.cc/PY8L-WDTW]; BUREAU OF
LABOR STATISTICS, PROBATION OFFICERS AND CORRECTIONAL TREATMENT SPECIALISTS, MAY
2018, https://www.bls.gov/oes/current/oes211092.htm#st [https://perma.cc/5K7D-EKWT]
(reporting 87,660 jobs in 2018).
6 Federal Judicial Caseload Statistics 2018, UNITED STATES COURTS,
https://www.uscourts.gov/statistics-reports/federal-judicial-caseload-statistics-2018 [https://
perma.cc/T8Q4-2XP4] (last visited May 5, 2019) (showing 81,553 newly-filed criminal cases
in 2018); R. Schauffler et al., ed., Court Statistics Project DataViewer, COURT STATISTICS
PROJECT (Jan. 11, 2017), www.courtstatistics.org [https://perma.cc/5EMT-CFUT] (last visited
May 5, 2019) (reporting 15,125,229 new criminal cases filed in 2017 in 40 states and the
District of Columbia).
7 DANIELLE KAEBLE & MARY COWHIG, CORRECTIONAL POPULATIONS IN THE UNITED
STATES, 2016, BUREAU OF JUST. STAT. 2 (2018), https://www.bjs.gov/content/pub/pdf/cpus16.
pdf [https://perma.cc/QWM5-3D3V].
8 See, e.g., Josh Salman, ‘A Game-Changer’: Florida House Passes Criminal Justice
Reform Bill: Newspaper Series on Sentencing Disparities Got Ball Rolling on Legislation,
SARASOTA HERALD-TRIBUNE (Feb. 22, 2018), http://www.heraldtribune.com/news/20180222/
a-game-changer-florida-house-passes-criminal-justice-reform-bill [https://perma.cc/9SRB-C
KQS] (last visited May 5, 2019) (reporting on new law to make criminal justice data more
transparent as a result of media reporting on data inaccuracies and racial disparities in
sentencing); Mike Nerheim & Meg Reiss, The Criminal Justice System Needs to Start
Learning from Its Mistakes, THE HILL (Nov. 18, 2017), http://thehill.com/opinion/criminal-
justice/360969-the-criminal-justice-system-needs-to-start-learning-from-its [https://perma.cc
/YU2R-ZC93] (last visited May 5, 2019) (“As democratically elected officials, prosecutors’
legitimacy is premised on public trust. Although they are tasked with holding people
accountable, they are rarely held accountable . . . . In order to actively build public trust and
foster public safety, prosecutors should scrutinize existing systems that allow any wrongdoing
or practice to occur . . . and support implementing processes and procedures that enhance
transparency of and accountability for their decisions.”); David A. Graham, What Can the U.S.

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
772 KLINGELE [Vol. 109

fundamental question is whether the system is worth its significant costs in

both resources and human lives.
While demands for accountability are both prudent and fair,
determining how to measure effectiveness is difficult. Whether the focus of
evaluation is on the success of system actors (such as police, prosecutors, or
judges), participants (such as arrestees, defendants, or prisoners), or specific
interventions (such as drug courts, probation supervision, or restorative
justice programs); it is not always clear how success should be defined or
measured. If the goal of the criminal justice system is to advance public
safety and promote proportional accountability for wrongdoing,9 then the
best metrics would be those that reveal how well system actors prevent
criminal harm or restore community confidence that justice has been served.
Such outcomes are difficult to quantify, however, and policy analysts often
default to measurements that are easier to gather: number of arrests made,
amount of restitution collected, or number of convictions secured.10
Those within the criminal justice system and those outside it rely
heavily on another measure of success: rates of recidivism. Recidivism rates
are one of the primary ways that legislators, policymakers, grant funders,
media outlets, and criminal justice system actors determine whether specific
criminal justice interventions have succeeded or failed.11 As Joan Petersilia
wrote in her authoritative article on recidivism:

Do to Improve Police Accountability?, THE ATLANTIC (Mar. 8, 2016), https://www.theatlantic.

com/politics/archive/2016/03/police-accountability/472524/ [https://perma.cc/96M8-R24G]
(last visited May 5, 2019) (“The lack of reliable information on policing has been a major
hindrance to discussions.”).
9 There are, of course, many purposes to the criminal justice system. See generally H. L.
A. HART, Prolegomenon to the Principles of Punishment, in PUNISHMENT AND
RESPONSIBILITY: ESSAYS IN THE PHILOSOPHY OF LAW (1968). Some of these exist in tension
with each other, and others co-exist more comfortably. See, e.g., Kevin M. Carlsmith et al.,
Why Do We Punish? Deterrence and Just Deserts as Motives for Punishment, 83 J.
PERSONALITY & SOC. PSYCH. 284 (2002) (discussing tensions between retributive and
deterrent philosophies of punishment). Among the most traditionally recognized of these are
the need for public protection and the desire to hold offenders morally accountable for their
willful wrongdoing. See generally Michael Tonry, Thinking About Punishment, in WHY
PUNISH? HOW MUCH? A READER ON PUNISHMENT (Michael Tonry ed., 2011).
10 Cf. Cecelia Klingele et al., Reimagining Criminal Justice, 2010 WIS. L. REV. 953, 995
(2010) (“Agencies like quantifiable statistics, and prevention is notoriously difficult to
quantify.”).
11 See, e.g., CONN. GEN. STAT. A NN. § 18-81w(b) (2007) (“The success of the reentry
strategy shall be measured by: (1) The rates of recidivism and community
revictimization . . . .”); TEX. GOV’T CODE § 509.003(a)(3) (2015) (“The [Community Justice
Assistance Division] shall propose and the board shall adopt reasonable rules
establishing . . . methods for measuring the success of community supervision and corrections
programs, including methods for measuring rates of diversion, program completion, and

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 773

Defining and measuring recidivism are . . . central to answering the question “How
well are we doing?” It has been said that recidivism rates are to the criminologist what
the Geiger counter is to the geologist. In other words, they are the most objective overall
basis we have for evaluating the performance of justice agencies.12
While criminologists have developed nuanced ways of gathering and
interpreting recidivism data, criminal justice agencies typically examine
recidivism rates in isolation from other available measures of success.13 On
its face, recidivism seems a sensible metric: re-offense is what recidivism
rates purport to measure, and a reduction in crime is undeniably a primary
goal of the criminal justice system. Moreover, when a convicted person does

recidivism.”); UTAH CODE ANN. 1953 § 64-13-25(4)(a) (2015) (“The department shall
collaborate with the Division of Substance Abuse and Mental Health to develop and
coordinate the performance goals and outcome measurements, including recidivism rates and
treatment success and failure rates.”); Alana Saulniera & Diane Sivasubramaniam, Restorative
Justice: Underlying Mechanisms and Future Directions, 18 NEW CRIM. L. REV. 510, 516
(2015) (“Rates of recidivism are perhaps considered the ultimate measure of success and are
one of the most frequently compared outcomes of restorative and retributive procedures.”);
Kevin Rector, Justice Department Gives Baltimore Police $750,000 Grant to Fight
Recidivism, BALTIMORE SUN, Sept. 28, 2017 https://www.baltimoresun.com/news/crime/bs-
md-ci-doj-recidivism-grant-20170928-story.html [https://perma.cc/CN3V-KC28]; NAT’L
REENTRY RESOURCE CENTER, SECOND CHANCE ACT STATEWIDE RECIDIVISM REDUCTION
PROGRAM https://csgjusticecenter.org/nrrc/second-chance-act-recidivism-reduction-grant-
program/ [https://perma.cc/74ZZ-SDXC] (last visited May 5, 2019) (“Among the most critical
priorities for . . . grant recipients is to implement evidence-based practices and core
correctional practices . . . . States that receive SRR grants utilize the funds to pursue an
intensive, collaborative process that brings the governor’s office, state policymakers, and
corrections leaders together to set measurable recidivism-reduction goals and develop
practical, data-driven plans to achieve them.”).
12 Joan Petersilia, Recidivism, in ENCYCLOPEDIA OF AMERICAN PRISONS 382, 382 (Marilyn
D. McShane & Frank P. Williams III eds., 1996) (citing Korn and McCorkle 24 (1966)); see
also Janet L. Jackson et al., A Critical Look at Research on Alternatives to Custody, 59 FED.
PROB. 43, 44–45 (1995) (“A considerable number of empirical studies comparing the
effectiveness of different sanctions are to be found in the literature. . . . Although other aspects
have been considered by some researchers, in general, recidivism has been the most important
and frequently exclusive indicator of effectiveness.”).
13 See, e.g., CENTER FOR STATE G OVERNMENTS, REDUCING RECIDIVISM: STATES DELIVER
RESULTS (2017) https://csgjusticecenter.org/wp-content/uploads/2017/06/6.12.17_Reducing-
Recidivism_States-Deliver-Results.pdf [https://perma.cc/BM43-ZDDV] (reporting on
cohort-based and revocation-based recidivism for seven states); NANCY LA VIGNE ET AL.,
JUSTICE REINVESTMENT AT THE LOCAL LEVEL: PLANNING AND IMPLEMENTATION GUIDEBOOK,
URB. INST. (2010), http://www.urban.org/UploadedPDF/412233-Justice-Reinvestment.pdf
[https://perma.cc/78DH-SDK2]; COUNCIL OF JUVENILE CORRECTIONAL ADMINISTRATORS
DEFINING AND MEASURING RECIDIVISM (2009), https://www.ncjrs.gov/App/Publications/abst
ract.aspx?ID=259797 [https://perma.cc/6XFP-QS7F]; Martha Lyman & Stefan LoBuglio,
‘Whys’ and ‘Hows’ of Measuring Jail Recidivism, AM. JAILS 9 (Mar./Apr. 2007), https://www.
montgomerycountymd.gov/COR/Resources/Files/PDF/jail_recidivism.pdf [https://perma.cc/
2A4H-FN7J]

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
774 KLINGELE [Vol. 109

not offend again, that is a success. Even so, there is a difference between
commending those who abandon crime entirely—whether by virtue of a
criminal justice intervention or otherwise—and saying that the criminal
justice system fails whenever it does not fully transform law-breakers into
models of perfect compliance.
One problem with recidivism is that it is a binary measure: either a
person commits a new crime, or he does not. Absent from data that measure
rates of recidivism is an appreciation for the nuances of human behavioral
change. The addict who stops selling drugs (but shoplifts a few canned
goods) and the batterer who again assaults his wife are both “recidivists,” but
there is a clear distinction between them. Recidivism as a metric is not
sensitive to reductions in the severity or frequency of offending, even though
such reductions often serve as markers of progress and indicate a reduction
in harm caused to the community.14 By over-relying on recidivism rates to
gauge success, policymakers and system actors alike risk underappreciating
change by individual defendants and undervaluing the criminal justice
interventions that move people forward. Moreover, by looking only at
whether past offenders have recidivated, rather than at how often and in what
ways they recidivate, system actors risk missing clues about escalating
dangers or underestimating the harm inflicted by ill-conceived interventions.
So, how should success and failure be measured when it comes to the
criminal justice system and those in it? Much of that depends on what we are
trying to measure and what resources we have available to capture data.
While there are many ways to gauge individual and programmatic success,15
this Article focuses on one particularly simple alternative to recidivism:
markers of desistance.
Criminologists have long studied what makes people stop committing
crimes. Studies have examined youth and adults at various life stages and
have identified a myriad psychological, social, physical, and environmental
factors that are correlated with desistance16—that is, the process by which
people disentangle themselves from criminal behavior and connect to

14 See infra Section III.A.

15 These include examining individuals’ pre and post-intervention health, substance abuse
rates, economic prosperity, community engagement, and self-reported satisfaction with
programs and program staff. Cf. Brandy L. Blasko et al., Performance Measures in
Community Corrections: Measuring Effective Supervision Practices with Existing Agency
Data, 80 FED. PROB., 3, 26, 28 (2016) (advocating for an examination of factors such as
supervisees’ program engagement, negative drug tests, speed of referral, and level of
supervision as important markers of success and failure for community supervision agencies).
16 See John H. Laub & Robert J. Sampson, Understanding Desistance from Crime, 28
CRIME & JUST. 1, 17–25 (2001) (reviewing studies of correlates of desistance).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 775

prosocial activities and associates.17 While the path to desistance is not

always a straight one,18 the data by which sociologists track behavior often
focuses on the severity and frequency of an individual’s criminal behavior.19
Serious “persisters” can be identified by their often frequent involvement in
crime and sometimes by an escalation in their offense severity.20 “Desisters,”
by contrast, can be identified by their sometimes instant, but more often
gradual, termination of criminal behavior.21 This more nuanced, academic
approach to gathering and analyzing crime data has, for the most part, not
translated into changes in the way criminal justice administrators measure
and report rates of recidivism.22
When desistance is discussed in the legal criminal justice literature, it is
often used as an antonym for recidivism, rather than as a description of the
process by which one progresses toward the end-goal of complete
compliance with the law.23

17 Cf. id. at 11 (“Termination is the time at which criminal activity stops. Desistance, by
contrast, is the causal process that supports the termination of offending.”); Shawn D.
Bushway et al., An Empirical Framework for Studying Desistance as a Process, 39
CRIMINOLOGY 491, 491–92 (2001) (“Most criminologists define desistance as the state of
having ‘terminated’ offending. But recently, criminologists have begun to reexamine and
expand the definition of desistance to include attention to the process by which people arrive
at this state of nonoffending.”); Shawn D. Bushway et al., Desistance as a Developmental
Process: A Comparison of Static and Dynamic Approaches, 19 J. QUANT. CRIMINOLOGY 129,
133 (2003) (defining desistance “as the process of reduction in the rate of offending from a
nonzero level to a stable rate empirically indistinguishable from zero. The key to this definition
is that it views desistance as a process, not a state”).
18 See generally Alex R. Piquero, Somewhere Between Persistence and Desistance: The
Intermittency of Criminal Careers, in AFTER CRIME AND PUNISHMENT: PATHWAYS TO
OFFENDER REINTEGRATION 102 (Shadd Maruna & Russ Immarigeon eds., 2004).
19 See Rolf Loeber & Marc LeBlanc, Toward a Developmental Criminology, 12 CRIME &
JUST. 375, 409 (1990) (“[D]esistance concerns a slowing down in the frequency of offending
(deceleration), a reduction in its variety (specialization), and a reduction in its seriousness (de-
escalation).”).
20 See Terrence P. Thornberry, Explaining Multiple Patterns of Offending Across the Life
Course and Across Generations, 600 A NN. AM. ACADEMY POL. & SOC. SCI. 156, 159 (2005)
(“[A]lthough some involvement in offending is common, relatively few offenders will have
extensive criminal careers. Extensiveness includes such dimensions as frequency, duration,
seriousness, and, especially, the co-occurrence of these dimensions.”).
21 See Loeber & LeBlanc, supra note 19, at 409.
22 See RYAN KING & BRIAN ELDERBROOM, IMPROVING RECIDIVISM AS A PERFORMANCE
MEASURE, URB. INST. 2–3 (2014), https://www.bja.gov/Publications/UI-ImprovingRecidivis
m.pdf [https://perma.cc/2QDD-NSZC] (advocating for the broader use by administrators of
some of the metrics used by criminologists, including time-to-new-offense).
23 See, e.g., Laub & Sampson, supra note 16, at 5 (discussing “conceptual issues” in the
definition of desistance).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
776 KLINGELE [Vol. 109

Some of the metrics gathered by criminologists—particularly intervals

between offenses and offense severity over time—are easily ascertainable by
reference to the recidivism data already collected and analyzed by the
criminal justice system. While such data remain an imperfect measure of
change, utilizing these “markers of desistance” would enable policymakers
and criminal justice stakeholders to develop a significantly more nuanced
picture of offenders’ aggregate and individual behavioral change and of how
criminal justice interventions positively and negatively affect that change.
Rather than limiting the definition of “success” to those state interventions
that (rather implausibly) claim to fully eradicate criminal behavior among
some fraction of their participants, success should also be understood to
include those programs that move people forward on the path of desistance.
Similarly, by seeing desistance as a process, system actors might better
discern both positive and negative behavioral changes in repeat offenders,
allowing them to contextualize the statistical predictions of individual
recidivism risk currently used by criminal justice agencies at all stages of the
criminal process.
This Article speaks broadly to criminal justice scholars, administrators,
and stakeholders but especially to those who lack specialized training in the
statistical methods used by criminologists to study recidivism and desistance.
It explains to a legal audience the limitations of relying on recidivism rates
as a measure of systemic or individual “success” and encourages the adoption
of desistance over time as an alternative measure. It does so for three reasons.
First, it is relatively easy to do: markers of desistance can be easily derived
from existing recidivism data. Second, focusing on markers of desistance
increases the system’s ability to accurately discern which criminal justice
programs are improving offender behavior and which are causing greater
harm. Finally, training system stakeholders to see success in markers of
desistance, and not just in the absence of recidivism, can shape the way
judges, lawyers, and correctional administrators employ the recidivism risk
prediction tools that are increasingly being adopted by criminal justice
agencies.
Part I describes the many ways recidivism is currently used to measure
and predict the success or failure of programs and people within the criminal
justice system. Part II dissects the concept of recidivism, examining the
many ways in which it has been defined and measured and translating for
legal audiences the meanings of recidivism rates as they are reported by
program analysts. Part III explores how criminology and analogous fields
understand the phenomenon of behavioral change and measure the success
of programmatic interventions. In light of the mechanisms of behavioral
change discussed in Part III, Part IV examines the problems created directly

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 777

and indirectly by overreliance on recidivism data, including incentives for

data gaming, the premature termination of otherwise promising programs for
advancing public safety, and excessively risk-averse behavior by system
actors. Part V concludes with a call for change in how policymakers and
criminal justice system actors think about and measure success in the
criminal justice system. It encourages administrators and policymakers to
draw on markers of desistance to provide a significantly more nuanced
picture of individual change over time, and of the role that criminal justice
interventions may play in promoting or hindering that change process.

I. DEFINING SUCCESS IN CRIMINAL JUSTICE

It is easy to imagine any number of ways in which the effectiveness of
the criminal justice system might be assessed. Efforts could be undertaken
to quantify reductions in crime24 or improvements in the degree to which
people feel safe when going about their daily business.25 Policymakers could
assess metrics of community health26 or the satisfaction of stakeholders, from
crime victims to defendants.27 We measure what we value, and for that
reason any metric we select is likely to tell us as much about those who
operate the criminal justice system as it is to tell us about the effectiveness
of the system itself. While more robust metrics (like those listed above) are
collected and analyzed in isolated instances, policymakers and criminal
justice stakeholders tend to give minimal attention to measures beyond
recidivism.28

24 Paul Ekbloma & Ken Pease, Evaluating Crime Prevention, 9 CRIME & JUST. 585, 643
(1995) (detailing the challenges of quantifying the effects of crime prevention efforts,
including problems with “[b]ackground fluctuation in the variables of interest, uncertainties
in the interpretation of cause and effect, and vagaries in how programs are implemented are
among the recurring problems”).
25 Wesley G. Skogan, Measuring What Matters: Crime, Disorder, and Fear, in
MEASURING WHAT MATTERS: PROCEEDINGS FROM THE POLICING RESEARCH INSTITUTE
MEETINGS, NAT’L INST. JUST. 37, 47–50 (Robert H. Langworthy, ed. 1999), https://www.ncjrs.
gov/pdffiles1/nij/170610.pdf [https://perma.cc/5G47-S28L] (discussing ways to measure fear
of crime).
26 See, e.g., Susanne Mayer et al., Health-Related Resource-Use Measurement
Instruments for Intersectoral Costs and Benefits in the Education and Criminal Justice
Sectors, 35 PHARMACOECONOMICS 895 (2017).
27 See JULIAN V. R OBERTS, UNDERSTANDING PUBLIC ATTITUDES TO CRIMINAL JUSTICE 32
(2005) (“Confidence or trust has been measured using a number of different questions.
Pollsters have asked questions that address issues of fairness and integrity, and ones that
address issues of competence or effectiveness.”).
28 John J. DiIulio, Rethinking the Criminal Justice System: Toward a New Paradigm, in
PERFORMANCE MEASURES FOR THE CRIMINAL JUSTICE SYSTEM, BUREAU OF JUST. STAT. 1
(1992) (“Rates of crime and recidivism have long served as critical measures for the
performance of the Nation’s criminal justice system. These measures represent the basic goals

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
778 KLINGELE [Vol. 109

Recidivism data, both historic and predictive, are used to define success
in multiple ways within the criminal justice system. Some focus on
individual defendants, as when judges or probation officers use actuarial
instruments to predict particular defendants’ statistical risks of future
recidivism.29 Others use recidivism data as “a policy outcome” to report on
the success (or failure) of “new correctional or reentry experiments.”30 Still
others assess the effectiveness of the criminal justice system as a whole, such
as reports by state and federal justice departments that quantify recidivism
rates for those exiting correctional institutions. 31
Governmental agencies are perhaps the largest consumers of recidivism
data. The website of the National Institute of Justice (NIJ) proudly proclaims
that “[r]ecidivism research is embedded throughout NIJ-sponsored research
in sentencing, corrections and policy intervention evaluations. Many NIJ-
funded studies of community supervision depend on recidivism measurement

of public safety to which all components of the criminal justice system contribute. At the same
time, however, rates of crime and recidivism are not the only, or necessarily the best, measures
of what criminal justice institutions do.”). See Chris Cunneen & Garth Luke, Recidivism and
the Effectiveness of Criminal Justice Interventions: Juvenile Offenders and Post Release
Support, 19 CURRENT ISSUES IN CRIM. JUST. 198 (2007) (“[O]ur argument here is that
[measures of recidivism] appear to now outweigh all other measures when considering the
impact of particular criminal justice policies, programs and other types of interventions.”).
29 See, e.g., J.S. v. State, 928 N.E.2d 576, 579 (Ind. 2010) (explaining that the function of
LSI-R “is to supplement and enhance a judge’s evaluation, weighing, and application of the
other sentencing evidence in the formulation of an individualized sentencing program
appropriate for each defendant”); Jones v. Com., No. 2000-CA-001746-MR, 2003 WL
21713776 (Ky. Ct. App. July 25, 2003) (“[S]exual-offender risk-assessment reports . . . are
now conducted as part of a comprehensive pre-sentencing evaluation. The purpose of such
reports [is] to provide the trial court with a recommendation assessing the defendant’s risk of
re-committing a sex crime, the threat which the defendant poses to public safety, the
defendant’s amenability to sex offender treatment, and the nature of the required treatment.”);
State v. Loomis, 881 N.W.2d 749 (Wis. 2016) (discussing use of COMPAS risk assessment
results at sentencing).
30 See Robert Weisberg, Meanings and Measures of Recidivism, 87 S. CAL. L. REV. 785,
788 (2014).
31 See, e.g., CALIFORNIA DEPARTMENT OF CORRECTIONS AND REHABILITATION 2017
OUTCOME EVALUATION REPORT: AN EXAMINATION OF OFFENDERS RELEASED IN FISCAL YEAR
2012–13 (2017), https://sites.cdcr.ca.gov/research/wp-content/uploads/sites/9/2018/04/2017-
Outcome-Evaluation-Report.pdf [https://perma.cc/CF2U-LKLZ]; RICHARD L. WIENER, UNIV.
OF NEBRASKA LAW & PSYCH. PROGRAM, R ECIDIVISM RATES FOR NEBRASKA ADULT
PROBATIONERS: 2006 TO 2012 (2017), https://supremecourt.nebraska.gov/sites/default/files/
recidivism-rates-adults-jan-2017.pdf [https://perma.cc/HU6K-GKG2]; JOSEPH R. TATAR II &
MEGAN JONES, RECIDIVISM AFTER RELEASE FROM PRISON, WIS. DEP’T OF CORRECTIONS.
(2016).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 779

to inform probation and parole policy.”32 The federal Second Chance Act,
which authorizes millions of dollars in grants to state governments and non-
profit service agencies assisting former prisoners, specifies that recipients
must develop a “comprehensive strategic re-entry plan” that includes a
measurable performance outcome of “reduc[ing] the rate of
recidivism . . . by 50 percent over a 5-year period for offenders released from
prison, jail, or a juvenile facility.”33 (In likely recognition of the
implausibility of achieving such dramatic reductions, grant applications
encourage applicants “to establish reasonable recidivism reduction goals for
their state based on current conditions and drivers.”34) Interest in measuring
recidivism rates extends to juvenile justice, where programmatic success is
also measured by reference to reductions in recidivism rates.35
Private charitable foundations have similarly embraced the idea of
recidivism as a reliable indicator of the success of criminal justice
interventions. The Bob Barker Company Foundation requires grant
applicants’ work to “result in reducing recidivism.”36 The D.C.-based Morris
and Gwendolyn Cafritz Foundation declares that its grants “ai[m] to reduce
the recidivism rate of incarcerated youth and adults.”37 But private interest
in reducing recidivism rates is not only philanthropic. Increasingly, private
funding in the form of social impact bonds is being used to fund programs
designed to reduce recidivism.38 In this model, private investors agree to

32 NAT’L INST. JUST., WHY RECIDIVISM IS A CORE CRIMINAL JUSTICE CONCERN,

https://www.nij.gov/topics/corrections/recidivism/pages/core-concern.aspx [https://perma.cc/
HX9J-2KD7] (last visited May 5, 2019).
33 SECOND CHANCE ACT OF 2007: COMMUNITY SAFETY THROUGH RECIDIVISM
PREVENTION, PUB. L. No. 110-199, § 101, 122 Stat. 664 (2008).
34 See BUREAU OF J UST. ASSISTANCE, Second Chance Act Statewide Adult Recidivism
Reduction Strategic Planning Program FY 2017 Competitive Grant Announcement, 5–6,
available at https://www.bja.gov/funding/SRR17.pdf [https://perma.cc/LQR7-6JFG].
35 PHIL W. HARRIS ET AL., COUNCIL JUV. CORRECTION ADMIN., DEFINING AND MEASURING
RECIDIVISM 1 (2009), http://cjca.net/attachments/article/55/CJCA-Recidivism-White-Paper.
pdf [https://perma.cc/CA42-GFGG] (“It is uncommon to conduct a program impact evaluation
in juvenile justice without measuring recidivism. Despite challenges posed by definitional
ambiguity and misuse of recidivism data, a program’s recidivism rate is generally regarded as
the most critical indicator of program success to the widest audience.”).
36 Eligibility and Grant Process, Bob Barker Company Foundation, http://www.bobbark
ercompanyfoundation.org/grant-process/ [https://perma.cc/HW9A-PBSU] (last visited May
5, 2019).
37 Apply for a Grant, Morris and Gwendolyn Cafritz Foundation, http://www.cafritzfound
ation.org/apply [https://perma.cc/P4A8-BMLA] (last visited May 5, 2019).
38 Richard A. Bierschbach & Stephanos Bibas, Rationing Criminal Justice, 116 MICH. L.
REV. 187, 223 (2017); Kevin W. Humphries, Note, Not Your Older Brother’s Bonds: The Use
and Regulation of Social-Impact Bonds in the United States, 76 L. & CONTEMP. PROBS. 433,
435 (2013).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
780 KLINGELE [Vol. 109

fund programs that deliver quantifiable concrete public goods such as

reduced rates of recidivism.39 If the programs are effective, as measured by
their ability to meet measurable targets, then the investors are reimbursed,
often at a profit to investors.40
These pay-for-success (PFS) models have been used with some success
in the United Kingdom, and have begun to gain traction in the United States
as well.41 In an attempt to reduce recidivism, Goldman Sachs in 2012
provided $7.2 million to create a cognitive therapy program for young people
incarcerated on Riker’s Island in New York City.42 If the program succeeded
in its goal of reducing recidivism by 10%, the city would reimburse the cost.43
In that case, a follow-up study conducted by the Vera Institute of Justice in
2015 found that the program did not measurably decrease recidivism at all,
and the program was therefore terminated.44 In a different experiment with
social impact bonding, New York State has partnered with private
philanthropists at the Center for Employment Opportunities to provide job
training and temporary employment for those leaving prison.45 If that
program meets its five-year goal of achieving an 8% reduction in recidivism
rates and a 5% improvement in employment rates, the government will
reimburse the program cost.46 If the program exceeds its benchmark goals,
investors will earn a profit.47 In both these examples, the pressure to reduce
recidivism is extremely high, since large sums of money turn on the ability
of the interventions to achieve their stated goals.48

39 Jo Max Liang et al., An Overview of Social Impact Bonds, 13 J. INT’L BUS. & L. 267,
268 (2014).
40 Id.
41 Id. at 268, 278, App. A.
42 JIM PARSONS ET AL., VERA INST. JUST., IMPACT EVALUATION OF THE ADOLESCENT
BEHAVIORAL LEARNING EXPERIENCE (ABLE) PROGRAM 23 (2016); Eduardo Porter, Wall St.
Money Meets Social Policy at Rikers Island, N.Y. TIMES (Jul. 28, 2015), https://www.nytimes.
com/2015/07/29/business/economy/wall-st-money-meets-social-policy-at-rikers-island.html
[https://perma.cc/4W3V-7ZGE].
43 PARSONS ET AL., supra note 42, at 6.
44 Id. at 23.
45 Alana Semuels, A New Investment Opportunity: Helping Ex-Convicts, THE ATLANTIC,
(Dec. 21, 2015), https://www.theatlantic.com/business/archive/2015/12/reducing-recidivism/
421323/ [https://perma.cc/UY4P-L67U].
46 Id.; New York State Increasing Employment and Improving Public Safety, NONPROFIT
FINANCE FUND (Dec. 7, 2017), http://www.payforsuccess.org/project/new-york-state-increas
ing-employment-and-improving-public-safety [https://perma.cc/SE4S-PXN6] (last visited
May 5, 2019).
47 Semuels, supra note 45.
48 See Jennifer Miller Oertel et. al., Proving That They Are Doing Good: What Attorneys
and Other Advisers Need to Know About Program Assessment, 59 WAYNE L. REV. 693 (2013)

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 781

Recidivism reduction has attracted investors in more ways than one.

Entirely apart from social impact bonds, private companies have begun to fill
the growing demand for tools that can predict the risk that criminal
defendants, probationers, prisoners, and parolees will commit future
crimes.49 Peddling proprietary algorithms that purport to predict risk and
identify individual treatment needs, private companies like Equivant
(formerly Northpointe) are finding a profitable market as criminal justice
stakeholders seek data to guide sentencing, supervision, and release
decisions.50 Studies of these risk-and-needs tools vary in appraisals of their
accuracy. Some studies claim that the tools outperform human judgment in
their predictions of future offending,51 and others claim that the tools increase
pre-existing racial and class disparities in punishment without accurately
sorting minor offenders from serious ones.52 However good or bad these
tools may prove to be, their popularity unquestionably entrenches the fear of
avoiding recidivism in any form.53

(“Because often millions of dollars will hinge upon whether or not social outcome goals have
been met, it is imperative that all parties understand social outcome measurement.”).
49 Michael Tonry, Legal and Ethical Issues in the Prediction of Recidivism, 26 FED.
SENT’G REP. 167, 167 (2014) (noting “[u]nprecedented private sector involvement . . . in
designing and marketing [risk prediction] instruments and providing services to
government”); NATHAN JAMES, RISK AND NEEDS ASSESSMENT IN THE CRIMINAL JUSTICE
SYSTEM, CONGRESS. RESEARCH SERV. 4 (2015), https://fas.org/sgp/crs/misc/R44087.pdf
[https://perma.cc/EE9X-MVCM] (discussing the various points in the criminal justice system
at which risk assessments can be utilized).
50 The most widely used proprietary tool is the Correctional Offender Management
Profiling for Alternative Sanctions, (COMPAS), sold by Equivant, and used by several states,
including Florida, Michigan, and Wisconsin. See ELECTRONIC PRIVACY INFORMATION
CENTER, ALGORITHMS IN THE CRIMINAL JUSTICE SYSTEM, https://epic.org/algorithmic-
transparency/crim-justice/ [https://perma.cc/CW89-CQXU] (last visited May 5, 2019).
51 See, e.g., Tim Brennan et al., Evaluating the Predictive Validity of the COMPAS Risk
and Needs Assessment System, 36 CRIM. JUST. & BEHAV. 21 (2009) (reporting favorably on
results of a company-produced study of COMPAS accuracy).
52 See, e.g., Julia Dressel & Hany Farid, The Accuracy, Fairness, and Limits of Predicting
Recidivism, 4 SCI. ADVANCES 1 (2018), http://advances.sciencemag.org/content/advances/4/1/
eaao5580.full.pdf [https://perma.cc/JE93-MPNQ] (asserting that both COMPAS was no better
than a randomly selected group of non-specialists at predicting future risk of recidivism); see
also Julia Angwin et al., Machine Bias, PRO PUBLICA (May 23, 2016), https://www.propublic
a.org/article/machine-bias-risk-assessments-in-criminal-sentencing [https://perma.cc/V7MH-
W4CP] (purporting to find evidence of inaccuracies and racial bias in COMPAS results). But
see Anthony W. Flores et al., False Positives, False Negatives, and False Analyses: A
Rejoinder to “Machine Bias: There’s Software Used Across the Country to Predict Future
Criminals. And It’s Biased Against Blacks,” 80 FED. PROBATION 38, 44–45 (2016) (finding no
evidence of racial bias in COMPAS results).
53 Alyssa M. Carlson, Note, The Need for Transparency in the Age of Predictive
Sentencing Algorithms, 103 IOWA L. REV. 303, 315–16 (2017) (discussing proprietary
algorithm used in one such recidivism prediction tool, COMPAS).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
782 KLINGELE [Vol. 109

The pressure to emphasize recidivism data has come from internal

stakeholders, too. The Conference of Chief Justices has promoted greater
use of risk assessment tools asserting that “the best research evidence has
shown that use of validated offender risk and need assessment tools is critical
in reducing recidivism.”54 Taking an equally optimistic view, individual
judges have also spoken in favor of greater attention to recidivism risks and
rates. Judge Roger Warren, for example, has advocated nationally for greater
consideration of recidivism risks at sentencing and for using evidence of
recidivism to evaluate the effectiveness of criminal justice programs.55 His
position turns in large part on an assessment that, unlike in decades past,
social scientists now know enough about drivers of recidivism to reduce risk
of future offense:
Today . . . we know—based on meticulous meta-analyses of rigorously conducted
scientific research—that unlike incarceration the right kinds of rehabilitation and
treatment programs carefully targeted at specific crime-related risk factors among
medium- to high-risk offenders can reduce offender recidivism by conservative
estimates of 10 to 20 percent.56
Judge Michael Marcus has similarly advocated for the broader use of
recidivism risk assessment tools and greater attention to recidivism rates,
optimistically asserting that properly-designed correctional programs “can
hope to produce roughly a 30% reduction in recidivism among many
common offenders.”57 He argues that because “the public is concerned, most
of all, with how successfully [the criminal justice system] prevent[s]
recidivism,” attention to recidivism risks and rates is an essential component
of modern sentencing.58
At times, academics have also embraced recidivism as a metric for
success in the criminal justice system, albeit with some caveats. Francis
Cullen, Cheryl Jonson, and Daniel Mears recently suggested “recidivism
reduction should be defined as the core goal of corrections, including
community-based agencies. Wardens, prison staff, probation and parole

54 State v. Loomis, 881 N.W.2d 749, 752 n.1 (Wis. 2016) (citing CONFERENCE OF CHIEF
JUSTICES, CONFERENCE OF STATE COURT ADMINISTRATORS, NATIONAL CENTER FOR STATE
COURTS, RESOLUTION 12: IN SUPPORT OF SENTENCING PRACTICES THAT PROMOTE PUBLIC
SAFETY AND REDUCE RECIDIVISM (2007), http://ccj.ncsc.org/~/media/Microsites/Files/CCJ/
Resolutions/08012007-Support-Sentencing-Public-Safety-Reduce-Recidivism.ashx [https://
perma.cc/UC99-WPAD] (last visited May 5, 2019)).
55 Roger K. Warren, The Most Promising Way Forward: Incorporating Evidence-Based
Practice into State Sentencing and Corrections Policies, 20 FED. SENT’G REP. 322, 324 (2008).
56 Id. at 323.
57 Michael Marcus, MPC—the Root of the Problem: Just Deserts and Risk Assessment, 61
FLA. L. REV. 751, 770 (2009).
58 Id. at 765.

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 783

chiefs, and officers should all be judged on whether offenders who pass
through their organizations return to crime.”59 They argue that correctional
officials should be held accountable for reducing recidivism in the same way
“[s]ales managers each year are given goals, and profit margins are tracked
carefully.”60 Why should corrections shoulder this responsibility? Because,
they assert, “[a]t some point, community corrections is going to have to
demonstrate its efficacy in reducing recidivism, or lose legitimacy.”61

II. RECIDIVISM AS A SUCCESS METRIC

Given the wide variety of ways in which recidivism measures and
predictions are used to assess the success of criminal justice programs and
the people upon whom they intervene, it is important to understand exactly
what recidivism means. Despite the criminal justice system’s heavy reliance
on recidivism rates, there is surprising variation in how recidivism is defined
and measured.62 Nevertheless, data purporting to track or predict recidivism
is put to a myriad of uses, from deciding which treatment programs to fund
to deciding what sentences to impose on individual defendants.63 The
following subsections discuss in greater detail how recidivism is defined and
measured in today’s criminal justice system.

A. DEFINING RECIDIVISM
Recidivism data attempt to quantify whether a person who has
committed a crime in the past has gone on to commit another crime in the
future.64 While the concept is simple, it is notoriously difficult to measure.65

59 Id.
60 Francis T. Cullen et al., Reinventing Community Corrections, 46 CRIME & JUST. 27, 50
(2017).
61 Id. at 49. Although Cullen, Jonson, and Mears embrace the idea of gathering and
analyzing recidivism data, they also acknowledge that there are risks associated with
emphasizing recidivism reduction, including the very real risk of data gaming. Id. at 51
(“When accountability is heightened, strong incentives emerge to game the systems.”). For a
full discussion on this point, see infra Section IV.A.
62 See NATHAN JAMES, CONG. R ESEARCH SERV., RL34287, OFFENDER REENTRY:
CORRECTIONAL STATISTICS, REINTEGRATION INTO THE COMMUNITY, AND RECIDIVISM 5–6
(2015); URBAN INSTITUTE, MEASURING RECIDIVISM AT THE LOCAL LEVEL: A QUICK GUIDE 1–
3 (n.d.), available at https://www.urban.org/sites/default/files/recidivism-measures_final-for-
website.pdf [https://perma.cc/36KA-R37C].
63 See Weisberg, supra note 30, at 790–94.
64 Petersilia, supra note 12, at 382.
65 Weisberg, supra note 30, at 794–95 (“[T]here often are simply too many variables to
account for, too many actors with too much and too many kinds of discretion, for us to
confidently credit a deliberate program for a measured outcome.”).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
784 KLINGELE [Vol. 109

The first challenge in quantifying recidivism is deciding what

constitutes proof of a subsequent crime. Most people do not boldly proclaim
their criminal exploits, and instead seek to hide—or at least downplay—their
involvement in illegal activities. This basic fact makes formal detection (and
therefore measurement) of criminal behavior difficult.
The most effective way to measure behavior is through longitudinal
studies.66 In such settings, researchers follow subjects over long periods of
time—often decades—periodically surveying, interviewing, and gathering
third party data about subjects’ behavior.67 Researchers conducting such
studies provide participants with guarantees of confidentiality and other
protections designed to reduce the risk that honest confessions about illegal
behavior will result in consequences of any kind.68 Longitudinal studies are
expensive and time consuming, however. For these reasons, they are
typically conducted by academic researchers, and not by criminal justice
program analysts.
Analysts who work within the criminal justice system are limited in
ways academics are not. They are frequently under pressure to produce
assessment results in short periods of time and to satisfy funders,
policymakers, and the public of the effectiveness of the programs they are
assessing.69 Moreover, as state agents, they are not always positioned to offer

66 David P. Farrington, Longitudinal and Experimental Research in Criminology, 42

CRIME & JUST. 453, 454 (2013).
67 Id. at 454–55; see also Loeb & Sampson, supra note 16, at 14 (analyzing evidence from
the Glueck longitudinal study of delinquent youth conducted from 1939 onward); Mark Warr,
Life Course Transitions and Desistance from Crime, 36 CRIMINOLOGY 183, 185–88 (1998)
(discussing several longitudinal studies and analyzing data from the National Youth Survey,
which collected data in 1976 and in nine subsequent waves thereafter, following participants
from childhood through young adulthood).
68 Institutional Review Boards (IRBs) are administrative bodies that are established to
protect the rights and welfare of human research subjects. See Christine Tartaro & Marissa P.
Levy, Criminal Justice Professionals’ Knowledge of Institutional Review Boards (IRBs) and
Compliance with IRB Protocol, 25 J. CRIM. JUST. EDUC. 321 (2014). IRBs ensure that research
on human subjects conducted by academic institutions and their affiliates complies with
applicable federal regulations, as well as local statutes and policies. A common protection
study participants receive is that all information collected by researchers shall remain private
and confidential. See, e.g., 42 U.S.C. § 241(d) (2012) (authorizing issuance of Certificates of
Confidentiality, which allow researchers to refuse attempts to compel disclosure of participant
information in legal proceedings); 45 CFR § 46.111(a)(7) (Health and Human Services policy
requiring all federally funded proposed research to demonstrate adequate safeguards to
“protect the privacy of subjects and to maintain the confidentiality of data”).
69 See Megan C. Kurlycheck et al., Long-Term Crime Desistance and Recidivism
Patterns–Evidence from the Essex County Convicted Felon Study, 50 CRIMINOLOGY 71, 79
(2012) (noting that in the majority of recidivism studies “follow-up periods are no more than
a few years long”).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 785

the same confidentiality protections researchers use to glean more accurate

information about illegal behavior.70 Not that anyone within the criminal
justice system regularly asks convicted people to provide such incriminating
information: the limited resources of most agencies and actors within the
system assure that the only recidivism data typically collected by state
agencies draws from formal court or law enforcement records, rather than
individual self-reports.71 The result is that while the criminal justice system
purports to measure recidivism, what recidivism data usually measure are
rates of re-capture—outcomes that turn as much on luck and policing patterns
as they do on deviant behavior.72
If recidivism rates are not being calculated based on longitudinal
studies, then from what records are they produced? Four types of documented

70 For example, disclosures of past misconduct made by participants in sex offender

programs operated by correctional facilities have sometimes served as the basis for later court
proceedings against these program participants. Corey Rayburn Yung, Sex Offender
Exceptionalism and Preventive Detention, 101 J. CRIM. L. & CRIMINOLOGY 969, 984 (2011).
This stands in contrast to the confidentiality protections that academic researchers must afford
research subjects under Institutional Research Board (IRB) protocols. See generally Leslie E.
Wolf et al., Certificates of Confidentiality: Protecting Human Subject Research Data in Law
and Practice, 43 J. LAW & MED. ETHICS 594 (2015). The tension between the role of
correctional agencies as executive branch law enforcers and their role as publicly-accountable
agencies can create ethical tensions in how they conduct research—and, consequently, in how
forthcoming their subjects are likely to be.
71 See Farrington, supra note 66 at 463 (reporting on results of a longitudinal study of
youthful offenders that “found that the prevalence, frequency, and duration of criminal careers
were all greater for reported offending than for arrests and that the escalation from minor to
more serious crimes was greater for reported offending than for arrests”); James L. Johnson,
Comparison of Recidivism Studies: AOUSC, USSC, and BJS, 81 FED. PROBATION 52, 54
(2017) (“[A]ny definition will underestimate the ‘true’ recidivism rate, because rates are based
on official criminal record data that only show crimes for which people have been arrested or
convicted.”).
72 See Jaclyn Hovda, Note, The Efficacy of Idaho’s Domestic Violence Courts: An
Opportunity for the Court System to Effect Social Change, 48 IDAHO L. REV. 587, 611 (2012)
(“[W]hile a reduction in recidivism seems like an obvious metric by which the efficacy of
D[omestic] V[iolence] courts could be measured, these rates can be deceiving because of
chronic underreporting by victims.”). Because re-capture data are based solely on behavior
detected by the criminal justice system, such data will necessarily be over-inclusive of the
“failures” of heavily-policed communities and underinclusive of the failures of those whose
behavior is not as closely surveilled. Given policing patterns in the United States, that suggests
that recidivism rates may fail to reflect the behavior of under-policed members of society, who
will be predominantly affluent and white. See, e.g., Decio Coviello & Nicola Persico, An
Economic Analysis of Black-White Disparities in the New York Police Department’s Stop-
and-Frisk Program, 44 J. LEGAL STUD. 315, 317 (2015) (discussing racial disparities in New
York City street stops); Charles R. Epp et al., Beyond Profiling: The Institutional Sources of
Racial Disparities in Policing, 77 PUB. ADMIN. REV. 168 (2017) (discussing racial disparities
in vehicular stops).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
786 KLINGELE [Vol. 109

events can be used as proxies for recidivism: arrest, charge, conviction, and
revocation from community supervision.73 Some studies use only one of
these measures, while others examine a combination of two or more. 74
Importantly, there is no national standard governing the choice of triggering
events, and typically no standardization within a single state or county.75
Comparing recidivism rates across programs or jurisdictions is therefore an
often futile task, since each study provides a different measure of re-
offense.76 This lack of uniformity is particularly problematic given the
different flaws inherent in each possible triggering event.
In some ways, arrest and charge data may get closer to capturing true
recidivism data than conviction records can because they track events in
which, at a minimum, the suspect behaved in a way that gave police or
prosecutors probable cause to believe he had re-offended.77 These records
are less likely than conviction records to be underinclusive of criminal
behavior, although they are less comprehensive than self-reported data
gathered by third-party researchers because they rely on official detection of
deviant behavior, which invariably will miss many instances of illegal
conduct.
Arrest and charge data are not necessarily better than conviction data,
since they risk being overinclusive. Police may arrest the wrong suspect or
may arrest for behavior that turns out not to be criminal at all once a full

73 Petersilia, supra note 12, at 383.

74 See generally Johnson, supra note 71, at 52 (2017) (comparing recidivism studies that
use different measures); see also Jessica M. Eaglin, Constructing Recidivism, 67 EMORY L.J.
59, 76–77 (2017) (reporting on the markers of recidivism used by a variety of recidivism risk
prediction tools).
75 See, e.g. Letter from Kamala Harris, Cal. Atty. Gen., to Public Safety Partners 1–2 (Oct.
16, 2014), https://oag.ca.gov/system/files/attachments/press_releases/Recidivism%20Definiti
on%20Letter%2C%20AG%20Harris.pdf [https://perma.cc/55U9-Z6CG] (observing that
“California lacks any uniform or standard way to measure the rate of individuals who re-
commit crimes” and proposing a statewide definition of recidivism as “[a]n arrest resulting in
a charge within three years of an individual’s release from incarceration or placement on
supervision for a previous criminal conviction”); see also Phil W. Harris et al., Defining and
Measuring Recidivism 1 COUNCIL JUV. CORRECTIONAL ADMINS. (2009) (observing that “using
the average of state juvenile recidivism rates for a small number of states, the national juvenile
rate could be anywhere between 25% and 55% depending on what measure of recidivism is
used to comprise the measure”).
76 Petersilia, supra note 12, at 382 (observing that “recidivism data in one study are seldom
comparable to the data in another”).
77 See Gerstein v. Pugh, 420 U.S. 103, 122 (1975) (probable cause required to bring
criminal charges); Draper v. United States, 358 U.S. 307, 314 (1959) (probable cause required
for lawful arrest).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 787

investigation has been completed.78 Charges may be brought against the

wrong defendant or may not align with the actual behavior in which the
defendant engaged.79 The fact that many of these cases do not proceed to
conviction gives rise to doubt about whether a crime occurred at all or
whether an error was made by system actors themselves.80
Conviction data have the advantage of being the least overinclusive
because they require that a defendant either admit to the charged conduct or
be found guilty beyond a reasonable doubt at trial. Nonetheless, conviction
data have the distinct disadvantage of being underinclusive in many cases.
Criminal cases may be disposed of pre-conviction for many reasons that have
nothing to do with innocence. Police error, prosecutorial resource
constraints, policy choices, and difficulty developing evidence all can lead to
dismissal of charges (or later acquittal at trial), even in cases when a
defendant has actually committed a crime.81 As a result, while conviction
data generate the fewest false positives among potential events suggestive of
recidivism, they include an unknown and potentially significant rate of false
negatives.82
In addition to conviction, arrest, or charging data, many studies of
recidivism also count revocation from probation or any form of post-release
supervision as a triggering event.83 While incarceration is an important event
to capture in recidivism data, individuals who are revoked from community
supervision have not always committed a new crime.84 New criminal
behavior can certainly trigger revocation (and often will) but so can so-called

78 Alfred Blumstein & Richard C. Larson, Problems in Modeling and Measuring

Recidivism, 8 J. RESEARCH CRIME & DELINQUENCY 124, 125 (1971) (“Whatever the choice [of
recidivism metric], there are two types of errors: the Type I error of commission involves the
erroneous counting as recidivists those who are improperly charged with recidivism . . . . The
Type I error includes erroneous arrests, convictions, and sentences.”).
79 Surell Brady, Arrests Without Prosecution and the Fourth Amendment, 59 MD. L. REV.
1, 41–42 (2000) (“Among the most commonly cited factors [for non-prosecution] are victim
and/or witness reluctance, unavailability of prosecution or defense witnesses, and writs of
extradition filed by other states. State prosecutors also report dismissals by courts due, in part,
to ‘search or seizure problems,’ and legal issues relating to self-incrimination, speedy trial
time restrictions, and the right to counsel.”).
80 Id.
81 Tayler Ruggero et al., Measuring Recidivism: Definitions, Errors, and Data Sources 3–
4, (Ctr. for Pub. Safety Initiatives, Working Paper No. 2015-03) (“The problem with using
only reconviction rates is that not everyone is convicted for all crimes they have committed;
authorities may not have become aware of the crime, or there may not have been enough
evidence to prosecute.”).
82 Id.
83 Fiona Doherty, Obey All Laws and Be Good: Probation and the Meaning of Recidivism,
104 GEO. L. J. 291, 295 (2016).
84 Id.

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
788 KLINGELE [Vol. 109

non-criminal violations, including absconding (that is, running away or

persistently avoiding required supervision meetings), failing to attend
mandated treatment programs, or failing to abide by restraint conditions, such
as curfews, no-contact orders, and abstention from alcohol or other drugs.85
Settling on what evidence proves recidivism is just the beginning of the
ambiguity surrounding what it means to “recidivate.” The next challenge is
defining a follow-up period that meaningfully captures the effect of a given
program on recidivism rates. Due to system resource constraints and a desire
for rapid feedback on the effect of various interventions, most studies on
recidivism follow convicted individuals for only two or three years from
conviction or from the completion of a sentence.86 A few employ a five year
follow-up window.87 As a result, data currently gathered on recidivism
provides a time-limited snapshot into the most recent behavior of convicted
persons following sentencing or imprisonment,88 not a full picture of how
convicted individuals go on to live their lives.
To determine whether a person subject to a criminal justice intervention
goes on to re-offend, perfect data would capture that person’s behavior for
the remainder of his life—or at least until enough time has passed that he no
longer remains at an elevated risk of committing future offenses. Although
individuals with prior records—like those without them—can commit

85 See generally Cecelia Klingele, Rethinking the Use of Community Supervision, 103 J.
CRIM. L. & CRIMINOLOGY 1015, 1032–35 (2013).
86 See Kurlycheck et al., supra note 69, at 73 (noting that in the majority of recidivism
studies “follow-up periods are no more than a few years long.”); see also David P. Farrington,
Age and Crime, 7 CRIME & JUST. 189, 222 (1986) (“Unfortunately, it is difficult to draw
conclusions about termination [of offending] from information about recidivism because those
who do not recidivate within two years include true desisters, undetected offenders, and those
who will persist later.”).
87 Shawn D. Bushway et al., Connecting Desistance and Recidivism: Measuring Changes
in Criminality Over the Lifespan, in AFTER CRIME AND PUNISHMENT: PATHWAYS TO OFFENDER
REINTEGRATION 85, 86 (Shadd Maruna & Russ Immarigeon eds., 2004); Laura M. Baber &
Mark Motivans, Extending Our Knowledge About Recidivism of Persons on Federal
Supervision, 77 FED. PROB. 23 (2013) (reporting on a five-year recidivism study of people
released from federal prison).
88 Recidivism data vary not only in definitions of recidivism and length of follow-up
windows, but also in the start and end times of follow-up window. In order to assess whether
people’s behavior has been altered by criminal justice interventions, it is necessary to study
them when they are “at hazard” of detection and punishment. See Kurlycheck et al., supra
note 69, at 72–79 (discussing methods of conducting long-term studies). While people commit
crime in jails and prisons, as well as in the community, fewer opportunities exist in such
closely-monitored environments. Consequently, follow-up studies for prisoners commence
typically upon release from custody and not immediately following sentencing. See, e.g.,
Mariel Alper et al., 2018 Update on Prisoner Recidivism: A 9-Year Follow-up Period (2005-
2014) BUREAU OF JUST. STAT. (2018), https://www.bjs.gov/content/pub/pdf/18upr9yfup0514.
pdf [https://perma.cc/Y5X9-8D7G].

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 789

criminal offenses spontaneously at any point in life, research suggests that as

time without re-offense lengthens, the odds of recidivism diminish
significantly. Al Blumstein and Kiminori Nakamura’s work on redemption
has examined how long a person convicted of a crime must go without re-
offending before he or she obtains a predicted risk of future arrest no higher
than that of a similarly-aged person in the general population.89 While the
exact window varies based on the age of the convicted person and the type
of crimes the person has committed, their research suggests that the average
“redemption time” was between ten and thirteen years after conviction.90
Blumstein and Nakamura’s findings suggest that a ten to fifteen year
follow-up period would provide a thorough picture of the effect of criminal
justice programs on long-term recidivism; however, a follow-up window of
that length is beyond the capacity of most criminal justice agencies.91 The
good news is that Blumstein and Nakamura’s work confirms what other
studies have repeatedly found: the vast majority of those who recidivate do
so within the first few years following release from custody. 92 As time
without re-offense lengthens, fewer people go on to recidivate at later points
in time.93 For this reason, shorter follow-up windows will yield useful, albeit
incomplete, information that can meaningfully capture the majority of those
who recidivate following any given intervention. That fact, coupled with the
observation that the effects of any criminal justice intervention, are likely to
fade over time, means that data about former offenders’ behavior in the years
that immediately follow sanction are not only the easiest data to obtain but
are also likely to be the most relevant for determining whether system
interventions are having a widespread effect on recidivism rates.

89 Alfred Blumstein & Kiminori Nakamura, Redemption in the Presence of Widespread

Criminal Background Checks, 47 CRIMINOLOGY 327 (2009).
90 Alfred Blumstein & Kiminori Nakamura, Paying a Price, Long After the Crime, N.Y.
TIMES, Jan. 9, 2012, at A23.
91 A recent, rare nine-year follow-up study by Mariel Alper, Matthew R. Durose, and
Joshua Markman of the Bureau of Justice Statistics provides a refreshing counter-example to
the overall tendency of state and government agencies to focus only on the time immediately
following release. See Alper et al., supra note 88, at 4. Importantly, the BJS study concludes
that recidivism rates, while reduced over time, persisted at higher levels than a three-year
snapshot would suggest. Id.
92 Blumstein & Nakamura, supra note 89, at 331 (“Studies on recidivism consistently
demonstrate that those who have offended in the past will have the highest probability of
reoffending within several years, and the probability will decline steadily afterward.”);
MATTHEW R. DUROSE ET AL., BUREAU OF JUST. STAT., RECIDIVISM PATTERNS OF PRISONERS
RELEASED IN 30 STATES IN 2005: PATTERNS FROM 2005 TO 2010 1 (2014) (“More than a third
(36.8%) of all prisoners who were arrested within 5 years of release were arrested within the
first 6 months after release, with more than half (56.7%) arrested by the end of the first year.”).
93 Blumstein & Nakamura, supra note 89, at 331.

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
790 KLINGELE [Vol. 109

B. REPORTING RECIDIVISM
Despite wide variations in how recidivism is defined and measured, it
remains a key metric used by criminal justice stakeholders to gauge the
effectiveness of criminal justice interventions. The following section
examines how recidivism data are reported and what those reports mean.
Typically, recidivism studies compare the performance of a defined
cohort to that of a baseline population over an equal period of time. 94 For
example, a study of participants in a prison education program might report
how program graduates fared during the three years following their release
from custody, as compared to a similar cohort of released prisoners who did
not participate in the educational program.95 Once variables other than the
intervention are controlled for, the difference between the performance of the
two groups is then attributed to the intervention itself.96
The results of recidivism studies are often used to determine whether
specific criminal justice interventions “reduce recidivism.” What that means
is not necessarily self-evident. When a program is found to reduce recidivism
by, 10%, that does not mean that each individual who completes the program
will, on average, go on to commit 10% less crime than the person otherwise
would have committed. Instead, it means that an aggregate group of people
who complete the program will have 10% fewer members of their cohort
recidivate than a comparable group of people who did not complete the
program. This is an important distinction for several reasons.
First, it reminds us that no retrospective study, or even any predictive
tool, can tell us what any individual person will do in the future. It is
impossible to know from a person’s characteristics and past behavior whether
the person’s most recent conviction will be the last or whether the person will
go on to commit serious violations.97 The very best risk prediction
instruments currently available can tell us which factors, on average, drive
recidivism, and what people with histories and characteristics similar to a
particular offender are statistically likely to do in the future.98 However,
given the variation of human behavior,99 no instrument can reliably predict

94 See, e.g., Ryang Hui Kim & David Clark, The Effect of Prison-Based College Education
Programs on Recidivism: Propensity Score Matching Approach, 41 J. CRIM. JUST. 196, 196–
97 (2013).
95 Id.
96 Id. at 198–203.
97 See Blumstein & Nakamura, supra note 89, at 332–33.
98 See Sonja B. Starr, Evidence-Based Sentencing and the Scientific Rationalization of
Discrimination, 66 STAN. L. REV. 803, 806 (2014).
99 There has been a resurgence in determinism among some neuroscientists and
punishment theorists. See, e.g., Luis E. Chiesa, Punishing Without Free Will, 2011 UTAH L.

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 791

whether any particular person will go on to re-offend or what specific form

any future re-offense will take.100
This observation has particular salience for the proper use of
increasingly popular recidivism risk prediction instruments, presently used

REV. 1403, 1406 (2011) (“[T]here are good moral reasons to conclude that the scientific
plausibility of determinism ought to lead us to abandon the notion of free will.”); Joshua
Greene & Jonathan Cohen, For the Law, Neuroscience Changes Nothing and Everything, 359
PHIL. TRANSACTIONS ROYAL SOC’Y LONDON B: BIOLOGICAL SCI. 1775, 1781 (2004) (arguing
that many of the traditional focuses of criminal law “will lose their grip in an age when the
mechanical nature of human decision-making is fully appreciated”). Nevertheless, the
criminal justice system (and the public at large) continues to embrace the traditional view that
variations in human behavior are best explained by free will and human agency—a factor life
course criminologists have found to be a meaningful component of desistance from crime.
See, e.g., Ros Burnett, To Reoffend or Not To Reoffend? The Ambivalence of Convicted
Property Offenders, in AFTER CRIME AND PUNISHMENT: PATHWAYS TO OFFENDER
REINTEGRATION 174 (Shadd Maruna & Russ Immarigeon eds., 2004) (stating research shows
“an important characteristic of desistance narratives [is] the theme of being in control . . . an
ex-offender may need to experience some level of personal success in the straight world before
they realise that they do not need to offend to regain a sense of personal agency”) (internal
citations omitted); Laub & Sampson, supra, note 16, at 55 (“[G]iven the role of human agency
in the desistance process, we need to find a way to measure individual motivation, free will,
and ultimately the decision to initiate and embrace the process of change.”).
100 See Starr, supra note 98, at 806 (explaining that instruments designed to predict
recidivism employ “underlying regression models [that] may provide reasonably precise
estimates of the average recidivism rates for the group of offenders sharing the defendant’s
characteristics, but the uncertainty about what an individual offender will do is much greater,
and when it comes to predicting individual behavior, the models offer fairly modest
improvements over chance”). Attempts have been made to develop instruments that can
differentiate between a person’s risk of committing any offense versus his risk of committing
a violent offense. Melissa Hamilton, Adventures in Risk: Predicting Violent and Sexual
Recidivism in Sentencing Law, 47 ARIZ. ST. L.J. 1, 24–25 (2015). In the main, actuarial tools
are poor at predicting future violence, in large part because (happily) violent behavior is
statistically aberrant. Id. at 20 (“violent and sexual recidivism are, contrary to popular belief,
low rate events, except in extraordinarily high risk populations”); see also Henriette Haas &
Maurice Cusson, Comparing Theories’ Performance in Predicting Violence, 38 INT’L J. L. &
PSYCH. 75, 82 (2015) (“Each single theory that we operationalized was quite good at
predicting non-violence (95.0% to 99%), but none of them could recognize more than thirty
percent of the violent cases. The interdisciplinary model however did raise the prediction of
true hits considerably to 36% compared to the theory driven models.”); Daryl G. Kroner et al.,
A Coffee Can, Factor Analysis, and Prediction of Antisocial Behavior: The Structure of
Criminal Risk, 28 INT’L J. L. & PSYCH. 360 (2005) (testing violent prediction abilities of the
Violence Risk Appraisal Guide (VRAG), General Statistical Information on Recidivism
(GSIR), Level of Service Inventory-Revised (LSI-R), and Psychopathy Checklist-Revised
(PCL-R) against several random, non-structured measures of criminal risk (the “coffee can”
measures) and finding no statistical difference in the accuracy of their predictions of
recidivism).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
792 KLINGELE [Vol. 109

at bail hearings,101 sentencing hearings,102 and in correctional decision-

making.103 Despite the de-personalized nature of risk predictions, reports of
predicted future risk can affect defendants’ access to bond, the nature and
length of a sentence, and the level of supervision and number of conditions
to which those serving community-based sentences will be subjected.104
Second, the aggregate nature of the data contained in recidivism studies
means that it is impossible to predict from recidivism rates alone which
specific people will benefit from any particular intervention. For example, if
a program claims to reduce recidivism by 25%, policymakers cannot assume
that all program graduates will leave the program 25% less likely to re-
offend. Rather, if the baseline re-offense rate for those who do not participate
in the program is 50%, then policymakers can predict that 37.5% of those
who do complete the program will go on to recidivate while the remainder
will not.105 From recidivism data alone, however, they cannot predict who
the recidivists will be or what characteristics they may share in common.
Finally, properly interpreting recidivism study outcomes is important
because it helps explain why those within the criminal justice system often
consider interventions successful even when they achieve very small
reductions in recidivism. To a layperson, it seems strange to say that a
program can be “successful” if it reduces a cohort’s (often high) risk of future
offending by a mere 10% or 15%. However, for many criminal justice
programs, a reduction of even 10% in baseline offending is considered a

101 See Crystal S. Yang, Toward an Optimal Bail System, 92 N.Y.U. L. REV. 1399, 1484–
85 (2017) (discussing bail risk prediction instruments, including the Arnold Foundation’s
Public Safety Assessment (PSA)).
102 See Dawinder S. Sidhu, Moneyball Sentencing, 56 B.C. L. REV. 671, 673 (2015)
(“More and more, courts today are adopting the use of risk-assessment tools in sentencing.”).
103 See J. C. Oleson et al., Training to See Risk: Measuring the Accuracy of Clinical and
Actuarial Risk Assessments Among Federal Probation Officers, 75 FED. PROB. 52, 53 (2011);
Amy Robinson-Oost, Evaluation As the Proper Function of the Parole Board: An Analysis of
New York State’s Proposed Safe Parole Act, 16 CUNY L. REV. 129, 136 (2012) (discussing
the use of risk instruments at parole release).
104 An entire body of literature has begun to address the philosophical, legal, and
sociological implications of relying on actuarial risk prediction instruments at stages
throughout the criminal justice process. See, e.g., BERNARD E. HARCOURT, AGAINST
PREDICTION (2007); Eaglin, supra note 74; Melissa Hamilton, Risk-Needs Assessment:
Constitutional and Ethical Challenges, 52 AM. CRIM. L. REV. 231, 232 (2015); Cecelia
Klingele, The Promises and Perils of Evidence Based Corrections, 91 N.D. L. REV. 537
(2015); Starr, supra, note 98.
105 Cf. Francis T. Cullen, Rehabilitation: Beyond Nothing Works, 42 CRIME & JUST. 299,
338 (2013) (describing various ways of reporting and computing recidivism reduction rates,
including as percentages).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 793

respectable result.106 The modest recidivism reduction demonstrated by

many popular correctional interventions is one of the reasons why a sizeable
number of policymakers and correctional administrators view penal
rehabilitation with skepticism.107 With treatment outcomes that reduce
recidivism by such modest percentages, it is not hard to see why the claim
that “nothing works” to reform repeat offenders has been a recurring mantra
since Robert Martinson’s now-infamous article was first published.108
When one looks harder at what is being measured, however, low-impact
rates are not surprising. A multitude of complex factors drive criminal
activity109 and influence behavioral change.110 No single criminal justice
program significantly eliminates criminal behavior in any sizable percentage
of participants.111 Such programs may reduce harm by reducing offense rates
and offense severity among participants and by improving participants’ lives
on any number of different measures.112 But recidivism data provide only

106 See Jeffrey A. Butts & John Roman, Better Research for Better Policies, in JUVENILE
JUSTICE: ADVANCING RESEARCH, POLICY, AND PRACTICE 513 (Francine Sherman & Francine
Jacobs eds., 2011). Larger effects sizes are sometimes claimed, often in the context of
assessments conducted by program administrators themselves. Such studies have “indicate[d]
that demonstration projects, managed by an involved evaluator or program designer, produce
larger treatment effects than the same programs implemented in a ‘real world’ setting.”
Christopher T. Lowenkamp et al., A Quasi-Experimental Evaluation of Thinking for a Change,
36 CRIM. JUST. & BEHAV. 137, 139 (2009).
107 Michelle S. Phelps, Rehabilitation in the Punitive Era: The Gap Between Rhetoric and
Reality in U.S. Prison Programs, 45 L. SOC’S REV. 33, 33–35, 37–40 (2011); see also
Christopher Ingraham, Even Violent Crime Victims Say our Prisons are Making Crime Worse,
WASH. POST: WONKBLOG (Aug. 5, 2016), https://www.washingtonpost.com/news/wonk/wp/
2016/08/05/even-violent-crime-victims-say-our-prisons-are-making-crime-worse/?utm_term
=.1e639db07ff0 [https://perma.cc/RW77-UN5N].
108 Robert Martinson, What Works? Questions and Answers About Prison Reform, 35
PUB. INT. 22 (1974).
109 Studies have suggested that multiple factors predict re-engagement in criminal
behavior, including antisocial attitudes and friends, relationship conflicts, substance abuse,
past criminal behavior, low educational and vocational attainment, marital status, poverty, age
and gender (young men are the most prolific offenders), challenges in family of origin,
including familial criminality and exposure to abuse or neglect, cognitive deficiencies, and
stress. Mandeep K. Dhami et. al., Prisoners’ Positive Illusions of Their Post-Release Success,
30 LAW & HUM. BEHAV. 631, 633 (2006) (citing Paul Gendreau et al., A Meta-analysis of the
Predictors of Adult Offender Recidivism: What Works!, 34 CRIMINOLOGY 575 (1996)).
110 See infra Section III.A.
111 Weisberg, supra note 30, at 795.
112 Programs can succeed in assisting people even when they fail to reduce or eliminate
criminal behavior. Helping participants conquer addictions, obtain an education, treat unmet
mental health needs, and improve job and parenting skills are all programmatic results that
may be achieved with or without a commensurate reduction in recidivism. It is important to
notice that neither recidivism rates nor the “markers of desistance” I advocate measuring in

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
794 KLINGELE [Vol. 109

one measure of success: whether people who complete a program go on to

commit any offense within the follow-up window. Recidivism data, as
typically reported, omit a host of important contextual information, including
whether the individuals being followed were first-time or chronic
offenders;113 whether the new offenses were minor or serious; or whether
program participants changed their rates of offending after having completed
the program. In short, pure recidivism data do not provide us with any sense
of whether a program or other intervention reduces harm but rather how often
it produces perfection. Not surprisingly, for many offenders, perfect
outcomes are difficult to achieve.

III. BEHAVIORAL CHANGE AND DESISTANCE FROM CRIME

The preceding sections have discussed the challenge of measuring
recidivism in a way that provides policymakers with a clear and accurate
picture of how criminal justice interventions affect subsequent behavior.
These challenges include detecting recidivism over time and discerning
which portions of recidivism data are related to the behavior of former
offenders and which are better attributed to the behavior of system actors.
Theoretically, with adequate resources and effort, many of these
methodological challenges could be overcome—or at least mitigated. But
even if policymakers are able to bring greater standardization to the
measurement of recidivism, larger problems remain.

A. UNDERSTANDING CHANGE
Although policymakers rarely articulate the assumptions that support
using recidivism as the primary measure of success, doing so has value, if
only to focus attention on how implausible many of those assumptions are.
Recidivism, as it is currently framed, is a binary metric: either a person
engages in behavior indicative of new criminal activity within the window of
time the person’s post-intervention behavior is being measured or the person
does not. If the person does, he or she is labeled a recidivist and is considered
a failure. If the person does not commit a new crime, then he or she is
successful. On a larger scale, the same evaluation applies to criminal justice
programs: if the programs reduce aggregate recidivism, they succeed. If they
do not, they have failed.

this Article capture the full range of positive ways in which the criminal justice system may
affect the lives of the people with whom it intervenes.
113 Richard Rosenfeld, Recidivism and its Discontents, 7 CRIMINOLOGY & PUB. POL’Y 311,
314 (2008) (noting “[t]he distinction between correctional veterans and first-timers has all but
disappeared in contemporary recidivism research”).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 795

To probe the plausibility of this approach for measuring success, it is

helpful to review what is known about the dynamics of human behavioral
change, of which the cessation of crime is but one variation. The commission
of crime is, after all, a subset of the much larger category of antisocial
behavior, which is combatted in various contexts not only by police and
correctional personnel but also by parents, teachers, doctors, industrial
engineers, and employers.114 The study and practice of motivating prosocial
behavioral change is a common human enterprise, and consequently, a wide
variety of fields offer examples of alternative ways in which change can be
documented and quantified.
Through education and conditioning, humans are socialized from
infancy to behave in ways that are culturally acceptable within their milieu.115
The process of acculturation is ongoing, and as environments and
expectations change throughout the life course, behaviors that were adaptive
in one environment or acceptable in one life stage must be discarded and new

114 See, e.g., MEME HIENEMAN ET AL., PARENTING WITH POSITIVE BEHAVIOR SUPPORT: A
PRACTICAL GUIDE TO RESOLVING YOUR CHILD’S DIFFICULT BEHAVIOR (2006) (providing
parenting tips for effectively re-directing problematic childhood behavior); MARILYN PINCUS,
MANAGING DIFFICULT PEOPLE: A SURVIVAL GUIDE FOR HANDLING ANY EMPLOYEE (2004)
(offering management tips for redirecting antisocial workplace behavior); Elizabeth E. Stewart
& Chester Fox, Encouraging Patients to Change Unhealthy Behaviors with Motivational
Interviewing, FAM. PRACT. MANAGEMENT 22 (2011), http://www.aafp.org/fpm/2011/0500/
p21.html [https://perma.cc/LPD2-PCJV] (describing cognitive behavioral techniques for
motivating behavioral change in medical patients); Melissa Stormont & Wendy Reinke, The
Importance of Precorrective Strategies and Behavior-Specific Praise and Strategies to
Increase Their Use, 18 BEYOND BEHAV. 26, 26 (2009) (observing that since “many young
children demonstrate behavior that teachers find challenging,” it is incumbent on educators to
“support children’s social development and use effective practices so that children will be less
likely to develop or sustain behavioral problems”); Edward C. Tomlinson & Jerald Greenberg,
Discouraging Employee Theft by Managing Social Norms and Promoting Organizational
Justice, in MANAGING ORGANIZATIONAL DEVIANCE 211 (Roland E. Kidwell & Christopher L.
Martin, eds. 2012) (describing industrial engineering interventions for preventing employee
crime).
115 Evidence suggests that the intellectual and psychological functions that control
learning are present from birth and do not change tremendously across the lifespan. See, e.g.,
Lewis P. Lipsitt, Learning Processes of Human Newborns, 12 MERILL-PALMER QUART.
BEHAV. & DEV. 45, 64 (1966) (presenting evidence of learning in early infancy and pointing
out that “perhaps not so strangely, the parameters and experimental techniques which seem to
have most promise at present [in promoting infant learning] are very similar to those known
for some time to be effective in manipulating animal behavior and adult behavior”).
Recognition of social norms—and deviation from them—also emerges very early in human
development. See Marco F. H. Schmidt & Michael Tomasello, Young Children Enforce Social
Norms, 21 CURRENT DIRECTIONS PSYCH. SCI. 232, 234 (2012) (finding that children as young
as two enforce moral and behavioral norms as part of self-identifying as members of a
distinctive cultural group).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
796 KLINGELE [Vol. 109

habits acquired.116 Each discipline, from public health to education to

criminology, has its own paradigm for explaining how behavioral change
occurs, but all share one feature in common: they see behavior change as a
process and not as an event.
In public health, the popular “transtheoretical model” identifies six
specific stages of change relevant to the abandonment of any maladaptive
behavior.117 These stages, which follow one another in roughly sequential
(though sometimes overlapping) order are pre-contemplation,
contemplation, preparation, action, maintenance, and termination.118 In the
pre-contemplation stage people deny or ignore the behaviors in need of
change, with no intention of altering their current behavior.119 In the
contemplation stage, they wrestle with ambivalence about whether they think
change is possible or desirable.120 In the preparation stage, they begin to
muster the skills and the will to break old habits and build new ones, thus
developing a plan for how they will make needed changes.121 In the action
stage, they begin to change behavior.122 Importantly, the change that occurs
in this stage is not always complete. In the context of smoking cessation,
action may involve reducing the number of daily cigarettes smoked: for
weight loss, it may involve calorie reduction.123 The change must be a
significant alteration from the pre-action behavior, but it need not be
perfected. As the new behaviors become more consistent and habitual, and
the old fade away, people move into the maintenance stage, which may last
from six months to five years.124 Termination is the final stage, one that may
be more ideal than real for most people:
Termination is the stage in which individuals have zero temptation and 100% self-
efficacy. No matter whether they are depressed, anxious, bored, lonely, angry, or
stressed, they are sure they will not return to their old unhealthy habit as a way of
coping.125

116 See generally James E. Thornton, Life-Span Learning: A Developmental Perspective,

57 INT. J. AGING & HUM. DEV. 55 (2003) (discussing social learning across the lifespan).
117 James O. Prochaska & Wayne F. Velicer, The Transtheoretical Model of Health
Behavior Change, 12 AM. J. HEALTH PROMOTION 38, 38 (1997).
118 Id. at 39.
119 Id.
120 Id.
121 Id.
122 Id.
123 Id.
124 Id.
125 Id. In its requirement of perfect abstinence from maladaptive behavior, termination
sounds a lot like recidivism: black and white with no margins of error.

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 797

As people move through each stage, back-sliding—or relapse—is not

uncommon. In fact, “relapse tends to be the rule when action is taken for
most health behavior problems,” though it often precipitates a fresh effort at
positive change.126
The key insight from the transtheoretical model is that “[b]ehavior
change is a process that unfolds over time through a sequence of stages.”127
The transtheoretical model has been tested across a wide range of behaviors,
ranging from addiction to overeating to smoking to delinquency to substance
abuse.128 Regardless of the behaviors studied, the model appears fairly
descriptive of the internal and behavioral processes that underlie change.129
In education, behavior change, in the form of social learning, is typically
explained as the acquisition and mastery of skills over time.130 In order to
move from maladaptive behaviors to more mature, prosocial behaviors,
students must be taught the correct way to be behave. That is, they must see
the desired behavior modeled, must be given opportunities to practice the
skill with corrective and formative feedback, and must have opportunities for
continued practice in multiple settings, both formal and informal.131 The

126 Id.
127 Id. at 41.
128 See, e.g., Ana Andrés et al., The Transtheoretical Model and Obesity: A Bibliometric
Study, 73 SCIENTOMETRICS 289 (2007); Fabiana Andrioni De Biaze Vilela et al., The
Transtheoretical Model and Substance Dependence: Theoretical and Practical Aspects, 31
REVISTA BRASILEIRA DE PSIQUIATRIA, 362 (2009), https://dx.doi.org/10.1590/S1516-4446200
9005000010 [https://perma.cc/PWH9-GLBK]; Russell C. Callaghan et al., Does Stage-of-
Change Predict Dropout in a Culturally Diverse Sample of Adolescents Admitted to Inpatient
Substance-Abuse Treatment? A Test of the Transtheoretical Model, 30 ADDICTIVE BEHAV.
1834 (2005); Christine F. Lerner, The Transtheoretical Model of Change: Self-change in
Adolescent Delinquent Behaviors, Dissertation Abstracts Int’l, 51(12-B, Pt 1), 6111; James O.
Prochaska et al., Stages of Change and Decisional Balance for 12 Problem Behaviors, 13
HEALTH PSYCH. 39 (1994) (summarizing research on change across multiple problem
behaviors including smoking cessation, quitting cocaine, weight control, high-fat diets,
adolescent delinquent behaviors, safer sex, condom use, sunscreen use, radon gas exposure,
exercise acquisition, mammography screening, and physicians’ preventive practices with
smokers).
129 Prochaska, supra note 128, at 39.
130 See, e.g., Pedro Gil-Madrona et al., Acquisition and Transfer of Values and Social
Skills through a Physical Education Program Focused in the Affective Domain, 12
MORTICIDADE 32 (2016) (reporting on effects of social behavior training on student behavior
in the context of physical education courses); David J. Hansen et al., Enhancing the
Effectiveness of Social Skills Interventions with Adolescents, 21 EDUC. & TREATMENT CHILD.
489 (1998); Paulo N. Lopes & Peter Salovey, Toward a Broader Education: Social,
Emotional, and Practical Skills, in BUILDING ACADEMIC SUCCESS ON SOCIAL AND EMOTIONAL
LEARNING: WHAT DOES THE RESEARCH SAY? (Joseph E. Zins et al., eds. 2004).
131 Rosemary Battalio & J. Todd Stephens, Social Skills Training: Teacher Practices and
Perceptions, 14 BEYOND BEHAV. 15, 15 (2015); see also Michael Bullis et al., A Promise

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
798 KLINGELE [Vol. 109

Positive Behavioral Support model, which has gained traction in many

American public schools as a model for advancing behavioral change,132
posits that:
[A]lthough learning and teaching processes are complex and continuous, and some
behavior initially is not learned (e.g., biobehavioral), key messages from this science
are that much of human behavior is learned, comes under the control of environmental
factors, and can be changed. The strength of the science is that problem behaviors
become more understandable, and, as our understanding grows, so does our ability to
teach more socially appropriate and functional behavior.133
In this model, the mechanisms for preventing antisocial behavior and
the mechanisms for responding to it are the same: prevention and early
intervention efforts; a “culture that provides multiple opportunities to display
and receive positive reinforcement for prosocial behavior;” a range of
strategies and responses that can be calibrated to the specific problem
behavior and to the needs of the learner; and a reliance on evidence-based
programs.134
While much of the literature on behavioral education is focused on
teaching children and young adult learners, the principles on which they rest
transfer readily to adult learners. Prosocial education is by no means limited
to the formation of children; adult learners can also manifest disruptive
behaviors that require intervention by educators, ranging from inattention to
threats of violence.135 Notably, the correctional programs that appear to be
most effective in reducing recidivism are those that utilize a cognitive-based
therapy (CBT) approach.136 The “goal of CBT is to help offenders develop

Unfulfilled: Social Skills Training with At-Risk and Antisocial Children and Youth, 9
EXCEPTIONALITY 67 (2001).
132 U.S. DEPT. OF ED., 22D A NNUAL REPORT TO CONGRESS ON THE IMPLEMENTATION OF THE
INDIVIDUALS WITH DISABILITIES EDUCATION ACT, III-10-III-22 (2000).
133 Id.; see also Anthony Biglan, Translating What We Know About the Context of
Antisocial Behavior into a Lower Prevalence of Such Behavior, 28 J. APPLIED BEHAV.
ANALYSIS 479 (1995).
134 Mack D. Burke et al., Preventing School-Based Antisocial Behaviors with School-
Wide Positive Behavioral Support, 1 J. EARLY & INTENSIVE BEHAV. INTERVENTION 65, 66
(2004).
135 See Robert Dobmeier & Joseph Moran, Dealing with Disruptive Behavior of Adult
Learners, 22 NEW HORIZONS ADULT ED. & HUM. RESOURCE DEV. 29, 39–41 (2008); Gerard
M. Schippers et al., Social Skills Training, Prosocial Behavior, and Aggressiveness in Adult
Incarcerated Offenders, 45 INT’L J. OFFENDER THERAPY & COMP. CRIMINOLOGY 244, 246
(2001) (finding positive results in study of Dutch prison program that teaches social skills
through “techniques as modeling, feedback, role playing, imitation, coaching, and social
reinforcement”).
136 See Cullen et al., supra note 60, at 50; Nana A. Landenberger & Mark W. Lipsey, The
Positive Effects of Cognitive–Behavioral Programs for Offenders: A Meta-Analysis of Factors
Associated with Effective Treatment, 1 J. EXPERIMENTAL CRIMINOLOGY 451, 472 (2005).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 799

a new way of thinking by providing them with a chance to model, role-play,

and practice pro-social skills.”137 That model, which employs the same tools
advocated by educators, has repeatedly outperformed other models of
correctional intervention in terms of its effect on recidivism rates.138 That
fact bolsters the key insight educators offer: the mastery of skills occurs over
time and comes only with repeated practice.139
The subfields of sociology known as developmental and life course
criminology embrace paradigms of behavioral change that complement those
used by educators and public health providers in many ways. Through
longitudinal studies of the kind discussed above, criminologists have
examined the ways in which antisocial behavior, both criminal and non-
criminal, manifests at different life stages. 140 In doing so, they have focused
on identifying the factors that correlate not only with commencement of
crime but with its termination. The term used most often to describe this
disentanglement from criminal behavior is desistance.141
While desistance is sometimes used as a synonym for termination of
offending, it is not a singular event.142 Rather, as Sampson and Laub explain
in their seminal work on the subject, desistance “is a social transition that
entails identity transformation, as from a smoker to a nonsmoker, from a
married or coupled person to a divorced or uncoupled person, or from an
offender to a nonoffender.”143 In other words, “just like quitting smoking or
uncoupling, desistance is best viewed as a process rather than a discrete
event.”144
Across an individual’s lifespan, engagement with crime varies
according to a wide number of variables, the most powerful being age.145 In
youth, crime is so statistically prevalent that it is considered sociologically

137 JASLENE LIZAMA ET AL., CTR. FOR PUB. POL’Y, WHAT WORKS? S HORT-TERM IN-
CUSTODY TREATMENT PROGRAMS 7 (2014), http://cpp.fullerton.edu/pdf/What%20Works.pdf
[https://perma.cc/4G8V-8U9N].
138 See Brandon A. Gaudiano, Cognitive-Behavioral Therapies: Achievements and
Challenges, 11 EVIDENCE BASED MENTAL HEALTH 5, 5–7 (2008).
139 See Susan H. Spence, Social Skills Training with Children and Young People: Theory,
Evidence and Practice, 8 CHILD & ADOLESCENT MENTAL HEALTH 84, 90 (2003) (“The practice
of target responses is essential for skill acquisition and improvement.”).
140 See discussion supra Section I.A.
141 Laub & Sampson, supra note 16, at 11.
142 Id.
143 Id. at 12.
144 Id.
145 Jeffrey T. Ulmer & Darrell Steffensmeier, The Age and Crime Relationship: Social
Variation, Social Explanations, in THE NURTURE VERSUS BIOSOCIAL DEBATE IN CRIMINOLOGY
378 (K. Beaver, B. Boutwell, and J.C. Barnes eds., 2014) (“It is a truism that age is one of the
strongest factors associated with criminal behavior.”).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
800 KLINGELE [Vol. 109

normative,146 particularly among males—a fact that explains why most

crimes are committed by men between the ages of fifteen and twenty-five.147
With age usually (and gradually) comes wisdom, lower energy, more peer
restraints, and, typically, desistance from crime.148 Although the pathways
into and out of criminal behavior vary from person to person, there are some
factors that repeatedly have been found to affect desistance:
[D]esistance stems from a variety of complex processes—developmental,
psychological, and sociological—and thus there are several factors associated with it.
The key elements seem to be aging; a good marriage; securing legal, stable work; and
deciding to “go straight,” including a reorientation of the costs and benefits of crime.
Processes of desistance from crime in general, specific types of crime, and multiple
forms of problem behavior seem to be quite similar.149
What is it about these elements that promotes desistance? Most scholars
conclude that it is not the events per se, but rather the moral and emotional
“turning points” they provide that create opportunities for people to redefine
their identities and accordingly reshape their behavior.150 No single turning
point can be guaranteed to change behavior, but each offers an opportunity
for re-orientation. In this model, criminal sanctions, educational attainments,
drug treatment, or a new relationship all serve a similar function with respect
to behavioral change: they provide a new chapter in a person’s life narrative
that may reinforce a negative self-identity or offer a chance to change one’s

146 Christopher Uggen & Michael Massoglia, Desistance from Crime and Deviance as a
Turning Point in the Life Course, in HANDBOOK OF THE LIFE COURSE 312 (Jeylan T. Mortimer
& Michael J. Shanahan eds., 2002) (“[A] long line of self-report research has demonstrated
that almost every adolescent admits to some form of delinquency.”); Ulmer & Steffensmeier,
supra, note 145, at 378 (“[A] significant portion of U.S. national crime rate trends over time
can be explained by fluctuations in the proportion of the population in the crime-prone age
group of fifteen- to twenty-four-year-olds.).
147 Ulmer & Steffensmeier, supra note 145, at 377–78.
148 See MICHAEL GOTTFREDSON & TRAVIS HIRSCHI, A GENERAL THEORY OF CRIME 124–
26 (1990); ROBERT J. SAMPSON & JOHN H. LAUB, CRIME IN THE MAKING: PATHWAYS AND
TURNING POINTS THROUGH LIFE 6 (1993); Travis Hirschi & Michael Gottfredson, Age and the
Explanation of Crime, 89 AM. J. SOC. 552 (1983).
149 Laub & Sampson, supra note 16, at 3.
150 Robert J. Sampson & John H. Laub, A Life-Course Theory of Cumulative
Disadvantage and the Stability of Delinquency, in DEVELOPMENTAL THEORIES OF CRIME AND
DELINQ. 1133, 1143 (Terence P. Thornberry ed., 1997) (quoting MICHAEL RUTTER &
MARJORIE RUTTER, DEVELOPING MINDS: CHALLENGE AND CONTINUITY ACROSS THE LIFESPAN
244 (1993) (“In our theoretical model, turning points may be positive or negative because they
represent ‘times of decision or opportunity when life trajectories may be directed on to more
adaptive or maladaptive paths.’”).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 801

story.151 Although there are documented cases in which a singular, salient

event precipitated the decision to “go straight,” a more common story is that
multiple opportunities for change combine, leading to a gradual
diminishment in criminal behavior. Whether this is understood as a change
in identity,152 a change in social role,153 or a re-interpretation of a life
narrative,154 the result is the same: through a process of reflection and
socialization, a person disentangles herself from crime.

B. MEASURING PROGRESS
If change is a process rather than an event, using a binary metric like
recidivism is too one-dimensional. While recidivism data tell us whether
those exposed to various interventions re-offended during the follow-up
window, they tell us nothing about the nature of the re-offense or whether the
trajectory of a person’s subsequent contacts with the law suggest a move
toward desistance or away from it. But what alternatives exist? Not
surprisingly, answers can be found by again considering how fields related
to criminal justice respond to the challenge of measuring change over time.
Education is a good place to start. Like criminal justice, American
primary education is a local enterprise with significant variation among states
in the contents and methods of teaching.155 Also like criminal justice,
education involves the outlay of significant public funds: in 2013–2014,
states and the federal government spent $634 billion on K–12 education.156
With that investment comes a demand for public accountability.
In recent years, there have been numerous attempts at the federal and
state levels to standardize curricula and require the generation and analysis
of metrics, such as standardized test scores, to hold teachers accountable for

151 See generally SHADD MARUNA, MAKING GOOD: HOW E X-CONVICTS REFORM AND
REBUILD THEIR LIVES (2001) (discussing differences in the autobiographical narratives of
“persisters” and “desisters”); Uggen & Massoglia, supra note 146, at 311.
152 Ray Paternoster & Shawn Bushway, Desistance and the “Feared Self”: Toward an
Identity Theory of Criminal Desistance, 99 J. CRIM. L. & CRIMINOLOGY 1103, 1105 (2009).
153 See generally ROBERT J. SAMPSON & JOHN H. LAUB, CRIME IN THE MAKING: PATHWAYS
AND TURNING POINTS THROUGH LIFE (1993); David Best, Social Identity, Social Networks and
Social Capital in Desistance and Recovery, in MOVING ON FROM CRIME AND SUBSTANCE USE:
TRANSFORMING IDENTITIES (Anne Robinson & Paula Hamilton eds., 2016).
154 MARUNA, supra note 151.
155 See Ralph D. Mawdsley & Paul Williams, Teacher Assessment and Credentialing: The
Role of the Federal Government in a State Function, 262 ED. L. REP. 735, 736 (2011)
(discussing federal legislation responding to “a lack of consistency among the states”).
156 U.S. DEPT. ED., NAT’L CTR. ED. STAT., THE CONDITION OF EDUCATION (2017).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
802 KLINGELE [Vol. 109

student learning.157 Standardized tests, which students across the country are
required to take at periodic intervals,158 are the educational analogs to
recidivism rates. They measure educational achievement in tested subject
matter and nothing more. When high percentages of students make adequate
yearly progress on grade level expectations, state and local education systems
are often lauded as successes, and when large numbers fail to make such
progress, schools and teachers are labeled as “failing.”159
A difference between education and criminal justice is that while
criminal justice administrators have largely accepted recidivism as a valid
metric of their own success, educators and educational administrators have
heavily resisted such simplistic measures of achievement.160 Recognizing
the wide variety of factors that affect student learning—many of which lie
outside educators’ control—teachers across the country have fought efforts
to define educational success by reference only to test scores.161 Through
unions lobbying, media campaigns, and grassroots coalition building,
educators have fought against what they assert has been an over-emphasis on
tests scores alone as a metric of student and teacher achievement.162
In response, some districts have introduced “value-added” assessments
to better capture the role of teachers in student growth over time. The goal

157 See, e.g., Jessica E. Ozalp & Rachel E. Snyder, School Accountability After “No Child
Left Behind,” 89 WIS. LAW 28, 28 (July/Aug. 2016) (explaining that both the Every Student
Succeeds Act and its predecessor legislation, No Child Left Behind, “leverage student
standardized test scores to pressure underachieving schools to improve on measures of student
progress” by “holding states, schools, and teachers accountable to test and report the
performance and progress of disadvantaged subgroups of students”).
158 ESSA ASSESSMENT NFR SUMMARY FACT SHEET FOR FINAL REG. TITLE I PART A AND
B, Dec. 7, 2017, https://www2.ed.gov/policy/elsec/leg/essa/essaassessmentfactsheet1207.pdf
[https://perma.cc/4JM7-AY3V].
159 See Monica Teixeira de Sousa, A Race to the Bottom? President Obama’s Incomplete
and Conservative Strategy for Reforming Education in Struggling Schools or the Perils of
Ignoring Poverty, 39 STETSON L. REV. 629, 656 (2010); James E. Ryan, The Perverse
Incentives of the No Child Left Behind Act, 79 N.Y.U L. REV. 932, 940 (2004);.
160 See, e.g., Noel K. Gallagher, Teachers Balk at New Performance Evaluation
Standard’s Reliance on Test Scores, PORTLAND HERALD PRESS (Jan. 28, 2013), https://www.
centralmaine.com/2013/01/28/teachers-resist-evaluation-proposal_2013-01-28/ [https://perm
a.cc/TV8B-HNNA]; Kathleen Megan, State Eliminates Test Scores From Teacher
Evaluations, HARTFORD COURANT (Apr. 5, 2017), https://www.courant.com/education/hc-stat
e-board-education-teacher-evaluations-20170405-story.html [https://perma.cc/U59U-8SCL].
161 Preston C. Green et al., The Legal and Policy Implications of Value-Added Teacher
Assessment Policies, 2012 B.Y.U. EDUC. & L.J. 1, 5–15 (2012).
162 See, e.g., Kate Taylor & Motoko Rich, Teachers’ Unions Fight Standardized Testing,
and Find Diverse Allies, N.Y. TIMES (Apr. 20, 2015), https://www.nytimes.com/2015/04/21/
education/teachers-unions-reasserting-themselves-with-push-against-standardized-testing.ht
ml [https://perma.cc/2VMP-GCYK].

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 803

of these tools is to control for “exogenous factors” such as poverty, friends,

and familial stability, “focus[ing] on achievement gains over time for the
same individual or groups of students.”163 Although such assessments
theoretically offer a more nuanced way of measuring student growth—and
the role of teachers in fostering it—they too have been met with pushback
from educators.164 Their objections are grounded in part on the difficulty of
“isolat[ing] one specific teacher’s contribution to students’ learning, leading
to situations where a teacher might be identified as a bad teacher simply
because her colleagues are ineffective.”165
Educators have argued instead that because “learning is a complex
process,” assessments of learning
should . . . emplo[y] a diverse array of methods, including those that call for actual
performance, using them over time so as to reveal change, growth, and increasing
degrees of integration. Such an approach aims for a more complete and accurate picture
of learning, and therefore firmer bases for improving . . . students’ educational
experience.166
In other words, without rejecting the role that data can play in assessing
the success of the educational enterprise, teachers and school administrators
have required that data be used to trace the learning process itself, looking
for multiple forms of evidence of growth over time rather than simple
snapshots of isolated performance.
Of course, there are key differences between education and corrections.
Failure to learn can reduce the quality of students’ lives, but it does not
imperil the lives of others. Crime sometimes does endanger others, and it
nearly always reduces the quality of life for those affected by it. In light of
that difference, it might seem fair to hold the criminal justice system to a
higher standard. The problem is that overreliance on recidivism provides
only one-dimensional data that masks important information about the
degree to which criminal justice interventions are, in the words of educational
assessment, “adding value”—or reducing it. Examining more nuanced data
about individual and aggregate behavioral change over time could yield
much more revealing information about not just whether, but in what ways

163 Ryan, supra note 159, at 981.

164 See, e.g., Valerie Strauss, Principals Reject ‘Value-Added’ Assessment that Links Test
Scores to Educators’ Jobs, WASH. POST (Dec. 4, 2014), https://www.washingtonpost.com/
news/answer-sheet/wp/2014/12/04/principals-reject-value-added-assessment-that-links-test-
scores-to-educators-jobs/ [https://perma.cc/NH6D-RY4S].
165 Green et al., supra note 161, at 10. This challenge is not dissimilar from the difficulty
of isolating the effects of any given prison program from other confounding factors that might
affect long-term desistance.
166 ALEXANDER ASTIN ET. AL., AMERICAN ASSOCIATION FOR HIGHER LEARNING: F ORUM,
NINE PRINCIPLES OF GOOD PRACTICE FOR ASSESSING STUDENT LEARNING 1(1992).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
804 KLINGELE [Vol. 109

people recidivate and the degree to which criminal justice programs move
them away from criminal behavior or propel them toward it.
So, what relevance does the experience of educators have for attempts
to measure success in the criminal justice system? First, like educators,
criminal justice system actors should demand an acknowledgement by
policymakers and the public that recidivism is affected by a host of
interacting, exogenous factors, many of which the criminal justice system
can do little to affect. Second, like educators, criminal justice system actors
should resist efforts to oversimplify the effectiveness or ineffectiveness of
interventions by reference to a measure as insensitive as pure recidivism
rates.
But what is the alternative to relying on recidivism? The answer to that
question can be found in the way criminologists measure change over time.
Although criminologists gather data on recidivism, they do not examine it in
isolation. Rather, in identifying “what works” to promote desistance,
criminologists look not at isolated recidivism rates but rather at how those
rates of offense change over time for individual offenders and among readily-
identifiable subgroups of offenders (“youth” or “burglars” for example) as
well as how those changes connect to other key contextual events.167 The
richest studies meld qualitative and quantitative data to create a three-
dimensional view of the effects of various life events and interventions.
Those in the criminal justice system do not typically have the resources
to extend follow-up windows or conduct surveys or interviews. That does
not mean, however, that they must settle for pure recidivism rates. In recent
years, policy-engaged criminologists have suggested that criminal justice
administrators should look beyond recidivism rates to identify the patterns of
behavior those rates can reveal over time. In a notable policy brief, Ryan
King and Brian Elderbroom of the Urban Institute urged state program
administrators to think more broadly about measuring programmatic
success.168 They explained, “Though failure rates should serve as the
foundation of recidivism research, it is critical to move beyond them to
improving recidivism as a performance measure.”169 Specifically, they urged
states to include in reports measures such as “time to failure, crime severity,
and behavior changes as indicators of success.”170 More recently, in a white
paper arising out of the Harvard Kennedy School Executive Session on

167 See, e.g., MARUNA, supra note 151.

168 RYAN KING & BRIAN ELDERBROOM, IMPROVING RECIDIVISM AS A PERFORMANCE
MEASURE, URB. INST. 2 (2014), https://www.bja.gov/Publications/UI-ImprovingRecidivism.
pdf [https://perma.cc/G4XD-KCAA].
169 Id.
170 Id. at 3.

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 805

Community Corrections, Jeffrey Butts and Vincent Schiraldi argued that

“[j]ustice policy should reduce the importance of recidivism and focus on
desistance.”171 They suggest that doing so might involve not only the
analysis of existing data but also the collection of additional information
about the changing “attitudes, perceptions, and beliefs” of people who have
been the subject of criminal justice intervention.172
Replete throughout the desistance literature are references to indicators
that signal desistance or its absence.173 Some of these, such as attitudes about
the law, are best captured through qualitative research. Several, however—
particularly changes in offense severity over time and alterations in the
frequency of offending—are easily derived from existing recidivism data by
calculating changes in crime severity over time, and the time between new
criminal incidents. Rarely, however, do these markers of desistance make
their way into reports assessing the effectiveness of any particular criminal
justice program.
To illustrate this point, imagine that you are a program analyst who has
decided to conduct a five-year recidivism study examining re-arrest rates of
all graduates of a prison-based anger management course. To assess
recidivism rates, you will need to collect arrest records for program graduates
and for members of a control group of prisoners who did not take the class
but are otherwise similar to those who did. With just those records, you could
report not just on recidivism but on a great deal more. The official record
that tells you John Doe was re-arrested within five years of release will also
tell you for what he was arrested and how long after his release the arrest
occurred. Moreover, if John Doe is arrested three times more, those arrest
records, when viewed together and compared against John’s criminal history,
will tell you whether he is slowing down or speeding up in terms of his
criminal productivity and whether his behavior is escalating, diminishing, or
holding steady in terms of severity. That kind of information, averaged out
across the larger population of graduates and controls, can also tell you
whether a study group is, in the main, offending more or less the same or
differently than they historically have done in addition to how their offending

171 JEFFREY A. BUTTS & VINCENT SCHIRALDI, RECIDIVISM RECONSIDERED: PRESERVING

THE COMMUNITY JUSTICE MISSION OF COMMUNITY CORRECTIONS, HARVARD KENNEDY
SCHOOL PROGRAM IN CRIMINAL JUST. POL’Y MANAGEMENT 13 (2018), https://www.hks.
harvard.edu/sites/default/files/centers/wiener/programs/pcj/files/recidivism_reconsidered.pdf
[https://perma.cc/QV7W-5ATX].
172 Id.
173 See Ros Burnett & Fergus McNeill, The Place of the Officer-Offender Relationship in
Assisting Offenders to Desist from Crime, 52 PROB. J. 221, 247–68 (2005); John H. Laub et
al., Trajectories of Change in Criminal Offending: Good Marriages and the Desistance
Process, 63 AM. SOC. REV. 225 (1998).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
806 KLINGELE [Vol. 109

patterns compare to those of the control group. This information would allow
a picture to emerge of change over time—a picture that is less focused on
whether former offenders have fully terminated their criminal careers
(though it could reveal that as well) and more interested in whether they are
desisting from or persisting in criminal behavior as a result of their
engagement with a specific intervention. Although this data would be subject
to all of the limitations of data on recidivism generally (inconsistent
definitions, over and under-inclusion of criminal behavior through reliance
on official court records, etc.)174, it would be far superior to current reliance
on recidivism rates alone.
It would be inaccurate to say that the methods described above, so
commonplace in academic criminological research, are never reflected in
evaluations by professionals within the criminal justice system. Periodically,
system analysts will issue a study or white paper demonstrating the ability to
use recidivism data to produce more nuanced information (reporting on the
age of recidivists or classifying the nature of their re-arrest, for example)175,
but these are the exception and not the rule. Far too often, recidivism
continues to be reported—and therefore understood—as a binary event rather
than as the dynamic process it is. As the following section explains, that is a
missed opportunity.

IV. THE LIMITS OF RECIDIVISM DATA

In light of the mechanisms of change discussed above, it is clear that
using markers of desistance to measure and report on the effectiveness of
criminal justice programs would have significant benefits. It may be less
clear, however, how the failure to utilize these more nuanced markers is
actively harmful to the criminal justice system and those affected by it. The
following section discusses three problems that can be created when the
binary metric of recidivism is used to measure the success of the criminal
justice system. These include data gaming, risk-aversive behavior by system
actors, and the premature termination of otherwise-promising programs for
advancing public safety.

174 See supra Section II; infra Section IV.

175 See, e.g., U.S. SENT’G COMM’N, THE EFFECTS OF AGING ON RECIDIVISM AMONG
FEDERAL OFFENDERS 24–25 (2017), https://www.ussc.gov/sites/default/files/pdf/research-and
-publications/research-publications/2017/20171207_Recidivism-Age.pdf [https://perma.cc/Q
89X-9C4G].

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 807

A. DATA GAMING
Criminal justice programs are under tremendous pressure to produce
data that demonstrates significant “good” outcomes, and that pressure can—
in ways both conscious and subconscious—affect the validity of the data they
generate.176 Scandals in education related to the falsification of standardized
test results are an analogous example;177 making criminal justice system
actors accountable for reducing recidivism incentivizes gaming the
underlying data. Placing a high premium on recidivism rates encourages
program administrators not to reduce re-offending so much as to reduce
statistical rates of recidivism.178 As a result, improvements in recidivism
rates often tell us more about the behavior of system actors than of
individuals with prior criminal records. Two examples help illustrate this
point.
First, consider recent efforts across the country to reduce failure rates
for individuals on community supervision. In states from California to North
Carolina, large numbers of people have, at times, been re-incarcerated while
serving sentences of probation or parole.179 Sometimes, the triggering event
for re-incarceration is the commission of a new crime.180 Other times,
incarceration is triggered by the violation of non-criminal rule, ranging from

176 See Mia Bird & Ryken Grattet, Realignment and Recidivism, 664 ANN. AM. ACAD.
POL’Y & SOC. SCI. 176, 192 (2016) (“[A]n observed change in a particular measure of
recidivism may reflect a change in offender behavior; it can also be inadvertently affected by
criminal justice process changes or even intentionally manipulated to produce indications of
success.”); Anthony Petrosino & Haluk Soydan, The Impact of Program Developers as
Evaluators on Criminal Recidivism: Results from Meta-Analyses of Experimental and Quasi-
Experimental Research, 1 J. EXPERIMENTAL CRIMINOLOGY 435, 444 (2005) (finding
“consistent with prior meta-analyses in offender rehabilitation, . . . that studies in which
evaluators were greatly influential in the design and implementation of treatment report
consistently and substantially larger effect sizes than other types of evaluators.”).
177 Motoko Rich, Scandal in Atlanta Reignites Debate Over Tests’ Role, N.Y. TIMES (Apr.
2, 2013), https://www.nytimes.com/2013/04/03/education/atlanta-cheating-scandal-reignites-
testing-debate.html [https://perma.cc/Q4V3-LLNK]; THOMAS S. DEE ET AL., THE CAUSES AND
CONSEQUENCES OF TEST SCORE MANIPULATION: EVIDENCE FROM THE NEW YORK REGENTS
EXAMINATIONS, NAT’L BUREAU ECON. RES. (2016), http://www.nber.org/papers/w22165.pdf
[https://perma.cc/76JF-HB5T].
178 Cullen et al., supra note 60, at 51 (“Another challenge is to ensure that correctional
personnel cannot influence outcome measures.”).
179 Klingele, supra note 85, at 1019, 1030–31.
180 PEW CTR. ON STATES, STATE OF RECIDIVISM: THE REVOLVING DOOR OF AMERICA’S
PRISONS 7, 9 (2011).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
808 KLINGELE [Vol. 109

minor infractions to more serious violations. 181 As discussed earlier, many

times these returns to custody are coded as instances of recidivism. 182
Recognizing the political, fiscal, and human costs associated with such
community supervision “failure,” many states have undertaken efforts to
reduce their revocation rates.183 They have often done so not by changing
the behavior of those being supervised but rather by changing the behavior
of those doing the supervising. California offers a prime example. Prior to
the United States Supreme Court’s ruling in Brown v. Plata,184 California had
one of the highest parole revocation rates in the nation at more than 66% in
2005.185 In Plata, the Court held that as a result of ongoing Constitutional
violations in the conditions of its prisons, California was required to reduce
its overall prison population by 37,000 people.186 Recognizing the enormity
of this reduction, the Court offered suggestions for how such a change could
be safely accomplished, including by diverting “technical parole violators to
community-based programs.”187
That is, in fact, exactly what California did (among other things).
Following Plata, parolee arrest and revocation rates plummeted, falling from
44.6% in 2011188 to 7% twelve months later.189 What drove this dramatic
change? Bird & Grattet report:
There is anecdotal evidence that arrest decisions are shifting from a lesser reliance on
arrests for supervision violations to a greater reliance on arrests for felonies . . . . These
shifting patterns will directly impact reported rates of recidivism, and it will be

181 Klingele, supra note 85, at 1030–31 (“One study followed individuals released from
prison in 2004 and found that in thirteen states, 25% or more of those released were
reincarcerated for purely ‘technical’ violations of community supervision within three
years.”).
182 See PEW CTR. ON STATES, supra note 180, at 33–37; discussion supra Section II.A, at
9.
183 See, e.g., PEW C HARITABLE TRUSTS, REDUCING INCARCERATION FOR TECHNICAL
VIOLATIONS IN LOUISIANA (2014), http://www.pewtrusts.org/~/media/assets/2014/11/psp
preducingincarcerationfortechnicalviolationsinlouisiana.pdf [https://perma.cc/6PRU-FUY6].
184 Brown v. Plata, 563 U.S. 493, 502 (2011) (ordering California prisons to reduce
overcrowding following years of Constitutional violations brought about by deficient
conditions).
185 Ryken Grattet et al., Parole Violations and Revocations in California: Analysis and
Suggestions for Action, 73 FED. PROBATION 2, 2 (2009).
186 Plata, 563 U.S. at 501.
187 Id.
188 CAL. DEP’T. OF CORRS. & REHAB., 2015 OUTCOME EVALUATION REPORT: AN
EXAMINATION OF OFFENDERS RELEASED IN FISCAL YEAR 2010–11 1 (2016), https://dev-multi
site.mystagingwebsite.com/research/wp-content/uploads/sites/9/2018/04/2015-Outcome-Eva
luation-Report.pdf [https://perma.cc/69BV-FU38].
189 Bird & Grattet, supra note 176, at 186 (“In one year’s time, California went from
having one of the highest return-to-custody rates in the nation to one of the lowest.”).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 809

challenging to separate changes in discretionary decision-making from changes in

underlying offender behavior.190
A similar phenomenon can be seen in Arizona. In 2008, 6,800 people
were revoked from supervision in the state.191 That same year, the state
legislature passed the Safe Communities Act, which, among other things,
shortened periods of supervision for people who were in compliance with the
terms of supervision (thereby reducing the time they were “at hazard” of
revocation) and “[c]reated incentives for county probation agencies to reduce
revocations.”192 The effort was met with modest success: while the
supervision population held steady, the number of annual revocations had
decreased to 4,800 by 2016.193
In a short period of time, both California and Arizona saw notable
reductions in recidivism-defined-as-revocation rates for those on community
supervision.194 In both cases, however, there is no reason to think that these
improvements in recidivism-as-revocation rates were driven by changes in
the behavior of people on supervision.195 Instead, the reductions appear to
reflect a conscious decision by the agencies overseeing probationers and
parolees to take a less punitive approach to violations of supervision rules,
either by choice (Arizona) or legal necessity (California).196 While such

190 Id. at 193.

191 COUNCIL OF STATE GOV’TS, REDUCING RECIDIVISM: STATES DELIVER RESULTS 3
(2017), https://csgjusticecenter.org/wp-content/uploads/2017/12/12.15.17_Reducing-Recidiv
ism.pdf [https://perma.cc/8NQ6-APU4].
192 Id. at 4.
193 Id. at 3.
194 Id. at 4 (showing 29% decline between 2008 and 2016 in Arizona in people admitted
to prison due to probation violations); Bird & Grattet, supra note 176, 186–87 (describing
California’s plummeting revocation rate).
195 See MIA BIRD ET AL., REALIGNMENT AND RECIDIVISM IN CALIFORNIA, PUB. POL’Y INST.
OF CAL. 9 (2017) (“[W]hen comparing the reconviction rates of individuals released before
and after realignment, it is important to consider how realignment changed the likelihood that
criminal justice systems would pursue formal convictions . . . . Under realignment, most
offenders can only be sent to prison following conviction on a new, prison-eligible offense—
meaning that correctional systems may be shifting away from revocations and toward more
formal rearrests and reconvictions. Therefore, if we were to see increases in reconviction rates
after realignment, we should not necessarily assume these differences were driven by an
increase in reoffending behavior.”).
196 Cf. ARIZONA ADULT PROBATION FY 2016 ANNUAL REPORT, AZ SUP. CT. ADMIN. OFF.
OF THE CTS. 5-6 (2016), https://www.azcourts.gov/Portals/25/AnnRepPop/FY2016_%20RE
PORT.pdf [https://perma.cc/55AN-98BW] (quoting the Director of Adult Probation Services:
“We have moved low risk offenders to a very different level of supervision that requires
accountability but not the usual resources utilized and required by higher risk offenders to be
successful. Based on the research, our focus is now on the higher risk individuals which are
the offenders with the higher priority on the caseloads for supervision . . . . [W]e continue to

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
810 KLINGELE [Vol. 109

outcomes may reflect wise public policy decisions, they do not suggest that
those on supervision desisted. Instead, they suggest that system actors
changed their response. These examples illustrate the ways in which
recidivism data can be altered in ways that reflect external, political factors
as much individual, behavioral ones.
Probation supervision is not the only place where overreliance on
recidivism rates can alter the behavior of system actors in unintended ways.
Specialty courts, such as drug courts, provide another example of how the
behavior of system actors—rather than offenders—sometimes changes in
order to improve reported recidivism rates for program participants. Drug
courts are popular specialty programs that developed out of concern about
the over-criminalization of addiction.197 Drug courts offer a more
collaborative model of decision-making, with a judge and lawyers as
members of a larger “treatment team.”198 Participants are given access to
drug treatment and other needed resources in lieu of traditional adversarial
adjudication.199 Graduates typically see their original charges dismissed or
reduced, while those terminated for non-compliance are sentenced
traditionally.200 Program success (and in turn, continued funding) is
ordinarily determined by the recidivism rates of program graduates as
compared to traditionally-prosecuted defendants.201

focus on supervising, engaging, coaching and supporting successful completions of probation

as we have since the adoption of Evidence Based Practice.”).
197 See Peggy Fulton Hora & Theodore Stalcup, Drug Treatment Courts in the Twenty-
First Century: The Evolution of the Revolution in Problem-Solving Courts, 42 GA. L.
REV. 717, 725 (2008); see also Drug Courts, NAT’L INST. JUST., https://www.nij.gov/topics/
courts/drug-courts/Pages/welcome.aspx [https://perma.cc/ZC9K-3N74].
198 See Eric J. Miller, Embracing Addiction: Drug Courts and the False Promise of
Judicial Interventionism, 65 OHIO ST. L.J. 1479, 1492 (2004).
199 Estimates suggest that in 2012, there were more than 1,300 drug courts operating in
the United States. SUZANNE M. STRONG ET AL., CENSUS OF PROBLEM-SOLVING COURTS, 2012,
BUREAU OF JUST. STAT. 3 (2016), https://www.bjs.gov/content/pub/pdf/cpsc12.pdf [https://per
ma.cc/6VYC-S6DA]. The popularity of treatment courts expands to other social problems as
well, including mental illness, domestic violence, veterans’ issues, and drunken driving:
estimates suggest there are currently over 1,700 problem-solving courts addressing issues
other than drug use. Id. Most of these, like drug courts, are required by their funding agencies
to track recidivism among program graduates.
200 Josh Bowers, Contraindicated Drug Courts, 55 UCLA L. REV. 783, 788 (2008).
201 In New Mexico, for example, the Legislative Finance Committee’s Program
Evaluation Unit presented findings to the state legislature on the performance of drug courts
statewide, making recommendations for certification that include “performance measures
such as graduation, recidivism, and costs . . . to ensure courts are meeting meaningful
performance targets.” N.M. LEG. FIN. COMM., PROGRAM EVALUATION: UPDATE ON NEW
MEXICO DRUG COURTS, 53rd Sess., at 26 (2017). Similarly, funding for adult drug courts from
the U.S. Department of Justice requires grant applicants to “[b]riefly describe the data
collection mechanism that the applicant will use to collect and report in-program progress as

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 811

Placing such significance on graduates’ recidivism rates creates two

well-documented problems in data generation. First, drug courts are
incentivized to limit entry to individuals with the highest prospect of
successfully completing the program and going on to avoid re-offense. In
many cases, this means that drug courts limit participation to first-time, non-
violent offenders with no co-occurring conditions202—the very people who
are least likely to need any intervention at all. For many of these individuals,
the program is likely to make no difference; they were already unlikely to
commit new crimes.
In addition to limiting entry to low-risk candidates, over emphasis on
recidivism rates also pressures drug courts to quickly terminate from the
program any participants who are having difficulty remaining sober—a
difficulty those with serious substance abuse problems would be expected to
have. Because these addicts are predictably less likely than their less-
addicted counterparts to avoid future reconviction, drug court administrators
have an incentive to exclude them from the cohort of graduates whose
performance will drive future funding of the program.203 This disincentive
to continue working with more serious drug offenders undercuts the very
purpose for which drug courts exist (i.e., to assist people in overcoming
addiction and related criminal behavior). It also leads to harsher punishment
for serious addicts. Research has consistently found that individuals
sentenced after being terminated from drug court programs face punishment
many times harsher than traditionally-sentenced drug offenders.204

well as post-program recidivism information.” BUREAU OF JUST. ASSISTANCE, Adult Drug

Court Discretionary Grant Program FY 2018 Competitive Grant Announcement, 20 (2018)
https://www.bja.gov/funding/DrugCourts18.pdf [https://perma.cc/ZRA4-9LVD].
202 AVINASH SINGH BHATI ET AL., URB. INST. JUST. POL’Y CTR., TO TREAT OR NOT TO
TREAT: EVIDENCE ON THE PROSPECTS OF EXPANDING TREATMENT TO DRUG-INVOLVED
OFFENDERS 7–8 (2008), https://www.urban.org/sites/default/files/publication/31621/411645-
To-Treat-or-Not-to-Treat.PDF [https://perma.cc/4X8Z-8EBY].
203 Cf. John A. Bozza, Benevolent Behavior Modification: Understanding the Nature and
Limitations of Problem-Solving Courts, 17 WIDENER L. J. 97, 118–19 (2007) (“Drug courts,
and probably mental health courts, are by definition dealing with a special population of
offenders. They are selective. Certain offenders, because of bad criminal histories or because
they have committed certain crimes, cannot participate. Moreover, of those who are eligible,
only those amenable to treatment are chosen and they are not necessarily individuals who are
addicted to drugs.”).
204 See Bowers, supra note 200. It is also true, however, that due to the length of drug
court interventions, even “defendants who progress through the entire treatment court regime
and who upon ‘graduation’ have their charges dismissed ultimately may have received ‘a more
onerous disposition in terms of the length of time [they are] subject to court control than they
would have received if the charges had been resolved through standard plea negotiations or
trial.” Richard C. Boldt, Rehabilitative Punishment and the Drug Court Treatment Court
Movement, 76 WASH. U. L. Q. 1205, 1256 (1998).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
812 KLINGELE [Vol. 109

Overreliance on recidivism rates as a success metric encourages drug

court programs to reject the very people who are most likely to benefit from
the extra resources the drug court offers. The irony is acute because
substance abuse treatment providers have long embraced the mantra that
“relapse is a part of recovery” and that the road to sobriety almost always
includes set-backs205—many of which would qualify as “recidivism” if
detected and punished.
The pressure to create good outcome statistics by limiting access to
programs is not limited to the drug court context. All criminal justice
programs that are assessed based on their reported reductions in recidivism
rates find themselves in a similar bind: admit those who most need help and
who are therefore at highest risk of recidivating, or admit those with lower
risks and needs who will be less likely to recidivate. When funding turns on
recidivism reductions, the pressure to game the data is particularly acute.206
If you admit those in most need of help, you increase your reported failure
rates; if you admit those who do not need your help, your programs success
rates skyrocket. Discussing this problem in the context of juvenile justice,
one foundation director explained; “One way to have low recidivism rates is
to sweep up kids who are low risk and formally handle them in the
system . . . .If the system concentrates only on very high-risk kids you’ll
probably have higher recidivism rates.”207
How would these outcomes be different if desistance were the measure
of success rather than recidivism? While there are opportunities for data
gaming in any system, the incentives to game decrease as emphasis shifts
away from demonstrating perfect outcomes and toward demonstrating
realistic, measurable progress. If programs were deemed successful when

205 See, e.g., Lisa Lightman & Francine Byrne, Addressing the Co-occurrence of Domestic
Violence and Substance Abuse: Lessons from Problem-Solving Courts, 6 J. CTR FOR FAM.,
CHILD. & CTS. 53, 65 (2005) (“Relapse . . . is generally considered a common element in the
process of recovery from substance abuse.”); see also David Sack, Why Relapse Isn’t a Sign
of Failure, PSYCH. TODAY (Oct. 19, 2012), https://www.psychologytoday.com/blog/where-sci
ence-meets-the-steps/201210/why-relapse-isnt-sign-failure [https://perma.cc/7JCR-FXFN].
206 See, e.g., CTR. FOR STATE GOV’TS, REDUCING RECIDIVISM 12 (2014) https://csgjustice
center.org/wp-content/uploads/2014/06/ReducingRecidivism_StatesDeliverResults.pdf [http
s://perma.cc/Y4E3-DAHV] (describing a Pennsylvania “performance incentive funding
model in which contractors work with [the department of corrections] establish baseline
recidivism rates and then to review their recidivism rates every six months. Contractors can
receive additional funds if they reduce recidivism below the baseline, or are at risk of having
their contracts revoked if recidivism is above the baseline.”).
207 Sanya Mansoor, Different Ways of Measuring Recidivism Leads to Incomplete Data,
Report Says, JUV. JUST. INFO. EXCHANGE (Sep. 11, 2014), https://jjie.org/2014/09/11/different-
ways-of-measuring-recidivism-leads-to-incomplete-data-report-says/ [https://perma.cc/5JYN
-JJGU] (quoting MacArthur Foundation program officer quoting Patrick Griffin).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 813

their participants improved and not just when they achieved perfection,
program administrators might become willing to treat more serious offenders
than they currently do. These more serious or prolific criminals might go on
to recidivate in some way but if the treatment they receive were to yield a
measurable decrease in later offense severity and frequency, public safety
would improve much more than if a minor offender is treated with a perfect
result. By removing the incentive for program administrators to focus their
efforts on those least likely to recidivate, policymakers could encourage the
development and expansion of programs that promote measurable desistance
among those with whom they intervene.

B. PREMATURE TERMINATION OF PROMISING PROGRAMS

A second problem created by overreliance on recidivism rates is the risk
that promising interventions may be prematurely declared failures. If the
only meaningful measure of a program’s success is the rate at which it
reduces recidivism, programs that demonstrably promote desistance, but do
not in themselves change recidivism rates, are at risk of termination.
As an initial matter, it is important to recognize that a sizeable
percentage of people who undergo any criminal justice intervention will not
recidivate. This is not because the criminal justice system exerts a pervasive
ameliorative influence, but rather because without intervention, most people
who come into contact with the criminal justice system once will not do so
again.208 As with any antisocial or maladaptive human behavior, crime can
be habitual, periodic, or anomalous. If the behavior is truly anomalous, then
desistance is not necessary: the offender moves directly from crime to
termination. The behavior will not repeat, and no intervention is needed.209
For those whose behavior is not anomalous, however, the process of
desistance “resides somewhere in the interfaces between developing personal
maturity, changing social bonds associated with certain life transitions, and
the individual subjective narrative constructions which offenders build

208 By one estimate, approximately two-thirds of first-time state prisoners will not return
to prison. Rosenfeld, supra note 113, at 313 (citing MICHAEL TONRY, THINKING ABOUT CRIME:
SENSE AND SENSIBILITY IN AMERICAN PENAL CULTURE 186–89 (2004)). A full 70% of juvenile
arrestees will never be re-arrested. DAVID MACALLAIR & TIM ROCHE, WIDENING THE NET IN
JUVENILE JUSTICE AND THE DANGERS OF PREVENTION AND EARLY INTERVENTION, JUST. POL’Y
INST. 5 (2001) http://www.cjcj.org/uploads/cjcj/documents/widening.pdf [https://perma.cc/P7
QE-RW7D] (“In other words, by doing nothing the state can achieve a 70 percent success
rate—meaning no subsequent arrests—with first time offenders.”).
209 In fact, a developing body of literature suggests that formal criminal justice
intervention in such cases may actually increase the risk of future criminal behavior. See D.
A. Andrews et al., Does Correctional Treatment Work? A Clinically Relevant and
Psychologically Informed Meta-Analysis, 28 CRIMINOLOGY 369 (1990).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
814 KLINGELE [Vol. 109

around these key events and changes.”210 In light of the complex interplay
between all of these factors, it is implausible to suggest that any one program
or intervention—however well-conceived or delivered—can reliably
transform persistent offenders into law-abiding citizens who will never again
run afoul of the law.
In reality, when a program claims to work dramatic changes in the
recidivism rates of its program graduates, it is likely that those individuals
did not need the program at all.211 A far more likely inference is that they
were already far along the path of desistance because their criminal behavior
was either anomalous or situational.
But while dramatic claims of recidivism reduction should always be a
cause for skepticism, the opposite is not always true. If a program does not
demonstrate its ability to significantly reduce recidivism rates, it is not
necessarily a poor investment. First, many programs have independent
humanitarian value. By providing convicted individuals with education,
emotional support, skill development, or treatment for addiction and other
challenges, the state recognizes the inherent dignity of those over whom it
exercises control.212 Second, at least some of these programs likely play an
important foundational role in fostering desistance.
Consider the transtheoretical model of change. Contemplation and
preparation are necessary prerequisites to action in that model, but those early
stages do not in themselves yield evidence of behavior change.213 Criminal
justice programs that provide opportunities for individuals to imagine change
where they previously could not or that assist individuals in devising plans
for how to make needed changes will not always result in reduced rates of
recidivism. But by laying a foundation for later change, such programs may
provide the necessary groundwork on which later progress will rest. To
identify such programs, it will be necessary to look for metrics more nuanced
than pure recidivism: metrics that better reflect the realities of behavioral
change and the influence that criminal justice interventions can have on
promoting that change over time.
Markers of desistance are a good start. By examining the pattern of
offending behavior that an individual displays over time, it is possible to
discern whether a person’s criminal behavior is escalating or diminishing and
whether it is gaining momentum or losing speed. Each alteration provides

210 Fergus McNeill, A Desistance Paradigm for Offender Management, 6 CRIMINOLOGY

& CRIM. JUST. 39, 47 (2006).
211 Cullen et al, supra, note 105.
212 FRANCIS T. CULLEN & KAREN E. GILBERT, REAFFIRMING R EHABILITATION 247–53
(1982).
213 See Prochaska & Velicer, supra note 117, at 38–39.

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 815

an opportunity for intervention, whether by connecting a freshly-desisting

young person with employment (a known pathway to long-term
desistance214) or by offering a cognitive-behavioral intervention to a person
whose behavior evinces a pattern of increasing entanglement with crime.

C. RISK AVERSION
Finally, overreliance on recidivism rates encourages risk aversive
behavior by system actors, thus exacerbating mass incarceration and limiting
opportunities for convicted individuals to master the skills that will
ultimately lead to long-term desistance.
The modern criminal justice system is risk averse by any measure. Over
the past half century, despite falling crime rates, criminal charges have
increased.215 Sentences have lengthened.216 Parole and other forms of back-
end release have been significantly restricted.217 Revocations from
community supervision have risen.218 Criminal justice bureaucrats are
haunted by the fear that, on their watch, a released defendant, probationer, or
parolee will commit a crime and wind up as front page news.219 The result
is a paralyzed system.
The emphasis that has been placed on recidivism rates and their
reduction psychologically reinforces the idea that the job of the criminal
justice system is to prevent crime at all costs, even if that means confining
behind bars people whose conduct does not pose a serious risk of harm to
others and those who have demonstrated that they are engaged in the process
of behavioral change. To the worried parole board member or probation
officer, a system that rises and falls on recidivism rates alone is one in which
any infraction, however small, counts as failure. In such a world, it will
always be easier to lock up the technical violator or never parole the model

214 See generally Christopher Uggen, Work as a Turning Point in the Life Course of
Criminals: A Duration Model of Age, Employment, and Recidivism, 65 AM. SOC. REV. 529
(2000).
215 See JOHN PFAFF, LOCKED IN: THE TRUE CAUSES OF MASS INCARCERATION 130 (2018)
(describing proliferation of criminal charges leveled by prosecutors).
216 Janice Williams, Serving Time: Average Prison Sentence in the U.S. is Getting Even
Longer, NEWSWEEK (July 22, 2017), http://www.newsweek.com/prison-sentences-increased-
2017-jail-639952 [archive] [https://perma.cc/TY4D-LFSA].
217 See NAZGOL GHANDNOOSH, THE SENTENCING PROJECT, DELAYING A S ECOND CHANCE:
THE DECLINING PROSPECTS FOR PAROLE ON LIFE SENTENCES (2017); Cecelia Klingele, The
Early Demise of Early Release, 114 W. VA. L. REV. 1, 19 (2011).
218 See generally Klingele, supra note 81.
219 See, e.g., Jacob Rosenberg, Will Another High Profile Crime Derail Parole Reform?,
ARK. TIMES (July 27, 2017), https://csgjusticecenter.org/jr/arkansas/media-clips/will-another-
high-profile-crime-derail-parole-reform/ [https://perma.cc/2AXU-UCUF].

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
816 KLINGELE [Vol. 109

inmate because such behaviors will always protect the political interests of
agencies and administrators.
If, instead, success were gauged by more holistic metrics—including
markers of desistance—incentives would shift. A chronic offender’s minor
violation, coming after a long period of good behavior, would be a cause for
attention but not necessarily alarm. Relapse by a chronic drug offender
would not be reason to declare treatment a failure if it reduced use and
associated new crime. Opportunities to celebrate success would increase for
both individuals in the system and the criminal justice system itself. Judges
and prosecutors, too, might begin to ask more critical questions about
defendants’ predicted risk of recidivism. Instead of reflexively balking at a
medium or high predicted risk of re-offense, system actors might ask
important interpretive questions such as, “What is this offender at risk of
doing?,” and “Does the evidence suggest that this person is on a path of
persistence or desistance?” The answers to those questions would guide the
use of predictive risk tools, placing recidivism risks into a larger, more
holistic context that is likely to yield sentences that would better advance
fairness and public safety.

CONCLUSION: TOWARD MARKERS OF DESISTANCE

In an article on recidivism, Robert Weisberg mused:
Perhaps we should not be asking whether a criminal recidivates or not, or whether
recidivist acts occur or not. Perhaps we should not even be focusing on the frequency
of recidivism for a person or a society . . . . [P]erhaps a better measure—or rough guide
to a better measure—is to ask whether as a result of a state intervention the offender
reoffends less frequently or less harmfully than he otherwise might, especially by
comparison to the likely downward arc of criminality due to aging.220
Indeed, given all that is known about how behavior change occurs and
what the criminal justice system can do to help or hinder that change process,
the present focus on recidivism rates misses the mark. It is only by looking
at the pattern of behavior revealed by recidivism data that those within and
outside the criminal justice system can begin to see the ways in which
criminal justice interventions may be promoting desistance and ways the in
which they may be hindering it.
Changing metrics of success from rates of recidivism to markers of
desistance will leave many problems unresolved. Like the simple recidivism
data on which they draw, markers of desistance can reveal only what
behavior has come to law enforcement attention, not what behavior is

220 Robert Weisberg, Meanings and Measures of Recidivism, 87 S. CAL. L. REV. 785, 804
(2014).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms
2019] MEASURING CHANGE 817

occurring on the streets and within homes. The underlying data will still be
skewed by inconsistencies in policing, charging, and correctional supervision
practices, as is the case with recidivism data more generally.221 Moreover,
and perhaps most importantly, markers of desistance will still fail to capture
many of the more holistic ways in which the pro-social progress of those
entangled in the criminal justice system might be measured. Markers of
desistance will not tell whether people are creating meaningful and healthy
relationships; or whether they are gainfully employed and able to meet their
basic human needs; whether they feel connected to their communities.222
These markers of well-being are not easily quantifiable, but they are no less
important measures of the ways in which criminal justice interventions may
be improving—or worsening—the condition of those with whom they
intervene.
Nevertheless, in a world committed to measurement, there are better
measurements than recidivism. By attending to markers of desistance,
policymakers and criminal justice system actors and administrators can begin
to develop a significantly more nuanced picture of individual change over
time and of the role that criminal justice interventions play in promoting or
hindering that change pro817cess.

221 See discussion supra Section I.A.

222 Cf. Michael M. O’Hear, The Second Chance Act and the Future of Reentry Reform, 20
FED. SENT’G REP. 75, 76 (2007) (expressing concern that “the tendency to frame and evaluate
reentry initiatives as solely, or even primarily, recidivism reduction measures” ignores “many
important social welfare and social justice concerns implicated in the treatment of returning
prisoners”).

This content downloaded from 186.121.194.185 on Sun, 10 Nov 2024 23:59:29 UTC
All use subject to https://about.jstor.org/terms