ArticlePDF Available

Abstract

Objectives Despite the dramatic expansion of the US correctional system in recent decades, little is known about the relative effectiveness of commonly used sanctions on recidivism. The goal of this paper is to address this research gap, and systematically examine the relative impacts on recidivism of four main types of sanctions: probation, intensive probation, jail, and prison. Methods Data on convicted felons in Florida were analyzed and propensity score matching analyses were used to estimate relative effects of each sanction type on 3-year reconviction rates. Results Estimated effects suggest that less severe sanctions are more likely to reduce recidivism. Conclusions The findings raise questions about the effectiveness of tougher sanctioning policies for reducing future criminal behavior. Implications for future research, theory, and policy are also discussed.
1 23
Journal of Quantitative Criminology
ISSN 0748-4518
Volume 30
Number 2
J Quant Criminol (2014) 30:317-347
DOI 10.1007/s10940-013-9205-2
Assessing the Effectiveness of Correctional
Sanctions
Joshua C.Cochran, Daniel P.Mears &
William D.Bales
1 23
Your article is protected by copyright and all
rights are held exclusively by Springer Science
+Business Media New York. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.
ORIGINAL PAPER
Assessing the Effectiveness of Correctional Sanctions
Joshua C. Cochran Daniel P. Mears William D. Bales
Published online: 13 August 2013
Springer Science+Business Media New York 2013
Abstract
Objectives Despite the dramatic expansion of the US correctional system in recent
decades, little is known about the relative effectiveness of commonly used sanctions on
recidivism. The goal of this paper is to address this research gap, and systematically
examine the relative impacts on recidivism of four main types of sanctions: probation,
intensive probation, jail, and prison.
Methods Data on convicted felons in Florida were analyzed and propensity score
matching analyses were used to estimate relative effects of each sanction type on 3-year
reconviction rates.
Results Estimated effects suggest that less severe sanctions are more likely to reduce
recidivism.
Conclusions The findings raise questions about the effectiveness of tougher sanctioning
policies for reducing future criminal behavior. Implications for future research, theory, and
policy are also discussed.
Keywords Sanctions Effectiveness Recidivism
J. C. Cochran (&)
Department of Criminology, University of South Florida, 4202 East Fowler Avenue, SOC 324, Tampa,
FL 33620-7200, USA
e-mail: jccochran@usf.edu
D. P. Mears (&)W. D. Bales
College of Criminology and Criminal Justice, Florida State University, 634 West Call Street,
Tallahassee, FL 32306-1127, USA
e-mail: dmears@fsu.edu
W. D. Bales
e-mail: wbales@fsu.edu
123
J Quant Criminol (2014) 30:317–347
DOI 10.1007/s10940-013-9205-2
Author's personal copy
Introduction
Over the past 30 years, the United States has witnessed a dramatic expansion of its cor-
rectional system. The increased growth in probation, jail, and prison populations has been
well documented (Glaze 2011) and has led to considerable scholarship aimed at under-
standing its causes and consequences (see, e.g., Garland 2001; Irwin 2005; Gottschalk
2006; Western 2007; Useem and Piehl 2008; Raphael and Stoll 2009; Blumstein 2011;
Cullen et al. 2011). One of the central justifications policymakers have invoked for tougher
sanctioning is a belief that it reduces offending, among those sanctioned, more so than less
severe punishment. Recently, however, community-based, non-custodial sanctions have
been promoted as a more effective approach, in part because their use may allow for the
provision of rehabilitative services that facilitate reintegration into society and, ultimately,
less offending (Petersilia 2003; Travis and Visher 2005; Nagin et al. 2009; Cullen et al.
2011).
Juxtaposed against these two perspectives—one arguing for tougher punishment and the
other for a balance of punishment and rehabilitation—is a paucity of credible empirical
evidence on the relative effectiveness of the central types of sanctions employed by most
states: probation, intensive probation, jail, and prison. Recent reviews and meta-analyses
have arrived at this same conclusion (Gendreau et al. 2000; Smith et al. 2002; McDougall
et al. 2003; Villettaz et al. 2006; Nagin et al. 2009; Durlauf and Nagin 2011; Jonson 2011).
Nagin et al.’s (2009) review of studies assessing the effect of imprisonment on reoffending
is illustrative. The authors identified few rigorous quasi-experimental studies of the relative
impact of custodial and non-custodial sanctions on offenders’ likelihood to reoffend, and
far fewer experimental studies. They also concluded that, among existing studies, mixed
evidence exists for the relative effectiveness of non-custodial versus custodial sanctions.
Specifically, some studies have identified criminogenic effects of custodial sanctions and
some have not. More importantly, as the authors emphasized, these studies typically have
suffered from methodological shortcomings, such as a failure to use matching designs and
other more rigorous methodological approaches, that render substantive conclusions
questionable.
A concern arising from such reviews is not just that little is known about the effect of
custodial versus non-custodial punishment. It also is that few studies have examined the
relative effectiveness of shorter versus longer terms of incarceration, of probation versus
intensive probation, and of these latter sanctions as alternatives to jail and prison sen-
tences. This research gap assumes particular importance given recent calls for using less
severe but potentially more certain sanctions because of the possibility that they can
reduce recidivism more so than incarcerative sanctions (see, e.g., McDougall et al. 2003;
MacKenzie 2006; Mears 2010; Durlauf and Nagin 2011; Cullen et al. 2011; Jonson
2011).
The purpose of this study is to respond to the calls by scholars for more rigorous
assessments of sanction effects. Using Florida Department of Corrections data on a cohort
of convicted felons, we use propensity score matching to examine the relative effectiveness
of four types of correctional sanctions: probation, intensive probation, jail, and prison. We
begin first by describing recent trends in corrections, what is known about the effectiveness
of correctional system sanctions, and the critical questions that remain unaddressed. We
then describe the data, methods, and findings, and conclude by discussing the study’s
implications.
318 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Background
Correctional Expansion and ‘‘Get-Tough’’ Punishment
Recent decades have been witness to historically unprecedented growth in the correctional
system (Gottschalk 2011). The US jail and prison population has grown from 501,886 in
1980 to over 2.2 million in 2010. Considerable scholarly and policy attention has been
given to this growth, yet the larger growth, in absolute magnitude, has occurred among the
population under non-custodial control. For example, during the same time period, the total
population on probation or parole grew from 1.3 to 4.8 million (Glaze 2011). Of the 7
million individuals under correctional system control in 2010, approximately one-third was
under some form of custodial control (11 % in jail and 21 % in prison) and over half
(57 %) was on probation.
Scholars have attributed the expansion of the correctional system to several factors.
They point, for example, to policymaker efforts to appear ‘‘tough on crime’’ and to be or
appear proactive in the fight against drugs and violence (Beckett 1997; Davey 1998;
Gottschalk 2006; Simon 2007; Mears 2010; Blumstein 2011). Others have suggested that
the increased use of sanctioning stems from an indirect response to perceived threats
among whites and the power elite from minorities, the poor, and other marginal groups
(Garland 2001; Beckett and Western 2001; Bobo and Thompson 2006). In part, the growth
may derive from disenchantment with rehabilitative approaches to crime control, as well as
dissatisfaction with the record of intermediate sanctions in reducing recidivism (Tonry and
Lynch 1996). This disenchantment in turn may have contributed to a belief that prison
terms, lengthy prison terms in particular, constitute the only viable way to reduce
offending. This belief—that more punitive sanctions reduce recidivism—provides the
central justification for many of the get-tough changes that have arisen in US sentencing
policies (Spelman 2000; Cullen et al. 2011).
The historically unprecedented change in the correctional system policy landscape has
led to considerable attention to investigating the effects of correctional system growth on
crime rates (e.g., Sampson 1986; Marvell and Moody 1994; Levitt 1996; Spelman 2000;
Kovandzic and Vieraitis 2005; Rosenfeld and Messner 2009). At the same time, there is,
however, the question of whether more severe sanctions reduce recidivism. Although
tougher punishment is motivated in part by retributive ideals, it also is motivated by a
belief that more severe sanctioning generates a specific deterrent effect. The underlying
theoretical premise is that such sanctions inspire a greater fear of further punishment and in
turn a greater likelihood of refraining from criminal behavior (Nagin et al. 2009). In
contrast to this view stands the theoretical argument that less severe sanctions can be more
effective. Among other things, they may allow for more rehabilitative services to be
provided and for ties to family and to the community to be maintained, thus not only
facilitating prosocial behavior but also enabling social support that can allow for successful
reentry (Braithwaite 1989; Lawrence 1991; Petersilia 1995; MacKenzie 2006; Pratt 2008;
Mears 2010). At the least, according to this argument, non-incarcerative sanctions avoid
the potentially criminogenic effects of incarceration (Nagin et al. 2009).
The Effectiveness of Correctional Sanctions
Given the growth in the US correctional system, the costs associated with such growth, the
seemingly compelling arguments for and against tougher sanctioning, and policymaker
calls for evidence-based policies (Welsh and Harris 2008; Mears 2010), it could reasonably
J Quant Criminol (2014) 30:317–347 319
123
Author's personal copy
be anticipated that a substantial body of rigorous empirical research has accumulated that
adjudicates between these different perspectives. As noted at the outset, however, few
studies exist that directly attend to this issue in ways that address a range of methodological
concerns, such as selection bias. In addition, the available evidence supports no clear or
consistent finding concerning the effectiveness of different types of sanctions, save to
suggest that more severe sanctioning exerts a null or criminogenic effect.
1
That assess-
ment, rendered most recently by Nagin et al. (2009) but also echoed by others (see,
generally, Gendreau et al. 2000; Smith et al. 2002; McDougall et al. 2003; Villettaz et al.
2006; Mears 2010; Durlauf and Nagin 2011; Jonson 2011), underscores the need for
studies that directly examine the relative effectiveness of different types of sanctions and
that do so using more rigorous research designs. As prior scholarship has emphasized, there
are two critical issues to address: identifying the relevant or appropriate counterfactual
condition and arriving at credible estimates of sanction impacts.
The concern about identifying the appropriate counterfactual condition derives from the
fact that any estimated impact of a sanction is relative to some other condition. In assessing
the impact of a prison term, for example, the relevant counterfactual is some other type of
sanction, such as jail or probation (Nagin et al. 2009:129; see, e.g., Smith and Akers 1993;
Bales and Piquero 2012) or length of time served (see, e.g., Loughran et al. 2009; Snod-
grass et al. 2011). Even so, a challenge here is that it is not always clear what sanction
constitutes the counterfactual condition. Among convicted felons, a sanction of some type
will occur. However, for a given group of sanctioned felons, it may not always be clear
what other sanction would have been administered. It is possible, for example, that pris-
oners would have been sent to jail, intensive probation, or even traditional probation. In a
context where multiple possibilities exist, it is important, as we discuss below, to assess
each of the different counterfactual conditions to arrive at estimates of the impacts of a
given sanction relative to the range of alternatives that might otherwise have occurred.
The concern about methodology stems from the idea that any assessment of a sanction’s
impact must take into account the fact that individuals typically are not randomly selected
into one type of sanction or another. In Nagin et al.’s (2009) review, the authors identified
three categories of prison recidivism studies: experimental, matching, and regression-
based. The authors noted that there were too few credible studies to draw firm conclusions
about the relative effectiveness of different sanctions. For example, they identified only 5
experimental studies of prison effects and only 12 studies that employed matching to
address selection biases (Caliendo and Kopeinig 2008; Guo and Fraser 2010). Finally, the
authors identified 31 regression-based studies of prison effects, which would seem to hold
promise for providing a robust estimate of sanction impacts. However, a critical limitation
of such studies is that they typically do not address selection bias as well as experimental or
matching designs (see, e.g., Chen and Shapiro 2007; Bales and Piquero 2012). In many of
the studies, the flaw was more fundamental—for example, of the 31 studies, only 16
controlled for age, race, sex, prior record, and offense type (Nagin et al. 2009:155).
1
There is debate about what constitutes more severe punishment. Some scholarship, for example, suggests
that offenders may perceive supervision-based sanctions as more severe than prison (Crouch 1993; Petersilia
and Deschenes 1994; Deschenes et al. 1995; Spelman 1995; Petersilia 1997; May et al. 2005). In general,
extant theory and research does not provide a clear answer (see, e.g., Paternoster 1987; Nagin 1998; Pratt
et al. 2006). Here, we recognize that although incarceration typically is viewed as a tougher sanction,
offenders’ perceptions of severity may vary depending on the conditions of incarceration and supervision.
320 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
What is the Relative Effectiveness of Correctional System Sanctions?
The purpose of this paper is to respond to calls for more methodologically rigorous
assessments of the relative effectiveness of correctional sanctions on recidivism (see, e.g.,
Smith et al. 2002; Chen and Shapiro 2007; Nagin et al. 2009; Bales and Piquero 2012). In
so doing, the paper aims to build on prior studies to examine heterogeneity within non-
custodial and custodial sanctions. Non-custodial sanctions can consist, broadly, of either
probation or intensive probation, while custodial sanctions can consist, broadly, of shorter-
term confinement in jail or longer-term confinement in prison. Accordingly, this study
examines the following question: What is the relative effectiveness of four different types
of sanctions—probation, intensive probation, jail, and prison—in reducing recidivism?
There can be, of course, heterogeneity within these broad categories of sanctioning. Such
variation itself bears investigation, but at the same time and as Nagin et al. (2009) have
highlighted, studies are needed that examine whether the general categories of sanctioning
most commonly used by the courts influence sanctioning.
Although several prior studies have examined the impact of incarceration on recidivism,
this study is, to our knowledge, the first to systematically investigate a series of different
counterfactual conditions specific to each of these sanctions and to do so using a meth-
odological approach, propensity score matching, called for in recent scholarship on
sanction effects (see, generally, Nagin et al. 2009; Bales and Piquero 2012). In particular,
we examine the following questions and the associated counterfactual conditions that they
involve. First, what is the effect of probation? More precisely, what is the effect of
probation as compared to what otherwise would have happened? The possibilities are that
the individuals instead would have been placed on intensive probation or in jail or prison.
Thus, to answer the question, we need to identify individuals from among each of the three
counterfactual conditions who resemble those who received probation. Implicitly, then, the
expectation is that there may be people in each of these three groups who have charac-
teristics similar to those of probationers. If in fact no matches exist, then it is not possible
to estimate a relative effect of probation. Second, if we view intensive probation as the
treatment, we want to know what the effect of this treatment is as compared to what
otherwise would have happened. Here, again, three possibilities present themselves—that
is, the individuals otherwise would have been placed on traditional probation or in jail or
prison. Third, for jail-as-treatment, the three possibilities are that the individuals otherwise
would have been placed on traditional probation or intensive probation or in prison,
respectively. Finally, if we view prison as the treatment, the three counterfactual conditions
are traditional probation, intensive probation, or jail, respectively.
A central implication that flows from identifying these different counterfactual condi-
tions is that the impact of a given sanction may vary depending on which counterfactual is
examined. From this perspective, there is no absolute effect of a given type of sanction.
Rather, its effect is always relative to the types of individuals who receive that sanction and
concomitantly to the types of particular sanctions that the individuals otherwise would
have received. For example, a study that examines the effects of imprisonment versus
intensive probation, in reality is assessing only one of several counterfactual conditions
relevant for determining the impact of imprisonment (see, e.g., Bales and Piquero 2012).
To illustrate the policy relevance of these observations and the salience of answering the
above questions, consider the case of a judge who must sentence a convicted felon. To
simplify matters, let us focus only on the individuals who the judge typically sentences to
prison. The judge may wonder if these individuals have lower levels of recidivism as
compared to what otherwise would have happened—that is, as compared to the sanctions
J Quant Criminol (2014) 30:317–347 321
123
Author's personal copy
that he or she otherwise typically would have administered. If the judge discovers that the
prison group had a higher level of recidivism as compared to what otherwise would have
happened, then he or she might consider a different approach to sanctioning the types of
individuals that, in the past, typically were sent to prison. Conversely, the judge may wonder
what would have happened to the individuals that he or she typically sanctions to intensive
probation. If the judge discovers that this group has a lower level of recidivism as compared to
individuals in two of the other sanction groups—for example, those in jail or prison but not
intensive probation—this finding might reinforce the judge’s view that it is the appropriate
intervention for the types of individuals who he or she typically places on intensive probation
instead of jail or prison. It also might raise questions about whether intensive probation, as a
sanction for the types of individuals who he or she otherwise would have sanctioned to
traditional probation, is worthwhile given that no difference in recidivism exists.
The implications of different counterfactual conditions associated with each of the four
types of sanctions bears emphasis—there is no fixed or absolute effect of a given sanction.
Rather, the effect of a given sanction on recidivism is always relative to what otherwise
would have happened. A central goal of this paper is to illustrate this point and, in
particular, to show that a rigorous assessment of sanction effects requires systematically
taking into account the sanctioning options that define the counterfactual universe of
options. In the absence of an experimental design, a series of counterfactual, matching-
based analyses provides one approach, among a range of approaches recommended by
scholars (see, e.g., Chen and Shapiro 2007; Nagin et al. 2009; Bales and Piquero 2012), to
address this complexity and arrive at more credible estimates of sanction effectiveness. A
related line of investigation involves assessing whether the theoretical underpinnings of
different sanctions contribute to identified effects. As Nagin et al. (2009) and Cullen et al.
(2011) have emphasized, however, any such undertaking requires first developing credible
assessments of sanction effects.
Data and Methods
The goal of the current study is to assess the relative effectiveness of four types of
sanctions: probation, intensive probation, jail, and prison. Using data on convicted felons
in Florida, the analyses employ propensity score matching to assess a series of counter-
factual scenarios, each comparing the recidivism of one group of felons to the recidivism
of other groups of felons who received a different sanction. Essentially, this approach asks
the following question: What is the effect of a given sanction, or ‘‘treatment,’’ as compared
to a particular counterfactual sanction? The data and methodology for answering this
question are described below.
Data
The data for this study came from the Florida Department of Corrections (FDOC) Sen-
tencing Guidelines database, and consist of a cohort of sentenced individuals who were
convicted of felonies in Florida and who were released between 1994 and 2002. All male
offenders in the guidelines database who were sentenced to probation, intensive probation,
jail, or prison were included in the dataset.
2
Also included are a series of demographic
2
In Florida, intensive probation is officially termed ‘‘community control.’’ It typically includes house arrest,
curfew, and contact restrictions greater than that of traditional probation.
322 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
characteristics and guidelines scores that describe each offender’s most serious offense and
whether or not an offender’s total sentencing score allowed for a prison sentence. The
guidelines records were then matched to information in the FDOC’s Offender Based
Information System to allow inclusion of prior record information, including prior prison
commitments, supervision violations, and felony convictions, and also to acquire infor-
mation on prison release dates. Recidivism was linked to each offender using the same
databases and is here defined as a felony reconviction within a 3-year follow-up period. For
prison and jail, the recidivism ‘‘clock’’ starts upon release from incarceration; for either
type of probation, it starts when the sentence begins. The dataset included 586,357 indi-
viduals in total.
Propensity Score Matching
The analyses here employed a matching methodology using propensity scores to reduce the
influence of sanction selection bias on estimated treatment effects (Rosenbaum and Rubin
1983; Becker and Ichino 2002; Apel and Sweeten 2010; Guo and Fraser 2010). As a
treatment, sanction types are not randomly assigned, and offenders are likely to vary
significantly on a number of characteristics related both to receiving a given sanction and
to reoffending. For example, prisoners are likely to differ, on average, from individuals
sentenced to probation, and in most cases this is by design. Prison is, for example, typically
considered to be a more severe sanction. Accordingly, individuals who are imprisoned
typically will differ from individuals who receive other sanctions. The record of prior
convictions, prior prison commitments, and the severity of their offense, for example, all
likely will be greater. That said, sentencing research highlights that there is considerable
heterogeneity in sanctioning (see, e.g., Reitz 2011), thus creating a situation in which
individuals who, as in this example, are sentenced to prison look similar in many respects
to individuals placed in jail or on traditional or intensive probation. The propensity score
matching technique is useful for reducing selection bias by matching groups of individuals
based on their propensity, or likelihood, of receiving various sanctions.
Analysis Plan
In instances when an experimental design is not feasible and offenders have not been
randomly selected into, in this case, one of the four sanction types, a matching procedure is
useful because it attempts to simulate independent treatment assignment (see, generally,
Rosenbaum and Rubin 1983; Apel and Sweeten 2010; Guo and Fraser 2010). Accordingly,
for this study, we undertook a three-step process involving matching analyses to produce
estimated average treatment effects on the treated (ATT). First, propensity scores were
created using logistic regression to predict the likelihood of individuals receiving a given
sanction relative to a given counterfactual condition, using matching information on
demographic characteristics, offense, and prior record. Second, for each comparison,
individuals from two different sanction groups—one designated to be the treatment and the
other the control—were matched using the estimated likelihood score. The matching
approach for all analyses involved 1-to-1 nearest neighbor matching without replacement
using a .005 caliper setting.
3
Under the propensity score framework, two individuals with
the same score have the same likelihood, based on the specified covariates, of receiving the
3
Ancillary analyses using replacement, 1-to-many matching, and various caliper specifications revealed
substantively similar findings. These results are available upon request.
J Quant Criminol (2014) 30:317–347 323
123
Author's personal copy
designated treatment. Once matched, these groups are now comparable, given the
assumption of no imbalance in unobserved confounders (Winship and Morgan 1999), and
differences between them can be more confidently attributed to the effect of the specified
treatment. Finally, the third step involved matching offenders from the designated treated
and control groups and—after excluding unmatched cases that were, in the terminology
of propensity score analysis, ‘‘off support’’—comparing the respective rates of recidivism
(for similar examples in the criminological literature, see King et al. 2007; Paternoster
and Brame 2008; Loughran et al. 2009; Bales and Piquero 2012). These steps were
followed for each of the four sanction groups and the respective counterfactual
comparisons.
The accuracy of the comparisons is based on two considerations. The first is the quality
of the matching variables. Nagin et al. (2009) suggest that, when assessing the effec-
tiveness of sanctions, it is critical to match, at a minimum, on the following characteristics:
race, gender, prior record, and offense information. Accordingly, we include measures that,
if omitted, might potentially bias the results (see Table 1). These consist of frequently
identified factors associated with sanctioning, including criminal record (a count variable
measuring number of prior convictions), age (continuous), race (Black, Hispanic, and
White dummy variables), current offense information (a dummy variable for each type of
offense, separated into 9 categories), and an indicator of the judicial circuit each offender
was sanctioned in (a dummy variable for each Florida judicial circuit, numbered 1 through
20). In addition, we include the number of prior prison commitments (count), prior
supervision violations (count), and two variables taken from the FDOC Sentencing
Guidelines data: a measure of offense severity (values =1–10, with 10 indicating the most
serious offense), and a binary measure of prison eligibility based on the offender’s total
sentencing score in accordance with the Florida sentencing guidelines (‘‘1’’ indicates an
offender’s sentencing score made them eligible for a prison sentence).
The second consideration is the ability to find matches to treatment group subjects. For
matching analyses, one ideally has a sufficient pool of potential comparison subjects to
ensure that individuals similar to those in the treatment group can be identified (Rosen-
baum and Rubin 1983). Here, we have access to information on 318,073 individuals on
traditional probation, 53,136 on intensive probation, 132,059 in jail, and 83,089 in prison.
To create a larger pool of comparison subjects for each treatment-to-control group
matching analysis, we created smaller treatment groups that nonetheless are substantially
larger than those used in many prior studies. For each sanction group (probation, intensive
probation, jail, and prison), we created treatment groups by randomly selecting 10,000
individuals. Then, we matched these individuals to individuals from a given comparison
pool of subjects. In each comparison, only the treated group is limited to 10,000 indi-
viduals. For example, 10,000 probationers were matched to individuals from the entire
pool of intensive probation individuals; then these same 10,000 individuals were matched
to individuals from the entire pool of individuals in jail; and, last, they were matched to
individuals from the entire pool of individuals in prison.
This process was repeated for each of the other three treatment groups along with two
additional comparisons when examining jail as a treatment alternative to prison. Because
jail typically involves a relatively short term of incarceration, it may be that the more
appropriate pool of subjects from which to make comparisons is not the full pool of
prisoners but rather those who serve shorter prison sentences. For this reason, in addition to
an analysis where jail inmates were matched to individuals from the general prison pop-
ulation, we conducted matching analyses that instead used individuals who served 1 year
324 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Table 1 Descriptive statistics
Probation Intensive prob. Jail Prison
Mean SD Mean SD Mean SD Mean SD
Reconviction, 3 years 0.31 0.46 0.33 0.47 0.43 0.50 0.47 0.50
Black 0.37 0.48 0.40 0.49 0.59 0.49 0.56 0.50
Hispanic 0.11 0.31 0.09 0.29 0.08 0.27 0.06 0.24
White 0.52 0.50 0.50 0.50 0.34 0.47 0.38 0.49
Age (years) 30.61 10.49 30.08 10.53 32.79 10.21 32.50 9.72
Offense-murder 0.00 0.04 0.01 0.08 0.00 0.02 0.02 0.12
Offense-sexual 0.02 0.13 0.05 0.21 0.00 0.05 0.04 0.20
Offense-robbery 0.02 0.12 0.04 0.20 0.01 0.12 0.09 0.28
Offense-other violent 0.15 0.36 0.18 0.38 0.08 0.27 0.16 0.37
Offense-burglary 0.11 0.31 0.15 0.35 0.08 0.27 0.19 0.39
Offense-property 0.23 0.42 0.15 0.36 0.18 0.39 0.13 0.34
Offense-weapons 0.04 0.19 0.04 0.19 0.03 0.16 0.04 0.19
Offense-drug 0.35 0.48 0.29 0.46 0.50 0.50 0.27 0.44
Offense-other 0.09 0.29 0.10 0.30 0.11 0.32 0.07 0.26
Prior convictions (#) 0.60 1.69 1.06 2.38 0.66 1.86 1.68 3.50
Prior prison com. (#) 0.23 0.74 0.43 0.99 0.57 1.16 1.28 1.61
Supervision viol. (#) 0.42 0.90 0.82 1.20 0.89 1.17 1.36 1.43
Off. seriousness (#) 3.75 1.79 4.67 2.02 3.46 1.51 5.36 1.92
Prison eligibility 0.14 0.34 0.40 0.49 0.15 0.35 0.78 0.41
Circuit 1 0.04 0.20 0.05 0.22 0.03 0.17 0.04 0.20
Circuit 2 0.02 0.14 0.02 0.12 0.01 0.07 0.03 0.16
Circuit 3 0.01 0.09 0.01 0.11 0.00 0.04 0.01 0.11
Circuit 4 0.04 0.20 0.04 0.20 0.12 0.32 0.07 0.25
Circuit 5 0.05 0.21 0.04 0.20 0.02 0.14 0.04 0.20
J Quant Criminol (2014) 30:317–347 325
123
Author's personal copy
Table 1 continued
Probation Intensive prob. Jail Prison
Mean SD Mean SD Mean SD Mean SD
Circuit 6 0.10 0.29 0.12 0.32 0.07 0.26 0.10 0.30
Circuit 7 0.04 0.19 0.04 0.20 0.04 0.19 0.04 0.21
Circuit 8 0.03 0.16 0.02 0.15 0.01 0.12 0.03 0.16
Circuit 9 0.08 0.27 0.06 0.23 0.10 0.30 0.07 0.25
Circuit 10 0.04 0.20 0.03 0.17 0.02 0.13 0.05 0.22
Circuit 11 0.10 0.30 0.08 0.28 0.24 0.43 0.08 0.26
Circuit 12 0.03 0.18 0.04 0.19 0.03 0.18 0.03 0.17
Circuit 13 0.09 0.29 0.16 0.36 0.05 0.21 0.09 0.28
Circuit 14 0.02 0.15 0.04 0.19 0.00 0.06 0.03 0.17
Circuit 15 0.05 0.21 0.02 0.15 0.12 0.32 0.05 0.21
Circuit 16 0.01 0.11 0.01 0.09 0.00 0.06 0.01 0.09
Circuit 17 0.13 0.34 0.12 0.33 0.09 0.29 0.14 0.35
Circuit 18 0.04 0.20 0.04 0.20 0.01 0.12 0.03 0.18
Circuit 19 0.03 0.17 0.02 0.13 0.02 0.14 0.03 0.18
Circuit 20 0.05 0.22 0.04 0.19 0.02 0.13 0.03 0.17
N 318,073 53,136 132,059 83,089
326 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
or less in prison and individuals who served 2 years or less in prison, respectively.
4
There
were, then, a total of 14 comparisons.
5
Findings
Post-matching Balance on Covariates
For each of the 14 total matching comparisons, we used logistic regression models to
predict an offender’s likelihood to receive the specified treatment (i.e., sanction) as
compared to the alternative sanction under consideration. Per the protocol recommended in
the propensity score literature (e.g., Rosenbaum and Rubin 1984; Becker and Ichino 2002),
interactional and polynomial specifications were introduced in the models to achieve
balance on the covariates.
6
The main goal of matching analyses is not these models but
rather the generation of scores that result in covariate balance between the treatment and
control groups and in turn the ability to draw more robust estimates of treatment effects.
That is, the goal is to ensure that any confoundedness of the included variables is ruled out
(Rosenbaum and Rubin 1983; Becker and Ichino 2002; Apel and Sweeten 2010).
Accordingly, we present post-test balance, or adjustment, statistics for each matching
scenario. These are presented in Table 2a–d, respectively, and include the post-matching
covariate means for the treated group and the matched group, the percent bias remaining,
the percent bias reduction achieved by matching, and the ttest values.
Prior to making comparisons between treatment and control groups, it is important to
eliminate any imbalance in the covariates. That is, after matching, there should be no
remaining significant differences in covariate means between the treatment and the control
group. Because of the large size of the samples, it is possible to identify statistically
significant differences even when no substantively significant imbalance exists. Thus, we
focus on both statistical significance and substantive significance when discussing post-
matching covariate balance. Table 2a–d present the balance statistics for all 14 compar-
isons. Table 2a focuses on probation as compared with intensive probation, jail, and prison,
respectively. Table 2b focuses on intensive probation as compared with traditional pro-
bation, jail, and prison, respectively. Table 2c focuses on jail as compared with probation,
4
Many inmates serve approximately a year in prison. For example, for Florida inmates released during the
years covered in this study, approximately 15–30 % of released inmates in a given year served a year or less
(see Table 4c, Time Served in DC Custody, Florida Department of Corrections Inmate Release reports—
http://www.dc.state.fl.us/pub/index.html). To illustrate, in 1994, over 28 % of inmates served one year or
less, and in 2002 almost 17 % did so. Nationally, the same pattern holds; for example, the median time
served among state prison inmates released in 2008 was 16 months (Bureau of Justice Statistics 2011).
5
These analyses differ from those that appear in an earlier study by Mears et al. (2012), which assessed
prison effects on recidivism, in several ways. There is no focus here on gender differences in the effects of
incarceration; we examine two groups of incarcerated prison inmates; we focus on the relative effects of four
types of sanctions to each other and not solely prison versus other sanctions; and we make no arguments
here about varying differences that the types of sanctions may exert on types of recidivism. Ancillary
analyses using the full samples (and thus 1-to-many matching analyses) and other treatment group sample
sizes identified results that were substantively and statistically similar; these analyses are available upon
request. Use of the sub-samples enables us to obtain estimates based on a more rigorous matching approach
(e.g., 1-to-1 matching and narrow caliper settings).
6
For all 14 models, the variables that typically predict sentences were statistically significant in the
expected directions. Because each model had a slightly different specification, there is no parsimonious way
to present the full set of regression results. They are available upon request.
J Quant Criminol (2014) 30:317–347 327
123
Author's personal copy
intensive probation, prison, prison for 1 year or less, and prison for 2 years or less,
respectively. Finally, Table 2d focuses on prison as compared with probation, intensive
probation, and jail, respectively.
Inspection of Table 1shows that substantial differences in many of the covariates
existed prior to matching. By contrast, inspection of Table 2a–d shows that the propensity
score matching process eliminated almost all statistically significant differences between
treated and control groups in the covariates. To illustrate, prior to matching, individuals in
the probation group had an average of .60 prior convictions and individuals in the intensive
probation group had an average of 1.06 prior convictions (see Table 1). By contrast, and as
shown in the first panel of Table 2a—the analysis in which traditional probation is the
treatment and this group is matched to individuals from the intensive probation pool—
there is no statistically significant difference; indeed, the mean prior record for both groups
is almost identical (.610 vs. .613).
The percent bias (%B) columns reinforce this assessment. For almost every covariate in
every comparison across Table 2a–d, the percent bias remaining after matching typically is
2 % or less. The percent bias reduction (%BR) column shows why. Across the different
covariates, the matching process generally reduced the pre-matching imbalance by 80 % or
more. In the end, then, only a few covariate comparisons remain statistically significant. In
these cases, the substantive differences are negligible. Consider, for example, Table 2a. Of
the three matching analyses—intensive probation in the first panel, jail in the second panel,
and prison in the third—only six statistically significant post-matching differences, out of
114 total comparisons, emerge. Closer inspection of the mean values in each of these six
cases identifies that the magnitude of the differences is trivial. In the panel 1 matching
analysis, for example, the percentages of individuals in the treated and matched groups,
respectively, who were convicted of murder are both less than 1 %; the prison eligibility
means for the two groups are within 1 % point of one another (.139 and .150, respectively);
and the percentages of each group tried in the 15th circuit also are substantively similar
(.044 and .052, respectively). Careful review of the other three statistically significant
differences (in panel 3) identifies no appreciable substantive differences in the covariates
after the matching analyses.
The same pattern can be seen in the other tables. In Table 2b, only 9 of the 114
comparisons are statistically significant, and in each of these 9 cases, there are no sub-
stantively significant differences that remain. In Table 2c, only 4 of 190 comparisons are
statistically significant and in each instance the substantive differences are slight. Finally,
in Table 2d, only 3 of the 114 comparisons are statistically significant; here, again, the
magnitude of difference in the mean values of the matching covariates is negligible. In
short, then, the propensity score matching resulted in treated and matched control groups
that are similar with respect to the matching covariates. This process thus enables the
estimation of effects that we can be more confident reflect the relative effectiveness of the
various sanctions rather than differences among the individuals who receive the different
types of sanctions.
We also examined sensitivity analyses for all of the reported results (Becker and Cal-
iendo 2007). These analyses provide conservative estimates of the degree to which the
results might be sensitive to unobserved covariates (DiPrete and Gangl 2004:291). Spe-
cifically, they estimate how large the effect of unobserved confounders would have to be to
alter the results. The size of this effect is expressed as gamma. For example, if gamma,
expressed as an odds ratio, is 2, the result is sensitive to bias that would double the
likelihood of receiving treatment. Across all 14 comparisons, the analyses yielded gamma
scores ranging from 1.1 to 1.8. As would be expected, larger gamma scores were
328 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Table 2 Adjustment balance statistics
Matching group 1 =intensive probation Matching group 2 =jail Matching group 3 =prison
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
(a) Treatment =probation
Black 0.371 0.376 -1.1 83.1 -0.81 0.374 0.372 0.6 98.7 0.41 0.444 0.431 2.8 92.8 1.54
Hispanic 0.106 0.106 -0.2 95.9 -0.14 0.108 0.100 2.7 71.7 1.84 0.085 0.090 -1.5 91.0 -0.85
White 0.524 0.518 1.2 65.9 0.87 0.518 0.529 -2.2 94.2 -1.52 0.470 0.479 -1.9 93.4 -1.05
Age 30.67 30.49 1.7 72.3 1.18 30.75 30.51 2.4 88.1 1.68 31.12 30.89 2.2 87.4 1.22
Offense-murder 0.001 0.002 -1.9 77.5 -1.98* 0.001 0.001 1.2 55.7 0.73 0.002 0.002 -0.7 95.6 -0.82
Offense-sexual 0.019 0.019 -0.1 99.3 -0.10 0.016 0.014 1.5 90.8 0.87 0.028 0.026 1.3 90.3 0.77
Offense-robbery 0.015 0.016 -0.4 97.5 -0.34 0.015 0.015 0.3 57.2 0.23 0.024 0.020 1.6 95.1 1.33
Offense-other viol 0.148 0.155 -1.9 77.0 -1.35 0.146 0.145 0.2 99.2 0.10 0.168 0.172 -1.3 63.4 -0.68
Offense-burglary 0.110 0.109 0.5 95.6 0.36 0.109 0.112 -1.1 89.1 -0.75 0.131 0.135 -1.2 94.5 -0.73
Offense-property 0.230 0.230 0.0 99.9 -0.02 0.235 0.237 -0.6 95.2 -0.42 0.219 0.214 1.4 95.0 0.71
Offense-weapons 0.037 0.037 -0.3 79.1 -0.19 0.037 0.035 0.9 82.6 0.61 0.040 0.037 1.8 -3967 0.97
Offense-drug 0.343 0.339 0.8 92.2 0.56 0.345 0.337 1.7 94.8 1.20 0.292 0.297 -1.2 92.7 -0.68
Offense-other 0.097 0.093 1.4 -40.6 0.97 0.097 0.104 -2.1 65.0 -1.54 0.098 0.097 0.3 95.6 0.18
Prior convictions 0.610 0.613 -0.2 99.2 -0.16 0.605 0.565 2.3 29.7 1.66 0.384 0.403 1.0 97.4 0.80
Prior prison comm. 0.247 0.249 -0.2 98.9 -0.19 0.246 0.241 0.5 98.5 0.45 0.639 0.636 -1.5 98.1 -1.14
Supervision violat. 0.433 0.424 0.9 97.7 0.75 0.431 0.414 1.6 96.4 1.38 4.207 4.170 0.2 99.7 0.14
Off. seriousness 3.743 3.764 -1.1 97.8 -0.81 3.708 3.668 2.4 84.7 1.62 0.216 0.241 2.0 97.7 1.13
Prison eligibility 0.139 0.150 -2.6 95.8 -2.21* 0.135 0.133 0.8 68.7 0.58 0.040 0.037 -6.5 96.1 -3.34*
Circuit 1 0.040 0.038 1.0 85.0 0.74 0.040 0.041 -0.8 84.7 -0.50 0.022 0.023 1.2 1.0 0.69
Circuit 2 0.019 0.020 -0.6 77.0 -0.41 0.019 0.019 0.5 96.3 0.26 0.011 0.010 -0.7 81.1 -0.42
Circuit 3 0.009 0.008 0.4 90.8 0.31 0.008 0.006 2.5 76.4 1.46 0.058 0.051 0.8 76.3 0.43
Circuit 4 0.038 0.039 -0.4 87.2 -0.26 0.038 0.041 -1.1 96.4 -1.02 0.044 0.043 3.3 76.8 1.84
Circuit 5 0.052 0.046 2.7 48.2 1.85 0.052 0.051 0.2 98.7 0.13 0.095 0.098 0.5 90.5 0.30
J Quant Criminol (2014) 30:317–347 329
123
Author's personal copy
Table 2 continued
Matching group 1 =intensive probation Matching group 2 =jail Matching group 3 =prison
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
Circuit 6 0.097 0.099 -0.4 94.5 -0.29 0.097 0.096 0.1 98.3 0.10 0.054 0.049 -1.2 30.9 -0.69
Circuit 7 0.040 0.039 0.4 60.7 0.29 0.040 0.040 0.3 76.4 0.18 0.026 0.030 2.5 -10.8 1.28
Circuit 8 0.028 0.027 0.1 92.6 0.09 0.027 0.025 1.5 81.9 0.98 0.078 0.073 -2.4 -143.1 -1.28
Circuit 9 0.076 0.077 -0.6 92.7 -0.37 0.076 0.080 -1.4 82.1 -1.03 0.053 0.047 1.8 17.5 1.01
Circuit 10 0.042 0.041 0.3 95.7 0.21 0.043 0.043 -0.4 97.5 -0.21 0.067 0.079 3.2 -6.9 1.71
Circuit 11 0.095 0.101 -1.9 59.3 -1.32 0.098 0.093 1.4 96.4 1.23 0.030 0.027 -4.3 42.9 -2.62*
Circuit 12 0.031 0.028 1.4 66.4 1.05 0.031 0.030 0.2 91.0 0.12 0.081 0.091 2.0 -1194 1.17
Circuit 13 0.093 0.096 -1.0 94.9 -0.78 0.093 0.093 -0.3 98.4 -0.17 0.033 0.031 -3.6 -149.0 -2.09*
Circuit 14 0.026 0.026 -0.1 98.1 -0.09 0.022 0.021 0.8 95.5 0.49 0.044 0.045 1.1 58.6 0.61
Circuit 15 0.044 0.052 -4.2 71.2 -2.54* 0.049 0.052 -0.7 97.1 -0.65 0.010 0.010 -0.8 -47.2 -0.47
Circuit 16 0.012 0.013 -0.5 87.2 -0.32 0.012 0.012 0.3 96.8 0.19 0.142 0.145 0.8 85.1 0.45
Circuit 17 0.133 0.128 1.5 31.3 1.04 0.132 0.132 -0.2 98.7 -0.10 0.037 0.036 -1.0 76.2 -0.53
Circuit 18 0.041 0.036 2.4 -185 1.75 0.040 0.038 1.0 93.7 0.59 0.043 0.038 0.4 88.3 0.24
Circuit 19 0.031 0.032 -0.5 95.0 -0.28 0.032 0.032 0.1 99.2 0.04 0.033 0.037 2.7 -140.1 1.40
Circuit 20 0.053 0.053 0.0 99.3 0.03 0.053 0.055 -1.2 93.8 -0.69 0.098 0.097 -1.9 83.0 -1.16
(b) Treatment =intensive probation
Black 0.409 0.411 -0.5 94.3 -0.35 0.424 0.420 0.9 97.4 0.62 0.424 0.431 -1.4 95.3 -0.98
Hispanic 0.090 0.088 0.7 89.8 0.50 0.089 0.086 1.2 69.5 0.78 0.086 0.087 -0.4 96.7 -0.23
White 0.501 0.501 0.1 98.2 0.06 0.486 0.494 -1.6 95.3 -1.05 0.490 0.482 1.6 93.4 1.10
Age 30.01 29.97 0.3 93.8 0.25 30.32 30.23 0.9 96.5 0.64 30.35 30.14 2.1 91.4 1.41
Offense-murder 0.005 0.005 1.6 78.3 0.91 0.004 0.003 1.8 81.0 1.17 0.006 0.004 1.5 84.7 1.45
Offense-sexual 0.047 0.018 0.5 97.0 0.30 0.025 0.026 -0.7 97.6 -0.46 0.050 0.043 3.2 -17.5 2.16*
Offense-robbery 0.039 0.039 -0.4 96.9 -0.26 0.036 0.040 -2.5 83.3 -1.45 0.041 0.037 1.6 91.8 1.36
Offense-other viol 0.175 0.189 -3.9 45.1 -2.66* 0.172 0.176 -1.2 95.9 -0.70 0.176 0.179 -0.9 79.4 -0.59
330 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Table 2 continued
Matching group 1 =intensive probation Matching group 2 =jail Matching group 3 =prison
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
Offense-burglary 0.143 0.145 -0.5 95.3 -0.34 0.138 0.132 1.9 91.0 1.16 0.148 0.149 -0.2 98.6 -0.12
Offense-property 0.150 0.147 0.6 97.0 0.50 0.159 0.161 -0.6 92.7 -0.44 0.150 0.152 -0.7 87.2 -0.47
Offense-weapons 0.041 0.039 0.9 54.4 0.61 0.043 0.043 -0.1 99.2 -0.04 0.040 0.040 -0.1 95.3 -0.07
Offense-drug 0.303 0.286 3.6 62.5 2.64* 0.321 0.301 4.2 89.8 2.94* 0.296 0.301 -1.2 84.2 -0.83
Offense-other 0.097 0.103 -2.1 -37.8 -1.44 0.102 0.117 -5.0 14.8 -3.33* 0.094 0.094 0.1 99.0 0.05
Prior convictions 1.021 0.984 1.9 91.2 1.21 1.023 0.998 1.2 93.1 0.76 1.062 1.058 0.1 99.4 0.10
Prior prison comm. 0.443 0.439 0.4 98.2 0.26 0.466 0.472 -0.5 95.9 -0.34 0.477 0.497 -1.5 97.5 -1.32
Supervision violat 0.849 0.858 -0.8 98.1 -0.47 0.890 0.910 -1.7 45.4 -1.18 0.885 0.916 -2.4 93.8 -1.73
Off. seriousness 4.632 4.670 -2.0 95.7 -1.33 4.461 4.434 1.6 97.6 1.02 4.730 4.694 1.9 95.0 1.25
Prison eligibility 0.391 0.388 0.6 99.0 0.38 0.354 0.350 1.0 98.2 0.63 0.420 0.420 0.0 100.0 0.01
Circuit 1 0.049 0.050 -0.6 82.3 -0.39 0.051 0.057 -3.3 65.4 -1.92 0.047 0.039 3.8 -16.8 2.63*
Circuit 2 0.014 0.015 -0.8 80.7 -0.60 0.013 0.014 -0.6 93.4 -0.32 0.015 0.015 -0.7 91.5 -0.54
Circuit 3 0.014 0.014 0.0 100.0 0.00 0.011 0.010 0.9 94.0 0.51 0.014 0.014 0.2 88.6 0.13
Circuit 4 0.043 0.046 -1.5 -0.5 -1.03 0.045 0.051 -2.0 92.6 -1.72 0.046 0.040 2.3 80.1 1.81
Circuit 5 0.044 0.043 0.7 66.0 0.49 0.045 0.044 0.6 95.3 0.35 0.045 0.043 1.1 42.9 0.75
Circuit 6 0.117 0.111 1.9 72.1 1.29 0.115 0.119 -1.4 90.8 -0.87 0.115 0.123 -2.7 46.1 -1.74
Circuit 7 0.043 0.044 -0.5 85.3 -0.31 0.046 0.045 0.4 84.5 0.28 0.046 0.047 -0.6 -13.0 -0.42
Circuit 8 0.023 0.026 -1.9 -39.2 -1.32 0.024 0.023 0.7 88.4 0.44 0.024 0.025 -0.3 81.7 -0.19
Circuit 9 0.055 0.060 -1.9 78.9 -1.46 0.058 0.059 -0.3 98.0 -0.25 0.057 0.055 0.8 86.5 0.57
Circuit 10 0.030 0.029 0.6 91.3 0.50 0.030 0.030 -0.4 94.7 -0.26 0.032 0.029 1.3 86.8 0.98
Circuit 11 0.082 0.083 -0.3 95.8 -0.21 0.088 0.077 3.1 92.9 2.78* 0.079 0.084 -2.0 16.1 -1.34
Circuit 12 0.038 0.037 0.7 80.9 0.44 0.040 0.042 -1.0 59.7 -0.63 0.038 0.036 0.6 85.5 0.43
Circuit 13 0.160 0.154 1.9 90.9 1.21 0.153 0.153 -0.3 99.3 -0.14 0.148 0.154 -1.9 91.5 -1.17
Circuit 14 0.038 0.038 0.2 97.1 0.15 0.031 0.030 0.3 98.8 0.17 0.038 0.036 1.2 72.9 0.78
J Quant Criminol (2014) 30:317–347 331
123
Author's personal copy
Table 2 continued
Matching group 1 =intensive probation Matching group 2 =jail Matching group 3 =prison
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
Circuit 15 0.020 0.021 -0.5 96.6 -0.45 0.022 0.020 0.6 98.5 0.71 0.021 0.021 0.4 97.7 0.31
Circuit 16 0.009 0.009 -0.2 94.8 -0.15 0.009 0.009 -0.1 98.1 -0.08 0.009 0.008 0.6 61.5 0.39
Circuit 17 0.122 0.122 0.0 99.0 0.02 0.124 0.122 0.8 91.6 0.54 0.129 0.139 -3.0 54.9 -2.07*
Circuit 18 0.044 0.045 -0.4 70.4 -0.31 0.041 0.039 1.5 91.7 0.86 0.044 0.043 0.3 95.0 0.18
Circuit 19 0.017 0.018 -0.7 91.6 -0.59 0.018 0.019 -1.0 49.8 -0.65 0.018 0.017 0.8 92.9 0.61
Circuit 20 0.038 0.036 0.7 88.9 0.52 0.038 0.038 0.0 100.0 0.00 0.036 0.029 3.6 8.8 2.47*
Matching group 1 =probation Matching group 2 =intensive probation Matching group 3 =prison
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
(c) Treatment =Jail
Black 0.585 0.591 -1.4 97.0 -0.95 0.548 0.545 0.6 98.3 0.39 0.588 0.594 -1.2 75.2 -0.76
Hispanic 0.077 0.070 2.4 78.3 1.90 0.085 0.077 2.8 49.3 1.85 0.065 0.062 1.2 82.0 0.74
White 0.338 0.339 -0.1 99.8 -0.06 0.367 0.378 -2.2 93.4 -1.45 0.347 0.343 0.7 92.4 0.40
Age 32.85 32.83 0.2 99.2 0.12 32.31 32.32 0.0 99.8 -0.03 32.57 32.52 0.6 83.8 0.35
Offense-murder 0.000 0.000 0.4 91.3 1.00 0.000 0.000 0.0 100.0 0.00 0.000 0.001 -0.5 97.3 -1.34
Offense-sexual 0.003 0.004 -1.0 93.2 -1.23 0.003 0.003 0.2 99.5 0.27 0.004 0.004 -0.1 99.6 -0.14
Offense-robbery 0.016 0.014 1.4 -3865 1.05 0.019 0.019 0.2 98.5 0.17 0.021 0.021 0.1 99.8 0.06
Offense-other viol 0.073 0.069 1.2 95.2 0.99 0.087 0.091 -1.1 96.4 -0.85 0.088 0.088 0.0 99.8 -0.03
Offense-burglary 0.084 0.081 0.9 89.0 0.67 0.093 0.087 1.9 90.4 1.36 0.103 0.098 1.6 95.0 1.07
Offense-property 0.180 0.176 1.1 91.7 0.80 0.195 0.195 -0.1 98.7 -0.06 0.208 0.206 0.4 97.2 0.20
Offense-weapons 0.027 0.026 0.7 86.8 0.57 0.032 0.033 -0.3 94.8 -0.22 0.030 0.028 1.3 74.1 0.84
Offense-drug 0.504 0.518 -2.8 91.2 -1.95 0.451 0.448 0.8 98.2 0.48 0.432 0.435 -0.6 98.9 -0.32
Offense-other 0.113 0.112 0.4 94.7 0.25 0.119 0.124 -1.8 60.2 -1.12 0.114 0.120 -2.0 84.7 -1.11
Prior convictions 0.702 0.657 2.5 55.2 1.68 0.783 0.794 -0.5 96.9 -0.35 0.846 0.867 -0.7 97.9 -0.58
332 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Table 2 continued
Matching group 1 =probation Matching group 2 =intensive probation Matching group 3 =prison
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
Prior prison comm. 0.557 0.535 2.2 93.4 1.32 0.564 0.539 2.3 81.4 1.38 0.717 0.753 -2.6 95.0 -1.81
Supervision violat 0.888 0.904 -1.5 96.7 -0.88 0.938 0.919 1.6 73.6 1.01 1.066 1.093 -2.1 94.1 -1.39
Off. seriousness 3.466 3.461 0.3 98.1 0.26 3.574 3.534 2.3 96.6 1.59 3.652 3.637 0.9 99.2 0.54
Prison eligibility 0.141 0.131 2.9 -187.1 2.05* 0.162 0.151 2.5 96.0 1.86 0.193 0.195 -0.4 99.7 -0.25
Circuit 1 0.030 0.026 2.3 63.3 1.84 0.036 0.034 1.1 90.0 0.80 0.038 0.036 0.8 87.2 0.48
Circuit 2 0.005 0.005 -0.3 97.8 -0.30 0.006 0.006 -0.1 98.9 -0.10 0.007 0.007 -0.6 96.6 -0.49
Circuit 3 0.001 0.002 -0.3 97.0 -0.37 0.002 0.002 -0.6 95.8 -0.71 0.002 0.001 0.8 93.4 1.04
Circuit 4 0.115 0.103 4.3 84.8 2.59 0.121 0.119 0.7 97.5 0.36 0.143 0.150 -2.3 85.4 -1.12
Circuit 5 0.022 0.024 -1.3 91.2 -1.08 0.026 0.027 -0.1 99.4 -0.05 0.029 0.028 0.6 94.7 0.35
Circuit 6 0.072 0.068 1.2 86.9 0.89 0.086 0.084 0.6 96.1 0.42 0.091 0.089 0.9 91.3 0.55
Circuit 7 0.037 0.038 -0.5 -919.2 -0.34 0.044 0.046 -0.6 75.2 -0.37 0.050 0.048 1.0 73.8 0.54
Circuit 8 0.013 0.012 0.7 92.2 0.65 0.015 0.014 0.6 92.9 0.45 0.017 0.015 1.3 86.3 0.85
Circuit 9 0.096 0.099 -1.3 79.8 -0.86 0.111 0.111 0.0 100.0 0.00 0.107 0.108 -0.3 96.9 -0.16
Circuit 10 0.016 0.017 -0.3 98.2 -0.28 0.019 0.018 0.6 93.2 0.40 0.022 0.021 0.5 97.5 0.34
Circuit 11 0.240 0.245 -1.4 96.4 -0.82 0.179 0.176 0.8 98.1 0.51 0.110 0.118 -2.2 95.4 -1.46
Circuit 12 0.034 0.034 0.5 60.7 0.35 0.041 0.043 -0.9 53.6 -0.54 0.040 0.039 0.5 77.0 0.30
Circuit 13 0.046 0.051 -2.0 89.1 -1.68 0.055 0.057 -0.4 98.8 -0.37 0.063 0.057 2.1 87.3 1.36
Circuit 14 0.004 0.004 0.2 99.0 0.23 0.005 0.005 -0.5 97.8 -0.66 0.005 0.006 -0.4 97.9 -0.44
Circuit 15 0.124 0.123 0.3 99.0 0.17 0.077 0.081 -1.4 96.4 -0.86 0.079 0.079 -0.2 99.1 -0.15
Circuit 16 0.004 0.005 -0.8 92.0 -0.77 0.005 0.004 0.6 90.0 0.47 0.005 0.005 0.6 88.3 0.35
Circuit 17 0.093 0.095 -0.7 94.7 -0.51 0.112 0.110 0.7 92.6 0.47 0.127 0.128 -0.5 97.1 -0.27
Circuit 18 0.014 0.014 -0.1 99.3 -0.12 0.017 0.022 -2.9 82.9 -2.19* 0.019 0.019 0.2 98.6 0.12
Circuit 19 0.020 0.021 -0.8 88.3 -0.60 0.024 0.022 1.8 -11.9 1.04 0.027 0.024 1.7 80.3 1.04
Circuit 20 0.015 0.014 0.2 99.2 0.18 0.017 0.019 -1.1 92.5 -0.87 0.019 0.020 -1.0 90.6 -0.66
J Quant Criminol (2014) 30:317–347 333
123
Author's personal copy
Table 2 continued
Matching Group 4 =Prison (B12 Mos.) Matching Group 5 =Prison (B12 24 Mos.)
Treated Matched %B %BR t-test Treated Matched %B %BR t-test
Black 0.584 0.594 -2.2 19.8 -1.23 0.586 0.596 -2.0 49.8 -1.20
Hispanic 0.061 0.058 1.3 87.4 0.72 0.064 0.059 2.0 75.2 1.22
White 0.355 0.348 1.6 79.5 0.90 0.349 0.344 1.0 87.5 0.62
Age 32.64 32.54 1.0 82.8 0.55 32.47 32.51 -0.4 94.7 -0.22
Offense-murder 0.000 0.000 -0.4 93.6 -0.58 0.000 0.000 -0.6 93.3 -1.00
Offense-sexual 0.004 0.005 -0.5 97.1 -0.40 0.004 0.004 -0.2 98.8 -0.27
Offense-robbery 0.024 0.021 1.3 93.5 0.85 0.022 0.022 -0.3 98.6 -0.23
Offense-other viol 0.094 0.090 1.2 94.2 0.72 0.090 0.086 1.2 95.1 0.77
Offense-burglary 0.108 0.111 -0.9 96.0 -0.49 0.100 0.104 -1.4 94.8 -0.89
Offense-property 0.209 0.205 1.2 45.5 0.64 0.210 0.202 2.2 69.2 1.21
Offense-weapons 0.032 0.029 1.7 58.9 0.94 0.029 0.031 -0.9 83.6 -0.54
Offense-drug 0.416 0.429 -2.7 92.3 -1.46 0.432 0.433 -0.2 99.5 -0.12
Offense-other 0.112 0.109 0.9 84.9 0.49 0.114 0.118 -1.2 84.6 -0.68
Prior convictions 0.896 0.903 -0.3 98.9 -0.16 0.859 0.867 -0.3 99.0 -0.22
Prior prison comm. 0.786 0.811 -1.8 96.1 -1.11 0.732 0.758 -1.9 96.2 -1.25
Supervision violat 1.134 1.148 -1.1 97.3 -0.66 1.085 1.099 -1.1 97.3 -0.70
Off. seriousness 3.772 3.771 0.1 99.8 0.07 3.668 3.657 0.7 99.2 0.42
Prison eligibility 0.215 0.232 -4.1 95.7 -2.35* 0.199 0.203 -1.0 99.3 -0.61
Circuit 1 0.035 0.033 1.1 15.9 0.55 0.038 0.037 0.5 87.8 0.27
Circuit 2 0.008 0.008 0.1 99.4 0.10 0.007 0.005 1.6 90.4 1.53
Circuit 3 0.002 0.002 0.2 97.9 0.19 0.002 0.001 0.8 93.3 0.82
Circuit 4 0.142 0.150 -2.8 76.0 -1.30 0.142 0.153 -3.8 75.2 -1.83
Circuit 5 0.033 0.032 0.3 97.1 0.15 0.030 0.031 -0.8 92.3 -0.49
Circuit 6 0.098 0.093 1.8 86.2 0.95 0.091 0.091 0.2 97.9 0.15
334 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Table 2 continued
Matching Group 4 =Prison (B12 Mos.) Matching Group 5 =Prison (B12 24 Mos.)
Treated Matched %B %BR t-test Treated Matched %B %BR t-test
Circuit 7 0.052 0.050 1.3 80.0 0.65 0.050 0.048 1.3 78.9 0.70
Circuit 8 0.019 0.020 -0.7 92.9 -0.39 0.017 0.016 0.8 91.5 0.53
Circuit 9 0.085 0.089 -1.7 88.7 -0.89 0.106 0.104 0.5 95.2 0.27
Circuit 10 0.026 0.025 0.6 97.0 0.40 0.023 0.024 -0.7 96.4 -0.50
Circuit 11 0.097 0.102 -1.4 97.3 -0.90 0.108 0.110 -0.8 98.5 -0.51
Circuit 12 0.035 0.036 -0.5 91.7 -0.24 0.040 0.037 1.9 42.0 1.01
Circuit 13 0.069 0.065 1.4 88.8 0.76 0.064 0.063 0.1 99.6 0.03
Circuit 14 0.006 0.005 1.4 90.7 1.09 0.006 0.006 -0.2 98.8 -0.22
Circuit 15 0.066 0.075 -3.3 87.9 -1.97* 0.074 0.079 -1.7 93.6 -1.05
Circuit 16 0.006 0.006 -0.7 85.1 -0.35 0.005 0.005 0.8 85.0 0.47
Circuit 17 0.148 0.141 2.0 93.3 1.10 0.131 0.128 0.9 95.9 0.50
Circuit 18 0.022 0.018 2.5 70.1 1.41 0.020 0.018 1.0 90.9 0.67
Circuit 19 0.032 0.028 2.2 78.4 1.22 0.028 0.024 2.5 73.8 1.53
Circuit 20 0.019 0.021 -1.0 86.4 -0.58 0.019 0.020 -0.3 97.0 -0.18
Matching group 1 =probation Matching group 2 =intensive probation Matching group 3 =jail
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
(d) Treatment =prison
Black 0.558 0.563 -0.9 97.7 -0.64 0.541 0.548 -1.4 95.6 -0.94 0.583 0.592 -1.7 67.8 -1.13
Hispanic 0.061 0.060 0.4 97.7 0.33 0.064 0.061 0.8 92.8 0.64 0.062 0.065 -1.2 83.5 -0.79
White 0.380 0.377 0.7 97.6 0.50 0.395 0.391 0.9 96.3 0.64 0.355 0.343 2.4 74.1 1.57
Age 32.49 32.60 -1.1 93.9 -0.83 32.11 32.16 -0.4 98.2 -0.30 32.59 32.51 0.8 72.4 0.53
Offense-murder 0.015 0.014 1.0 93.5 0.53 0.016 0.016 -0.7 91.8 -0.41 0.006 0.005 1.1 93.6 0.86
Offense-sexual 0.038 0.038 0.1 99.5 0.04 0.040 0.043 -1.4 71.2 -0.95 0.027 0.025 1.2 95.3 0.68
J Quant Criminol (2014) 30:317–347 335
123
Author's personal copy
Table 2 continued
Matching group 1 =probation Matching group 2 =intensive probation Matching group 3 =jail
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
Offense-robbery 0.087 0.078 4.4 86.7 2.47* 0.082 0.075 2.9 85.7 1.77 0.066 0.064 0.5 98.4 0.31
Offense-other viol 0.152 0.161 -2.6 -160.4 -1.77 0.160 0.161 -0.5 92.9 -0.32 0.160 0.161 -0.4 98.0 -0.25
Offense-burglary 0.195 0.189 1.8 92.7 1.15 0.194 0.189 1.4 89.0 0.93 0.176 0.173 0.9 97.5 0.51
Offense-property 0.134 0.138 -0.9 96.4 -0.72 0.133 0.137 -1.2 76.7 -0.81 0.148 0.149 -0.4 97.3 -0.24
Offense-weapons 0.036 0.038 -1.0 -41.6 -0.68 0.037 0.043 -3.0 -98.2 -2.01* 0.039 0.039 0.1 98.6 0.04
Offense-drug 0.268 0.268 0.0 100.0 0.00 0.263 0.263 0.0 99.2 0.03 0.300 0.294 1.2 97.5 0.81
Offense-other 0.074 0.076 -0.9 86.4 -0.70 0.075 0.073 0.9 90.0 0.67 0.080 0.089 -3.4 76.3 -2.30*
Prior convictions 1.715 1.706 0.4 99.2 0.19 1.648 1.715 -2.3 89.8 -1.38 1.500 1.478 0.8 97.9 0.45
Prior prison comm. 1.264 1.233 2.5 97.0 1.33 1.102 1.098 0.3 99.5 0.19 1.275 1.302 -1.9 96.2 -1.08
Supervision violat 1.374 1.407 -2.8 96.5 -1.56 1.289 1.319 -2.3 94.6 -1.40 1.399 1.419 -1.5 96.0 -0.90
Off. seriousness 5.363 5.342 1.1 98.7 0.77 5.376 5.348 1.4 96.0 0.97 5.063 5.037 1.5 98.7 0.94
Prison eligibility 0.784 0.789 -1.3 99.2 -0.86 0.769 0.777 -1.6 98.1 -1.19 0.745 0.754 -2.5 98.5 -1.42
Circuit 1 0.040 0.040 0.2 85.9 0.11 0.042 0.038 1.7 72.7 1.27 0.042 0.042 -0.3 95.0 -0.15
Circuit 2 0.024 0.023 1.0 73.2 0.70 0.024 0.025 -0.3 95.1 -0.19 0.021 0.021 0.5 96.9 0.27
Circuit 3 0.013 0.015 -1.7 67.2 -1.02 0.013 0.014 -0.8 -109.8 -0.51 0.007 0.007 0.1 99.0 0.09
Circuit 4 0.067 0.060 3.1 74.5 2.01 0.063 0.062 0.7 93.6 0.42 0.077 0.084 -2.4 86.1 -1.61
Circuit 5 0.042 0.045 -1.5 48.4 -1.08 0.043 0.043 -0.1 92.6 -0.04 0.038 0.035 1.6 86.9 0.94
Circuit 6 0.099 0.095 1.1 -31.1 0.79 0.102 0.110 -2.5 61.2 -1.71 0.101 0.094 2.6 72.0 1.58
Circuit 7 0.048 0.050 -0.7 86.7 -0.49 0.047 0.046 0.6 80.3 0.42 0.050 0.050 0.1 98.9 0.04
Circuit 8 0.029 0.032 -2.0 11.3 -1.31 0.030 0.030 -0.1 95.3 -0.09 0.029 0.028 0.6 94.1 0.32
Circuit 9 0.071 0.072 -0.5 83.5 -0.33 0.070 0.064 2.1 66.2 1.38 0.078 0.084 -2.0 79.5 -1.30
Circuit 10 0.046 0.046 -0.1 84.4 -0.10 0.044 0.042 0.8 91.4 0.50 0.034 0.034 0.5 96.6 0.34
Circuit 11 0.074 0.069 2.0 79.6 1.54 0.078 0.072 2.3 34.6 1.59 0.086 0.086 0.0 99.9 0.03
Circuit 12 0.028 0.029 -0.3 87.5 -0.21 0.029 0.030 -0.2 96.8 -0.13 0.030 0.027 1.8 46.6 1.19
336 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Table 2 continued
Matching group 1 =probation Matching group 2 =intensive probation Matching group 3 =jail
Treated Matched %B %BR t-test Treated Matched %B %BR t-test Treated Matched %B %BR t-test
Circuit 13 0.092 0.087 1.5 -473.2 1.04 0.097 0.103 -2.0 89.5 -1.51 0.090 0.089 0.2 98.7 0.13
Circuit 14 0.031 0.031 -0.4 88.8 -0.29 0.032 0.034 -1.2 63.6 -0.82 0.020 0.019 0.5 97.8 0.28
Circuit 15 0.048 0.050 -1.0 -111.8 -0.72 0.042 0.040 0.8 94.1 0.52 0.056 0.057 -0.4 98.5 -0.30
Circuit 16 0.007 0.008 -0.6 88.8 -0.49 0.008 0.007 0.8 47.9 0.61 0.006 0.007 -1.0 82.4 -0.57
Circuit 17 0.143 0.145 -0.5 84.2 -0.34 0.142 0.143 -0.2 96.1 -0.15 0.141 0.143 -0.6 96.3 -0.35
Circuit 18 0.033 0.035 -0.9 80.6 -0.63 0.034 0.034 -0.3 94.0 -0.20 0.027 0.025 1.6 87.1 0.96
Circuit 19 0.036 0.037 -0.6 81.2 -0.38 0.030 0.030 0.1 99.4 0.04 0.035 0.037 -1.3 86.5 -0.74
Circuit 20 0.031 0.033 -1.3 87.1 -1.00 0.032 0.034 -0.9 78.1 -0.65 0.031 0.032 -0.5 94.1 -0.31
%B =percent bias; %BR =percent bias reduction
*p\.05
J Quant Criminol (2014) 30:317–347 337
123
Author's personal copy
associated with substantively larger effects. In these cases, the effects were more robust
and less vulnerable to unobserved confounding. The robustness of these larger effects,
coupled with the consistent pattern in the estimated effects, discussed below, provides
greater confidence in the assessment of the relative effectiveness of the different sanctions.
Traditional Probation
We turn now to the results of the matching comparisons, presented in Table 3. For the
probation, intensive probation, jail, and prison ‘‘treatment’’ groups, it was possible to
match substantial proportions of the individuals to counterparts in the control groups. We
focus here on the first panel, which presents the matching analyses for the traditional
probation group. Before discussing the results, there is the critical question of whether
matches to this group can be found. Inspection of the last column in Table 3provides this
information. Specifically, we were able to match 99 % of the treatment group (n =10,000)
to individuals in the intensive probation and jail populations, respectively. When matched
to the prison population, approximately 36 % of probationers fell off support; this indicates
that it was more difficult to find comparable matches for the probationers among those
sanctioned to prison. On the one hand, this loss of cases to off support limits the gener-
alizability of an estimated effect of probation versus prison. On the other hand, it highlights
that in fact large numbers of individuals in prison have counterparts with similar profiles
who received probation instead of prison. In short, the fact that probation counterparts can
be found in prison suggests that prison is used as a sanction for individuals who, in many
respects, appear to be similar to individuals placed on probation.
The question addressed in the first panel is as follows: Among individuals on probation,
what is the effect of probation as compared to what would have happened had they not
received this sanction? Focusing first on the comparison to intensive probation, the anal-
yses show that, in absolute terms, 31 % of the ‘‘treated’’ individuals—that is, those on
probation—recidivated as compared to 34 % for the matched individuals on intensive
probation (.310 vs. .319 respectively, or a -.029 difference, statistically significant at the
p\.001 level, the level of statistical significance used for all of the analyses). This
difference is only slightly more pronounced if being placed in jail defines the counter-
factual. Here, again, those on probation are less likely to recidivate. Whereas 31 % of
individuals in the ‘‘treated’’ (probation) group recidivated, 35 % of the jail group recidi-
vated (.311 vs. .347, or a -.036 difference).
The more appreciable difference surfaces when prison is the counterfactual condition.
Here, we see that 33 % of probationers recidivated as compared to 47 % of ex-prisoners
(.328 vs. .469, or a -.141 difference). This effect reflects the expected difference in
recidivism only among those prisoners who resembled individuals in the probation sample.
Many prisoners had no counterparts on probation, which partially restricts the generaliz-
ability of the assessment. Notably, though, 64 % of the probation sample could be matched
to prisoners. For these individuals who do have counterparts in prison, placement on
probation appears to be associated with a substantially lower likelihood of recidivism.
Intensive Probation
What about when intensive probation is the treatment? Here, paralleling the steps taken
above, we first need to find matches for each of three distinct counterfactual conditions. As
with the probation analyses above, matching was not a problem. Specifically, for each of
the three comparisons, over 90 % of the treatment group could be matched to the control
338 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Table 3 Sanction effects on recidivism: an assessment using propensity score matching and average effect on the treated (ATT) estimates
Treated Matched Difference SE t-test % Off support
Treatment =probation vs.
Intensive probation 0.310 0.339 -0.029* 0.007 -4.300 1.2
Jail 0.311 0.347 -0.036* 0.007 -5.340 0.8
Prison 0.328 0.469 -0.141* 0.009 -16.400 36.4
Treatment =intensive probation vs.
Probation 0.336 0.332 0.004 0.007 0.550 0.0
Jail 0.346 0.375 -0.029* 0.007 -4.180 6.8
Prison 0.332 0.453 -0.121* 0.007 -17.070 7.0
Treatment =jail vs.
Probation 0.430 0.398 0.032* 0.007 4.550 0.1
Intensive probation 0.422 0.395 0.027* 0.008 3.550 17.0
Prison (all) 0.447 0.538 -0.091* 0.008 -11.060 27.1
Prison (12 mos. or less) 0.450 0.551 -0.101* 0.009 -11.280 38.2
Prison (24 mos. or less) 0.447 0.550 -0.103* 0.008 -12.280 29.4
Treatment =prison vs.
Probation 0.474 0.370 0.104* 0.007 15.000 0.3
Intensive probation 0.467 0.329 0.138* 0.007 19.440 6.7
Jail 0.498 0.441 0.058* 0.008 7.510 15.6
*p\.001
J Quant Criminol (2014) 30:317–347 339
123
Author's personal copy
group. Thus, it again appears that we have evidence that the policy question at hand is far
from academic. That is, many individuals who receive traditional probation, jail, or prison
sanctions in fact appear to be similar to individuals who receive intensive probation as a
sanction.
We turn now to the second panel of Table 3. When compared to traditional probation,
what in many cases would be viewed as a less serious sanction, we find no significant
difference in the effectiveness of intensive probation in reducing recidivism. This null
effect is notable, given the greater supervision associated with intensive probation, and, in
turn, the greater costs (Smith and Akers 1993; Piehl and LoBuglio 2005). A different
pattern surfaces when we turn to custodial sanctions. Compared to jail or prison, intensive
probation is associated with a reduced likelihood of recidivism, a finding that parallels the
first set of analyses that centered on traditional probation. For individuals on intensive
probation as compared to matched counterparts in jail, recidivism is slightly lower (.346 vs.
.375, respectively, or a -.029 difference). As with the traditional probation analyses
presented in the first panel, this recidivism-reducing effect is considerably more pro-
nounced when the comparison is with prison. Specifically, the estimated recidivism for the
probationers is 33 % rather than 45 %, what amounts to a 12 % reduction in recidivism in
absolute percentage terms (.332 vs. 453, respectively, or a -12.1 difference).
Jail
With the focus on jails, we now turn our attention to the effects of custodial sanctions.
What, in particular, is the effect of jail? We begin first with examining the extent to which
matches to the jail group could be obtained for four different counterfactual groups (tra-
ditional probation, intensive probation, prison, \1 year in prison, \2 years in prison).
Although almost all individuals in the jail sample could be matched to individuals on
traditional probation, approximately 17 % of the sample was off support when matching to
intensive probation. That is, it was more possible to identify matches among individuals on
traditional probation than it was among individuals on intensive probation. Finding mat-
ches to the prison population was more difficult. Among individuals in jail, 27 % could not
be matched to individuals from the prison population. Surprisingly, when the prison control
group was limited to just those inmates who served 1 year or less, 38 % of jailed offenders
were off support. When we focused on individuals who served 2 years or less, 29 % were
off support. The fact that matches could be identified at all indicates that probation and
prison terms are used for individuals who look, in many respects, comparable to indi-
viduals who received jail as a sanction. At the same time, the loss of some cases to off
support limits the generalizability of the estimated effects, which apply only to compari-
sons between the jail population and the types of individuals in these other groups who
could be matched on offense type, prior record, and the other measures.
What, then, is the relative effect of jail? Consistent with the previous two panels, a clear
pattern is present—tougher sanctioning, jail in this instance, is associated with increased
recidivism. Among individuals who received a jail sanction, as compared to those who
were placed on traditional probation, the likelihood of recidivism is modestly increased,
from 40 to 43 % (.398 vs. .430, or a ?.032 difference). As compared to matched indi-
viduals on intensive probation, the likelihood of recidivism among those who were placed
in jail is also slightly increased from 40 to 42 % (.395 vs. .422, or a ?.027 difference). The
pattern is evident, too, when prison, arguably a more severe sanction, serves as the
counterfactual. Here, the recidivism of the treated jail sample is 45 % compared to the
54 % recidivism of the matched prison group (.447 vs. .538, respectively, or a -.091
340 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
difference). This difference essentially is the same when the comparison is to prisoners
who served 1 year or less in prison or 2 years or less in prison.
Prison
The effect of prison has perhaps received the most attention in the sanctions literature, but
its impact has not been systematically compared to the full spectrum of alternative sanc-
tions—for example, probation, intensive probation, and jail. As with the preceding anal-
yses, the initial question is how comparable the prisoners are to individuals in the other
sanction groups. As can be seen in the table, almost all prisoners (99 %) could be matched
to individuals on traditional probation and 93 % could be matched to individuals on
intensive probation. Thus, few prisoners were off support. The matching was slightly more
limited for the jail population. In that analysis, 84 % of the prison sample could be
matched to individuals in jail.
The comparisons in the final panel reinforce the notion that sanctions typically viewed
as more severe are associated with increased recidivism. For example, placement in prison,
as compared to traditional probation, is associated with an increase in recidivism, from
37 % to 47 % (.370 vs. .474, respectively, or a ?.104 difference). A somewhat greater
increase can be seen when the comparison is to intensive probation—here, the increase is
from 33 to 47 % (.329 vs. .467 respectively, or a ?.138 difference). And prison appears,
not least, to be a more criminogenic alternative to jail. Here, the effect is not as pronounced
but nonetheless is notable. Specifically, when the counterfactual condition is jail, the
estimated effect of a prison sanction is to increase recidivism from 44 to 50 % (.441 vs.
.498, respectively, or a ?.058 difference).
Conclusion
In the United States and in many other parts of the world, a dramatic increase in more
punitive sanctioning occurred in recent decades, driven in no small part by the view that
tougher sanctions ‘‘work’’—that is, they reduce recidivism and they reduce crime rates.
This ‘‘get-tough’’ trend has been challenged by critics who claim that tougher sanctioning
does not produce these benefits and, at the same time, carries with it substantial costs in the
form of increased recidivism and missed opportunities to invest in potentially more
effective approaches to reducing crime (see, e.g., McDougall et al. 2003; Raphael and Stoll
2009; Mears 2010; Cullen et al. 2011). In support of such arguments are those who have
argued that more certain sanctioning, coupled perhaps with a range of services and sup-
ports, may do more to reduce the offending of individuals who enter the criminal justice
system (see, e.g., MacKenzie 2006; Pratt 2008; Durlauf and Nagin 2011). Notably, how-
ever, as Nagin et al. (2009) and others have shown, few rigorous studies of the relative
effectiveness of correctional system sanctions exist.
The goal of this study was to contribute to efforts to address this research gap. To this
end, it is the first study of which we are aware to systematically compare the effectiveness
of four commonly used types of sanctions—traditional probation, intensive probation, jail,
and prison—relative to the unique counterfactual conditions specific to each. The latter
emphasis is especially important because the effectiveness of a given sanction funda-
mentally depends on the basis of comparison. The effect of intensive probation, for
example, may be different if the counterfactual condition is probation or jail or prison.
J Quant Criminol (2014) 30:317–347 341
123
Author's personal copy
The matching analyses here proceeded from that premise and estimated the effects of a
given sanction in comparison to all possible counterfactual conditions. Two broad findings
emerged. First, we found that across most comparisons, tougher sanctioning was consis-
tently and positively associated with recidivism. Second, it was possible to find matches
between groups, suggesting that, at least in Florida and presumably in other states, a
naturally occurring experiment of sorts has been unfolding. That is, convicted felons who
resemble one another with respect to offense type, prior record, and other such charac-
teristics receive very different sanctions.
These findings are qualified by the fact that unobserved confounding might have
influenced the estimated effects, a problem central to all quasi-experimental assessments of
sanctioning impacts (Nagin et al. 2009), and by a loss of cases ‘‘off support’’ in some of the
analyses. In this latter instance, the generalizability of the estimated effects from the
analyses is limited to those cases for which matches could be identified. This limitation is,
however, precisely what the matching analyses highlight more clearly than traditional
regression analyses—that is, the effects of a given sanction, as compared to another
sanction, should only be expected for individuals who resemble one another. The research
design here responded to calls for using matching analyses to estimate more credibly the
effectiveness of sanctions. More research will be needed, however, before a strong claim
can be made concerning the effects of different sanctions on recidivism.
We turn now to several possibilities suggested by prior theory and research that may
account for the finding that sanctions that typically are viewed as tougher are associated
with more rather than less recidivism. First, it may be that tougher sanctions provide less
support and fewer services. By contrast, opportunities for providing more support and
services may be available with less severe sanctions. For example, being placed on pro-
bation or in a local jail may more readily allow for community-based reintegration and
treatment. In turn, these effects may translate into reduced offending. Some research, for
example, indicates that rehabilitative services, treatment, and community support and
assistance can contribute to lower levels of offending (see, e.g., Lawrence 1991; Petersilia
and Turner 1993; Petersilia 1995; Cullen and Gendreau 2000; MacKenzie 2006; Mears
2010; White et al. 2012).
Another possibility is that less severe sanctions reduce exposure to potentially crimi-
nogenic environments (Nagin et al. 2009). Jails and prisons, for example, are settings in
which substantial deprivations can occur and in which cultures of violence and criminality
may exist (Adams 1992; Bottoms 1999). Exposure to such conditions, and to the conse-
quences that may attend to incarceration (e.g., an even greater reduction in the ability to
find employment or housing), may increase the likelihood of recidivism. Conversely, a lack
of exposure to them may reduce recidivism even in the absence of rehabilitative pro-
gramming or various social supports that may be available while on probation.
Yet another possibility is that less severe sanctions are associated with increased per-
ceptions of punishment certainty among convicted felons who experience them. That is,
these individuals may perceive there to be a greater certainty that, if they commit an
offense, they will be sanctioned. Such a possibility would generate a reduced likelihood of
sanctioning if, as recent scholarship suggests, it is the certainty of punishment more so than
the severity of punishment that exerts a specific deterrent effect (Durlauf and Nagin 2011).
Not least, an intriguing possibility is that sanctions, such as probation, that typically are
viewed as less severe than other sanctions, such as prison, in fact may be perceived to be
more severe. Although seemingly counter-intuitive, this finding has emerged in several
studies (see, e.g., Crouch 1993; Deschenes et al. 1995; Petersilia 1997). Research suggests,
for example, that for some offenders a prison sentence may be preferable to a community
342 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
sanction, primarily because of perceptions that prison is less severe than supervision and
involves fewer restrictions upon release back into society (Spelman 1995; May et al. 2005).
In short, community sanctions, including jail, that provide access to the community and
links to a variety of potential supports, may be more effective than prison in reducing
recidivism because they may better allow both for more punishment and for more reha-
bilitation. That expectation certainly accords with the arguments made for reintegrative
approaches to punishment (Braithwaite 1989; Lawrence 1991). What the precise balance of
punishment and such services needs to be, under such arguments, remains largely
unknown.
Given prior research and this study, there are, in our view, several implications for
future scholarship and policy discussions. First, studies are needed that, as Nagin et al.
(2009) and others have advocated, employ more rigorous approaches to estimating the
effects of various sanctions on recidivism. In so doing, they will ideally want to examine a
range of counterfactual conditions. Comparing incarcerative and non-incarcerative sanc-
tions likely obscures important variation within these categories. Here, for example, there
were clear differences in the effects of each of the four different types of sanctions as
compared to the others. This study was not able to investigate further heterogeneity in
sanctioning, but such a step would be justified. Traditional probation is illustrative. In some
cases, traditional probation may consist of only a few contacts with an officer, while in
others it might involve more contact and supervision and also a strong emphasis on
facilitating access to social services and supports (Petersilia 2003; Piehl and LoBuglio
2005). In addition, community sanctions will not always provide better conditions than a
prison setting, so what about instances when prisons provide higher quality services and
features? This type of heterogeneity is typical of correctional system sanctions (see Chen
and Shapiro 2007; Bonta et al. 2008; Jonson 2011; Cullen et al. 2011; Durlauf and Nagin
2011; Listwan et al. 2011) and so constitutes an important avenue of research.
In a related vein, it will be important for future research to investigate the extent to
which incarceration effectiveness, relative to other types of sanctions, is moderated by the
quantity and quality of post-release supervision. Such work will need to confront the
challenge of specifying appropriate counterfactuals. For example, for ex-prisoners released
to lengthy terms of intense post-release supervision, the appropriate matches from the
probation pool might be those individuals on probation with comparable periods and
amounts of supervision. However, to the extent that ex-prisoner post-release supervision
derives in part from in-prison behavior, this approach would not necessarily result in
equivalent groups.
Second, studies are needed that investigate the extent to which a given type of sanction
may exert a differential effect for different groups of individuals. It is, for example,
possible that the effects of a particular type of sanction may vary along such dimensions as
age, race or ethnicity, gender, prior prison experience, offense type, and the community
context from which individuals come or to which they return (Clear and Hardyman 1990;
Spelman 1995; Bonta et al. 2000; Kubrin and Stewart 2006; Hipp et al. 2010).
Third, research is needed that identifies and assesses empirically the theoretical
mechanisms that would lead less severe sanctions to be associated with less recidivism. Is
it, for example, the avoidance of criminogenic conditions in prison, the greater access to
rehabilitative services and supports, the perception that community supervision sanctions
and consequences are more certain or severe, or some other mechanism (Nagin et al.
2009)?
Fourth, future research ideally will continue to employ quasi-experimental designs
aimed at estimating sanction effects. Experiments allow for greater internal validity—that
J Quant Criminol (2014) 30:317–347 343
123
Author's personal copy
is, we can trust more that a given sanction produced a particular effect. At the same time,
they typically do not allow for investigating the more nuanced ways in which sanctioning
occurs (Mears 2010; Sampson 2010). One example is the fact that a given sanction can
serve two purposes. Intensive probation might be viewed as an alternative to traditional
probation. Here, then, we need to devise a study—experimental or otherwise—that
addresses this particular use. It may also be viewed, however, as an alternative to jail or
prison. Once, again, this use requires its own study design. What would not be appropriate,
or sufficiently nuanced in an assessment of the effectiveness of intensive probation, would
be to limit our focus to the one use or the other. As suggested by the present study, the
effectiveness of a given sanction may well be relative to its particular use. In addition, it
may be relative to the populations to which, and settings in which, the given sanctions
occur. These are issues that future research ideally will investigate.
Fifth, studies should consider other effects of sanctions. The current study focused
exclusively on recidivism. There are, however, other dimensions along which to evaluate
the effectiveness of sanctions. One dimension consists of whether sanctions are used for
the populations for which they are intended. As this study’s results show, there are many
individuals in each of the four major sanction groups who greatly resemble one another
with respect to such dimensions as offense type and prior record. That does not mean that
the sanctions have been used inappropriately. However, it does raise questions about
whether they are. Given the calls for greater federal and state government accountability,
these types of assessments may facilitate efforts to show that sanctions are used in the
manner in which they are intended.
There is, of course, also the important task of identifying the effects of various sanc-
tioning regimes on crime rates and other outcomes. For example, there is a need to identify
the extent to which various sanctions result in the types and amount of retribution that are
intrinsic to the sanctioning process, how these sanctions affect families and communities,
how they affect racial or ethnic groups in a differential way, and how cost-effective
different sanctions may be (Western 2007; Gottschalk 2011; Tonry 2011; Austin 2011).
Such assessments are difficult to make, but nonetheless are critical for balanced assess-
ments of policy effectiveness (Mears 2010).
Finally, a straight-forward policy implication stems from this study and that of several
recent reviews (e.g., Nagin et al. 2009; Cullen et al. 2011)—specifically, greater reductions
in recidivism may be obtained through the use of less severe sanctions. As the above
discussion highlights, there are many considerations other than recidivism to consider, and
more research unequivocally is needed. Yet, as states deliberate on how best to allocate
scarce resources, they may well want to revisit assumptions about the benefits of tougher
sanctioning.
Acknowledgments We thank Peter Austin, Sam Field, and Brian Stults for their helpful comments and
suggestions during the development of this paper. We also thank the Florida Department of Corrections for
permission to use their data. The views expressed here are those of the authors and do not reflect those of the
Department of Corrections.
References
Adams K (1992) Adjusting to prison life. In: Tonry MH (ed) Crime and justice. University of Chicago Press,
Chicago, pp 275–359
Apel RJ, Sweeten G (2010) Propensity score matching in criminology and criminal justice. In: Piquero AR,
Weisburd D (eds) The handbook of quantitative criminology. Springer, New York, pp 543–562
344 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Austin J (2011) Making imprisonment unprofitable. Criminol Public Policy 10:629–635
Bales WD, Piquero AR (2012) Assessing the impact of imprisonment on recidivism. J Exp Criminol
8:71–101
Becker SO, Caliendo M (2007) Sensitivity analysis for average treatment effects. Stata J 7:71–83
Becker S, Ichino Andrea (2002) Estimation of average treatment effects based on propensity score. Stata J
2:358–377
Beckett K (1997) Making crime pay: law and order in contemporary American politics. Oxford University
Press, New York
Beckett K, Western B (2001) Governing social marginality: welfare, incarceration, and the transformation of
state policy. Punishm Soc 3:43–59
Blumstein A (2011) Bringing down the US prison population. Prison J 91:12S–26S
Bobo LD, Thompson V (2006) Unfair by design: the war on drugs, race, and the legitimacy of the criminal
justice system. Soc Res 73:445–472
Bonta J, Wallace-Capretta S, Rooney J (2000) A quasi-experimental evaluation of an intensive rehabilitation
supervision program. Crim Justice Behav 27:312–329
Bonta J, Rugge T, Scott T-L, Bourgon G, Yessine AK (2008) Exploring the black box of community
supervision. J Offender Rehabil 47:248–270
Bottoms A (1999) Interpersonal violence and social order in prisons. In: Tonry MH, Petersilia J (eds)
Prisons. University of Chicago Press, Chicago, pp 205–281
Braithwaite J (1989) Crime, shame, and reintegration. Cambridge University Press, Cambridge
Bureau of Justice Statistics (2011) Table 8. State prison releases, 2008: time served in prison, by offense and
release type. Bureau of Justice Statistics, Washington, DC
Caliendo M, Kopeinig S (2008) Some practical guidance for the implementation of propensity score
matching. J Econ Surv 22:31–72
Chen MK, Shapiro JM (2007) Do harsher prison conditions reduce recidivism? A discontinuity-based
approach. Am Law Econ Rev 9:1–29
Clear TR, Hardyman PL (1990) The new intensive supervision movement. Crime Delinquency 36:42–60
Crouch BM (1993) Is incarceration really worse? Analysis of offenders’ preferences for prison over pro-
bation. Justice Q 10:67–88
Cullen FT, Gendreau P (2000) Assessing correctional rehabilitation: policy, practice, and prospects. In:
Horney J (ed) Policies, processes, and decisions of the criminal justice system. National Institute of
Justice, Washington, DC, pp 109–175
Cullen FT, Jonson CL, Nagin DS (2011) Prisons do not reduce recidivism: the high cost of ignoring science.
Prison J 91:48S–65S
Davey JD (1998) The politics of prison expansion: winning elections by waging war on crime. Praeger,
Westport
Deschenes E, Piper S, Turner JP (1995) Intensive community supervision in Minnesota: a dual experiment
in prison diversion and enhanced supervised release. RAND Corp, Santa Monica
DiPrete TA, Gangl M (2004) Assessing bias in the estimation of causal effects: Rosenbaum bounds on
matching estimators and instrumental variables estimation with imperfect instruments. Sociol Meth-
odol 34:271–310
Durlauf SN, Nagin DS (2011) Imprisonment and crime: can both be reduced? Criminol Public Policy
10:13–54
Garland D (ed) (2001) Mass imprisonment: social causes and consequences. Sage, Thousand Oaks
Gendreau P, Goggin C, Cullen FT, Andrews DA (2000) The effects of community sanctions and incar-
ceration on recidivism. Forum Correct Res 12:10–13
Glaze LE (2011) Correctional populations in the United States, 2010. Bureau of Justice Statistics, US
Department of Justice, Washington, DC
Gottschalk M (2006) The prison and the gallows: the politics of mass incarceration in America. Cambridge
University Press, New York
Gottschalk M (2011) The past, present, and future of mass incarceration in the United States. Criminol
Public Policy 10:483–504
Guo S, Fraser MW (2010) Propensity score analysis: statistical methods and applications. Sage, Thousand
Oaks
Hipp JR, Petersilia J, Turner S (2010) Parolee recidivism in California: the effect of neighborhood context
and social service agency characteristics. Criminology 48:947–979
Irwin J (2005) The warehouse prison: disposal of the new dangerous class. Roxbury Press, Los Angeles
Jonson CL (2011) The impact of imprisonment on reoffending: a meta-analysis. Unpublished doctoral
dissertation, University of Cincinnati, OH
J Quant Criminol (2014) 30:317–347 345
123
Author's personal copy
King RD, Massoglia M, Macmillan R (2007) The context of marriage and crime: gender, the propensity to
marry, and offending in early adulthood. Criminology 45:33–65
Kovandzic TV, Vieraitis LM (2005) The effect of county-level prison population growth on crime rates.
Criminol Public Policy 5:213–244
Kubrin ChE, Stewart EA (2006) Predicting who reoffends: the neglected role of neighborhood context in
recidivism studies. Criminology 44:165–197
Lawrence R (1991) Reexamining community corrections models. Crime Delinquency 37:449–464
Levitt SD (1996) The effect of prison population size on crime rates: evidence from prison overcrowding
litigation. Q J Econ 111:319–351
Listwan SJ, Sullivan CJ, Agnew R, Cullen FT, Colvin M (2011) The pains of imprisonment revisited: the
impact of strain on inmate recidivism. Justice Q 30:144–168
Loughran TA, Mulvey EP, Schubert CA, Fagan J, Piquero AR, Losoya SH (2009) Estimating a dose-
response relationship between length of stay and future recidivism in serious juvenile offenders.
Criminology 47:699–740
MacKenzie DL (2006) What works in corrections: reducing the criminal activities of offenders and
delinquents. Cambridge University Press, New York
Marvell TB, Moody CE Jr (1994) Prison population growth and crime reduction. J Quant Criminol
10:109–140
May DC, Wood PB, Mooney JL, Minor KI (2005) Predicting offender-generated exchange rates: impli-
cations for a theory of sentence severity. Crime Delinquency 51:373–399
McDougall C, Cohen MA, Swaray R, Perry A (2003) The costs and benefits of sentencing: a systematic
review. Ann Am Acad Polit Soc Sci 587:160–177
Mears DP (2010) American criminal justice policy. Cambridge University Press, New York
Mears DP, Cochran JC, Bales WD (2012) Gender differences in the effects of prison on recidivism. J Crim
Just 40:370–378
Nagin DS (1998) Criminal deterrence research at the outset of the twenty-first century. Crime Justice
23:1–42
Nagin DS, Cullen FT, Jonson CL (2009) Imprisonment and reoffending. Crime Justice 38:115–200
Paternoster R (1987) The deterrent of the perceived certainty and severity of punishment: a review of the
evidence and issues. Justice Q 4:173–217
Paternoster R, Brame R (2008) Reassessing race disparities in Maryland capital cases. Criminology
46:971–1008
Petersilia J (1995) A crime control rationale for reinvesting in community corrections. The Prison J
75:479–496
Petersilia J (1997) Probation in the United States. Crime Justice 22:149–200
Petersilia J (2003) When prisoners come home: parole and prisoner reentry. Oxford University Press, New
York
Petersilia J, Turner S (1993) Intensive probation and parole. Crime Justice 17:281–335
Petersilia J, Deschenes EP (1994) Perceptions of punishment: Inmates and staff rank the severity of prison
versus intermediate sanctions. The Prison J 74:306–328
Piehl AM, LoBuglio SF (2005) Does supervision matter? In: Travis J, Visher C (eds) Prisoner reentry and
public safety in America. Cambridge University Press, Cambridge
Pratt TC (2008) Addicted to incarceration: corrections policy and the politics of misinformation in the
United States. Sage, Thousand Oaks
Pratt TC, Cullen FT, Blevins KR, Daigle LE, Madensen TD (2006) The empirical status of deterrence
theory. In: Cullen FT, Wright JP, Blevins KR (eds) Taking stock: the status of criminological theory.
Transaction Books, New Brunswich, pp 367–396
Raphael S, Stoll MA (2009) Why are so many Americans in prison? In: Raphael S, Stoll MA (eds) Do
prisons make us safer? The benefits and costs of the prison boom. Russell Sage Foundation, New York,
pp 27–72
Reitz KR (2011) Sentencing. In: Wilson JQ, Petersilia J (eds) Crime and public policy. Oxford University
Press, New York, pp 467–498
Rosenbaum PR, Rubin DB (1983) The central role of the propensity score in observational studies for causal
effects. Biometrika 70:41–55
Rosenbaum PR, Rubin DB (1984) Reducing bias in observational studies using subclassification on the
propensity score. J Am Stat Assoc 79:516–524
Rosenfeld R, Messner SF (2009) The crime drop in comparative perspective: the impact of the economy and
imprisonment on American and European burglary rates. Br J Sociol 60:445–471
Sampson RtJ (1986) The effects of formal and informal social control. Crime Justice 8:271–311
346 J Quant Criminol (2014) 30:317–347
123
Author's personal copy
Sampson RJ (2010) Gold standard myths: observations on the experimental turn in quantitative criminology.
J Quant Criminol 26:489–500
Simon J (2007) Governing through crime: how the war on crime transformed American democracy and
created a culture of fear. Oxford University Press, New York
Smith LG, Akers RL (1993) A comparison of recidivism of Florida’s community control and prison: a five-
year survival analysis. J Res Crime Delinquency 30:267–292
Smith P, Goggin C, Gendreau P (2002) The effects of prison sentences and intermediate sanctions on
recidivism: general effects and individual differences. Solicitor General of Canada, Ontario
Snodgrass G, Matthew AA, Blokland J, Haviland A, Nieuwbeerta P, Nagin DS (2011) Does the time cause
the crime? An examination of the relationship between time served and reoffending in the Netherlands.
Criminology 49:1149–1194
Spelman W (1995) The severity of intermediate sanctions. J Res Crime Delinquency 32:107–135
Spelman W (2000) The limited importance of prison expansion. In: Blumstein A, Wallman J (eds) The
crime drop in America. Cambridge University Press, Cambridge, pp 97–129
Tonry M (2011) Making peace, not a desert: penal reform should be about values not justice reinvestment.
Criminol Public Policy 10:637–649
Tonry M, Lynch M (1996) Intermediate sanctions. Crime Justice 20:99–144
Travis J, Visher C (eds) (2005) Prisoner reentry and crime in America. Cambridge University Press, New
York
Useem B, Piehl AM (2008) Prison state: the challenge of mass incarceration. Cambridge University Press,
New York
Villettaz P, Killias M, Zoder I (2006) The effects of custodial vs. noncustodial sentences on re-offending: a
systematic review of the state of knowledge. The Campbell Collaboration Crime and Justice Group,
Philadelphia
Welsh WN, Harris PW (2008) Criminal justice policy and planning, 3rd edn. LexisNexis, Anderson Pub-
lishing, Dayton
Western B (2007) Punishment and inequality in America. Russell Sage Foundation, New York
White MD, Saunders J, Fisher C, Mellow J (2012) Exploring inmate reentry in a local jail setting: impli-
cations for outreach, service use, and recidivism. Crime Delinquency 58:124–146
Winship C, Morgan SL (1999) The estimation of causal effects from observational data. Annu Rev Sociol
25:659–706
J Quant Criminol (2014) 30:317–347 347
123
Author's personal copy
... Despite its popularity, serious questions about the effectiveness of incarceration have been raised over the past several decades (e.g., see Smith & Akers, 1993;Cullen, Jonson, & Nagin, 2011;Bales & Piquero, 2012;Cochran, Mears, & Bales, 2014) and especially for drug offenders (Mitchell, Wilson, & Mackenzie, 2007). In the United States, 68% of those who are released from prison and jail are rearrested within three years post-release (Alper, Durise, & Markman, 2018); and in some studies, approximately 77% of drug offenders are rearrested within five years of release from prison (Durose, Cooper, & Snyder, 2014). ...
... Since the early 2000s there has been a major push to divert offenders from prison to community-based corrections. It has been suggested that these practices provide numerous benefits and successes compared to incarceration (Cochran et al., 2014), especially with drug offenders (Mitchell et al., 2007). For example, Cochran et al. (2014) utilized a sample of convicted offenders in Florida employing a propensity score matching technique, finding that less severe community sanctions were more likely to reduce recidivism. ...
... It has been suggested that these practices provide numerous benefits and successes compared to incarceration (Cochran et al., 2014), especially with drug offenders (Mitchell et al., 2007). For example, Cochran et al. (2014) utilized a sample of convicted offenders in Florida employing a propensity score matching technique, finding that less severe community sanctions were more likely to reduce recidivism. ...
Article
Nordic countries incarcerate offenders at much lower rates in comparison with incarceration rates in the United States, and reincarcerate fewer people per capita. Non-custodial alternatives to sanctions, including fines and community service, are used extensively in Finland to reduce negative effects of institutionalization and subsequent disadvantage caused by incarceration. The nature of drug-involved offenders within the Finnish system is reviewed in light of current research about the effectiveness of incarceration and deterrence-based approaches for drug offenders. Employing a 2014 sample from register data (consisting of official government records) of drug offenders in Finland with a 3-year recidivism period, this study utilizes a genetic matching procedure to compare offenders who received fines, conditional sentences (probation), or incarceration. While recognizing that numerous confounding variables affect incarceration, we compare a matched sample of drug offenders and the sanctions they have received from the Finland judicial system to determine whether offenders who initially receive a fine or a conditional sentence reenter the correctional system at different rates than those who are incarcerated. After matching, results found no significant differences between offenders receiving incarceration sentences or those who received noncustodial sentences (fine, or conditional sentence) for general and drug-related recidivism. These results are presented within the context of the Finnish corrections system in order to inform the criminal justice community about culture, incarceration, and process differences that could positively affect working with drug offenders in other localities.
... Shorter stay while awaiting trial were linked to increased criminal intent. These findings contrast Cochran, Mears and Bales (2014), who found that the time served and recurrence rates were not significantly related, and Chen & Shapiro (2007), which argued that the longer prison service do not reduce recurrence. The research further corroborates the findings of Jason and Kyleigh (2016), which observed that the risk of relapse among the longest serving parole inmates is significantly greater than the shorter serving probation prisoners. ...
Article
Full-text available
Criminal intent is a major precursor of reoffending behaviour and one of the big issues for ex-convicts and investigations have been undertaken to determine reasons of re-offending within imprisoned population. Evidence is still needed on the influence of family support and incarceration duration on intent to reoffending among inmates in Nigeria. Prison inmates (291) were selected for a cross-sectional survey in the Agodi prison. Results revealed a significant inverse relationship among family and partner support and criminal intent behavior. Family and partner support predicted criminal intent. Short duration of stay (less than 2 years) in the prison among inmates awaiting trial was associated with higher criminal intent. It was concluded that low family/partner support and shorter stay in prison were possible precursor of reoffending among inmates in Agodi prisons. Persistent assessment of institutional environment risk for re-offending among inmates using valid risk assessments tools was recommended.
... For example, most prior research is based upon mixed-gender samples, and analysts generally do not conduct supplemental analyses to determine whether incarceration works differently for males and females (e.g., Wermink et al. 2010;Bales and Piquero 2012a;Mears and Cochran 2018). However, some researchers investigated sanction effects exclusively among males (e.g., Cochran, Mears, and Bales 2014;Jolliffe and Hedderman 2015), exclusively among females (e.g., Hedderman and Jolliffe 2015), or in analyses stratified by gender (e.g., Mitchell et al. 2017;Caudy, Skubak Tillyer, and Tillyer 2018). Another source of heterogeneity is the age composition of samples. ...
Article
Beginning in the 1970s, the United States began an experiment in mass imprisonment. Supporters argued that harsh punishments such as imprisonment reduce crime by deterring inmates from reoffending. Skeptics argued that imprisonment may have a criminogenic effect. The skeptics were right. Previous narrative reviews and meta-analyses concluded that the overall effect of imprisonment is null. Based on a much larger meta-analysis of 116 studies, the current analysis shows that custodial sanctions have no effect on reoffending or slightly increase it when compared with the effects of noncustodial sanctions such as probation. This finding is robust regardless of variations in methodological rigor, types of sanctions examined, and sociodemographic characteristics of samples. All sophisticated assessments of the research have independently reached the same conclusion. The null effect of custodial compared with noncustodial sanctions is considered a “criminological fact.” Incarceration cannot be justified on the grounds it affords public safety by decreasing recidivism. Prisons are unlikely to reduce reoffending unless they can be transformed into people-changing institutions on the basis of available evidence on what works organizationally to reform offenders.
... With mounting political and economic pressures stemming from mass incarceration and the recent gradual shift toward decarceration, the need to prepare inmates for reentry through meaningful correctional programs has become even more critical in the United States (Cochran et al., 2014). Emotional Literacy (EL) correctional programs arespotlighted here, as there is some empirical evidence that they can impact recidivism. ...
Article
Full-text available
Applying a random-effect meta-analysis, the current study examines previous empirical findings about three main purposes of Emotional Literacy (EL) correctional programs: improving empathy, mindfulness, and self-regulation among inmate populations. A total of 22 programs in 15 studies in Germany, Portugal, the Netherlands, and the US are analyzed. Results suggest that EL programs significantly improve levels of empathy, mindfulness, and self-regulation, although a possibility of publication bias is observed. Based on these findings, this study concludes that EL programs retain a potential as an effective rehabilitation strategy and should be sought and implemented in U.S. correctional settings.
... For instance, while individuals released from prison have a high recidivism rate, with estimates indicating that around 75% of individuals released from prison will be rearrested within 5 years, there is still a substantial percentage of individuals who are able to avoid future crime (Durose et al., 2014). Similarly, across both traditional community supervision and prison diversion programs, findings have indicated that these programs can reduce recidivism in some instances (e.g., Aos et al., 2006;Cochran et al., 2014;Kelsay et al., 2019;Padgett et al., 2006;Paparozzi & gendreau, 2005), while other findings have suggested that these programs have a limited effect on recidivism (e.g., Hyatt & Barnes, 2017;Latessa et al., 1998;Renzema & Mayo-wilson, 2005), or are even associated with increases in reoffending (e.g., Aos et al., 2006;gendreau et al., 1993;Petersilia & Turner, 1993). To better understand what works with justice-involved individuals, researchers have investigated the sources of variation in effectiveness based on program characteristics. ...
Article
Prior research on various correctional sanctions has sought to explain variation in recidivism. The goal of the current study was to examine whether social disorganization mechanisms, such as residential mobility and family disruption, significantly influenced the recidivism rates of justice-involved individuals supervised under prison diversion programs, traditional community supervision programs, as well as individuals released from prison. Multilevel modeling of justice-involved individuals nested within 25 counties was used to examine the recidivism rates of 2,855 prison diversion cases, 2,278 traditional community supervision cases, and 497 incarceration cases. The results indicated that social disorganization mechanisms, specifically family disruption and residential mobility, significantly influenced the recidivism rates of justice-involved individuals supervised on all three types of correctional sanctions. Therefore, the results indicate that social disorganization mechanisms may partly explain the variation in recidivism rates of justice-involved individuals supervised under various correctional sanctions.
... For instance, while individuals released from prison have a high recidivism rate, with estimates indicating that around 75% of individuals released from prison will be rearrested within 5 years, there is still a substantial percentage of individuals who are able to avoid future crime (Durose et al., 2014). Similarly, across both traditional community supervision and prison diversion programs, findings have indicated that these programs can reduce recidivism in some instances (e.g., Aos et al., 2006;Cochran et al., 2014;Kelsay et al., 2019;Padgett et al., 2006;Paparozzi & gendreau, 2005), while other findings have suggested that these programs have a limited effect on recidivism (e.g., Hyatt & Barnes, 2017;Latessa et al., 1998;Renzema & Mayo-wilson, 2005), or are even associated with increases in reoffending (e.g., Aos et al., 2006;gendreau et al., 1993;Petersilia & Turner, 1993). To better understand what works with justice-involved individuals, researchers have investigated the sources of variation in effectiveness based on program characteristics. ...
Article
Prior research on various correctional sanctions has sought to explain variation in recidivism. The goal of the current study was to examine whether social disorganization mechanisms, such as residential mobility and family disruption, significantly influenced the recidivism rates of justice-involved individuals supervised under prison diversion programs, traditional community supervision programs, as well as individuals released from prison. Multilevel modeling of justice-involved individuals nested within 25 counties was used to examine the recidivism rates of 2,855 prison diversion cases, 2,278 traditional community supervision cases, and 497 incarceration cases. The results indicated that social disorganization mechanisms, specifically family disruption and residential mobility, significantly influenced the recidivism rates of justice-involved individuals supervised on all three types of correctional sanctions. Therefore, the results indicate that social disorganization mechanisms may partly explain the variation in recidivism rates of justice-involved individuals supervised under various correctional sanctions.
Article
Recidivism is a foundational concept that represents the continued criminal behavior after system contact. Evidence suggests that how recidivism is operationalized can produce varying estimates of recidivism, yet this disparity has not been estimated within the juvenile justice system. This study examined — through event history analysis — whether using different official measures of recidivism produced disparate estimates of recidivism. This study compared the hazards of recidivism among three unique operational definitions of recidivism – offense date, referral date, and adjudication date – among a cohort of 10,830 juvenile offenders from a large southern state. Two hypotheses were tested – (a) the use of different operational definitions of recidivism produces disparate recidivism hazards and (b) the use of different operational definitions of recidivism produces disparate effects across the correlates of recidivism. The results suggest that official measures of juvenile recidivism produce significantly different hazard estimates and the operational definition of recidivism had significant effects on the correlations between the recidivism measure and predictor variables among juvenile offenders.
Chapter
Tough-on-crime policies and strategies, such as mandatory minimum and determinate sentencing laws and more severe punishments for juveniles, sex offenders, and drug offenders, have caused a substantial increase in the U.S. incarceration rate. Mass incarceration has generated several negative consequences, including racial bias and disparities, economic and social costs, and prison overcrowding. As such, the use of community corrections programs as an alternative form of sentencing has significantly increased. To effectively reduce crime and recidivism, community corrections programs must 1) utilize a validated risk and needs assessment when determining program placement; 2) provide participants with comprehensive, evidence-based services, including substance abuse and mental health treatment; 3) focus on modifying participants' behavior through treatment goal setting and the use of graduated incentives and sanctions; and 4) strive to hire and retain qualified staff, provide both initial and ongoing training, and monitor treatment staff with regular clinical supervision.
Chapter
In diesem Kapitel geht es zunächst um eine Klärung dessen, was Ziele und angestrebte Wirkungen des Strafvollzugs sind. Es wird herausgearbeitet, dass Strafvollzugsgesetze eine Reihe von Wirkungen thematisieren und dass die Befähigung zu einem Leben in sozialer Verantwortung ohne Straftaten das zentrale Ziel darstellt. Aus diesem Grund werden nachfolgend Forschungsbefunde zur Legalbewährung nach Entlassung aus dem Strafvollzug, auch im Vergleich mit ambulanten Sanktionen, referiert. Probleme bei der Messung der Legalbewährung, die anschließend besprochen werden, führen zur Frage, welche alternativen bzw. ergänzenden Wirkungsindikatoren erfasst werden könnten. Hierzu werden Vorschläge unterbreitet.
Book
Criminal Justice Policy and Planning: Planned Change, Fifth Edition, presents a comprehensive and structured account of the process of administering planned change in the criminal justice system. Welsh and Harris detail a simple yet sophisticated seven-stage model, which offers students and practitioners a full account of program and policy development from beginning to end. The authors thoughtfully discuss the steps: analyzing a problem; setting goals and objectives; designing the program or policy; action planning; implementing and monitoring; evaluating outcomes; and reassessing and reviewing. Within these steps, students focus on performing essential procedures, such as conducting a systems analysis, specifying an impact model, identifying target populations, making cost projections, collecting monitoring data, and performing evaluations. In reviewing these steps and procedures, students can develop a full appreciation for the challenges inherent in the process and understand the tools that they require to meet those challenges. https://www.routledge.com/products/9780323298858
Book
http://johnbraithwaite.com/monographs/
Article
Every year, hundreds of thousands of jailed Americans leave prison and return to society. Largely uneducated, unskilled, often without family support, and with the stigma of a prison record hanging over them, many, if not most, will experience serious social and psychological problems after release. Fewer than one in three prisoners receive substance abuse or mental health treatment while incarcerated, and each year fewer and fewer participate in the dwindling number of vocational or educational pre-release programs, leaving many all but unemployable. Not surprisingly, the great majority is rearrested, most within six months of their release. As long as there have been prisons, society has struggled with how best to help prisoners reintegrate once released. But the current situation is unprecedented. As a result of the quadrupling of the American prison population in the last quarter century, the number of returning offenders dwarfs anything in America's history. A crisis looms, and the criminal justice and social welfare system is wholly unprepared to confront it. Drawing on dozens of interviews with inmates, former prisoners, and prison officials, the book shows us how the current system is failing, and failing badly. Unwilling merely to sound the alarm, it explores the harsh realities of prisoner re-entry and offers specific solutions to prepare inmates for release, reduce recidivism, and restore them to full citizenship, while never losing sight of the demands of public safety. As the number of ex-convicts in America continues to grow, their systemic marginalization threatens the very society their imprisonment was meant to protect.
Book
The number of people incarcerated in U.S. prisons and jails more than quadrupled between 1975 and 2005, reaching the unprecedented level of over two million inmates today. Annual corrections spending now exceeds 64 billion dollars, and many of the social and economic burdens resulting from mass incarceration fall disproportionately on minority communities. Yet crime rates across the country have also dropped considerably during this time period. In Do Prisons Make Us Safer? leading experts systematically examine the complex repercussions of the massive surge in our nation?s prison system. Do Prisons Make Us Safer? asks whether it makes sense to maintain such a large and costly prison system. The contributors expand the scope of previous analyses to include a number of underexplored dimensions, such as the fiscal impact on states, effects on children, and employment prospects for former inmates. Steven Raphael and Michael Stoll assess the reasons behind the explosion in incarceration rates and find that criminal behavior itself accounts for only a small fraction of the prison boom. Eighty-five percent of the trend can be attributed to "get tough on crime" policies that have increased both the likelihood of a prison sentence and the length of time served. Shawn Bushway shows that while prison time effectively deters and incapacitates criminals in the short term, long-term benefits such as overall crime reduction or individual rehabilitation are less clear cut. Amy Lerman conducts a novel investigation into the effects of imprisonment on criminal psychology and uncovers striking evidence that placement in a high security penitentiary leads to increased rates of violence and anger-particularly in the case of first time or minor offenders. Rucker Johnson documents the spill-over effects of parental incarceration-children who have had a parent serve prison time exhibit more behavioral problems than their peers. Policies to enhance the well-being of these children are essential to breaking a devastating cycle of poverty, unemployment, and crime. John Donohue?s economic calculations suggest that alternative social welfare policies such as education and employment programs for at-risk youth may lower crime just as effectively as prisons, but at a much lower human cost. The cost of hiring a new teacher is roughly equal to the cost of incarcerating an additional inmate. The United States currently imprisons a greater proportion of its citizens than any other nation in the world. Until now, however, we?ve lacked systematic and comprehensive data on how this prison boom has affected families, communities, and our nation as a whole. Do Prisons Make Us Safer? provides a highly nuanced and deeply engaging account of one of the most dramatic policy developments in recent U.S. history.
Article
The propensity score is the conditional probability of assignment to a particular treatment given a vector of observed covariates. Previous theoretical arguments have shown that subclassification on the propensity score will balance all observed covariates. Subclassification on an estimated propensity score is illustrated, using observational data on treatments for coronary artery disease. Five subclasses defined by the estimated propensity score are constructed that balance 74 covariates, and thereby provide estimates of treatment effects using direct adjustment. These subclasses are applied within sub-populations, and model-based adjustments are then used to provide estimates of treatment effects within these sub-populations. Two appendixes address theoretical issues related to the application: the effectiveness of subclassification on the propensity score in removing bias, and balancing properties of propensity scores with incomplete data.
Chapter
The answer to the question, “Does supervision matter?” is central for designing postincarceration policy for those released from prison. Knowing how much supervision can help and which elements of supervision are efficacious could lead to better outcomes for communities and for recent inmates. The improvement could take the form of reduced criminal offending, the particular emphasis of this book, or other benefits such as reduced substance abuse or better employment outcomes. Before attempting to answer the question posed by the title of this chapter, it is helpful to be more precise. What do we mean by supervision? This term generally refers to the structured monitoring and support by law enforcement following release from prison, such as “parole supervision.” Even if the term parole has ever been sufficient to describe the various state practices for monitoring ex-inmates in the community, it certainly is not sufficient now. Depending on the state, postincarceration supervision can be provided by parole departments (some of which fall under departments of correction), probation departments, other entities, or some combination of these. The status of being supervised in the community may be the result of a decision by a parole board to offer conditional release for the remainder of the sentence or it may result from the original sentence from the court. The latter is generally referred to as “mandatory supervision.” Those states that “abolished parole” have by and large replaced it with some other form of community supervision, lending yet another dimension of complexity to the nomenclature.
Article
Within the past 25 years, the prison population in America shot upward to reach a staggering 1.53 million by 2005. This book takes a broad, critical look at incarceration, the huge social experiment of American society. The authors investigate the causes and consequences of the prison buildup, often challenging previously held notions from scholarly and public discourse. By examining such themes as social discontent, safety and security within prisons, and impact on crime and on the labor market, Piehl and Useem use evidence to address the inevitable larger question, where should incarceration go next for American society, and where is it likely to go?.
Article
Over the last three decades the United States has built a carceral state that is unprecedented among Western countries and in US history. Nearly one in 50 people, excluding children and the elderly, is incarcerated today, a rate unsurpassed anywhere else in the world. What are some of the main political forces that explain this unprecedented reliance on mass imprisonment? Throughout American history, crime and punishment have been central features of American political development. This book examines the development of four key movements that mediated the construction of the carceral state in important ways: The victims' movement, the women’s movement, the prisoners' rights movement, and opponents of the death penalty. This book argues that punitive penal policies were forged by particular social movements and interest groups within the constraints of larger institutional structures and historical developments that distinguish the United States from other Western countries.