FOR SURAHSIGN ONLY Due wed.

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=rjqy20

Download by: [Capella University] Date: 15 July 2016, At: 17:01

Justice Quarterly

ISSN: 0741-8825 (Print) 1745-9109 (Online) Journal homepage: http://www.tandfonline.com/loi/rjqy20

Supervision Intensity and Parole Outcomes: A Competing Risks Approach to Criminal and Technical Parole Violations

Ryken Grattet & Jeffrey Lin

To cite this article: Ryken Grattet & Jeffrey Lin (2016) Supervision Intensity and Parole Outcomes: A Competing Risks Approach to Criminal and Technical Parole Violations, Justice Quarterly, 33:4, 565-583, DOI: 10.1080/07418825.2014.932001

To link to this article: http://dx.doi.org/10.1080/07418825.2014.932001

Published online: 27 Jun 2014.

Submit your article to this journal

Article views: 354

View related articles

View Crossmark data

Citing articles: 2 View citing articles

Supervision Intensity and Parole Outcomes: A Competing Risks Approach to Criminal and Technical Parole Violations

Ryken Grattet and Jeffrey Lin

Recent scholarship about parole supervision indicates that higher supervision intensity is associated with an increased risk of parole violations. However, parole violations can take many forms—some minor and some serious—and theory suggests that supervision intensity might have differential effects depending upon the type of violation. We use “competing risks” survival models to identify supervision effects on five types of parole violations among 79,082 individuals released from prison in California: absconding, technical violations, drug use, violent offenses, and sexual offenses. We find that supervision effects are strongest for absconding violations. Past sexual offending also triggers significant supervision effects for technical viola- tions, drug use violations, and violent violations. We conclude that parole violation patterns are influenced by parolee behaviors, the amount of atten- tion the state is paying to those behaviors, and official markers of criminal dangerousness that are attached to particular parolees.

Keywords parole violators; recidivism; competing risks; supervision regime; societal reaction; liberation hypothesis

Introduction

The reincarceration of parolees has contributed significantly to America’s high rate of imprisonment (Blumstein & Beck, 2005; Lin, 2010; Travis, 2007).

Ryken Grattet is a professor of Sociology at the University of California, Davis, and research fellow at the Public Policy Institute of California. He writes about punishment and criminal law, focusing on the development and enforcement of hate crime law and the California prison and parole sys- tem. Jeffrey Lin is an assistant professor of Sociology and Criminology at the University of Denver. His research focuses on decision-making in criminal justice institutions, with specific attention to the treatment of juveniles, parolees, and sex offenders. Correspondence to: R. Grattet, University of California, Davis and the Public Policy Institute of California, USA. Email: [email protected].

� 2014 Academy of Criminal Justice Sciences

Justice Quarterly, 2016 Vol. 33, No. 4, 565–583, http://dx.doi.org/10.1080/07418825.2014.932001

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

Parolee reincarceration often results from the detection and punishment of violations by parole agencies. Thus, the identification of individual and institu-

tional factors that explain parole violations has emerged as an important topic in correctional research. Research on parole violations has a long lineage in

association with the creation of actuarial instruments designed to predict paro- lee misconduct (Harcourt, 2007). However, more recent work has focused less on prediction and more on developing a scholarly understanding of the sources

of parolee behavior and systemic responses to it. Recently, Grattet, Lin, and Petersilia (2011) considered the effects of “supervision regimes” on parolee

deviance, focusing on the ways that the system of parole supervision interacts with the backgrounds and offending histories of parolees to affect the overall

risk of parole violations. They found that, holding constant a parolee’s risk factors, more intensive supervision increases the likelihood of a violation.

Moreover, certain pivotal categories of offenders—namely sex offenders and those with two strikes under California’s “three strikes and you’re out” law

(also known as “second strikers”)—experienced particularly elevated violation hazards under more intensive supervision. Some parole agent characteristics, as well as regional and bureaucratic factors, also affected the chances of

detecting and reporting parole violations. Thus, the authors conclude that par- ole violations are a complex result of individual, institutional, and geographic

(i.e. cultural) factors that combine to “produce” violation events, marking a significant advance from the conceptualization of parole violations as simply

the product of offender-level criminal risk factors. While Grattet et al.’s (2011) study represents an advance in understanding

the ways that parole violations emerge from an interaction between what the parolee does, who the parolee is, and institutional orientations toward those factors, their analysis collapses all types of violations—from relatively minor

technical violations to very serious violent and sexual offenses—into a single dependent variable. In this paper, we refine this analysis by examining how dif-

ferent levels of supervision intensity affect the reporting and detection of dif- ferent types of violations—specifically, absconding, technical violations, drug

offenses, property offenses, sexual offenses, and violent offenses. Given that supervision intensity affects violation likelihood overall, we wish to determine

whether the determinants of parolee deviance vary according to the type of violation under investigation and the type of parolee that commits the viola-

tion. In other words, do supervision effects differ across violation types? How do key parolee characteristics influence this dynamic?

To answer these questions we rely on data from the California Parole Study,

collected by Grattet, Petersilia, and Lin (2008), on parolees released from California prisons during 2003 and 2004. We apply a competing risks hazard

model to these data and find that supervision intensity matters more for some kinds of violations than others, and certain parolee characteristics affect the

degree to which supervision intensity matters for certain violation types. However, before turning to these results we describe the theory and

background literature that motivate our study.

566 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

Literature

Recent research on parole supervision indicates that the risk of recidivism— either for new criminal activity or the violation of technical conditions of

supervision—is a function of both individual risk and the orientation of control institutions toward offenders (Grattet et al., 2011; Lin, Grattet, & Petersilia,

2010; see also Simon, 1993). This perspective extends earlier work that almost exclusively conceptualized recidivism risk as a product of offender characteris- tics such as demographics, criminal history, mental health, substance use, and

social relationships (Feeley & Simon, 1992; Harcourt, 2007; Petersilia, 2009; Silver & Miller, 2002). These offender-level factors are central to the scholarly

literature about offender behaviors, and, importantly, they are used in a vari- ety of actuarial instruments adopted in correctional practice to make release

and supervision decisions (Harcourt, 2007; Petersilia, 2009). However, this per- spective mostly ignores the effects of institutional circumstances on the likeli-

hood of parole violation. The integration of institutional factors into thinking about parolee deviance

emerges from societal reaction (or labeling) theory, which includes a focus on organizational conditions that shape the day-to-day operations of social control agencies (Becker, 1963; Cicourel, 1976; Cavender & Knepper, 1992; Drass &

Spencer, 1987; Emerson, 1983; Kitsuse & Cicourel, 1963; McCleary, 1977). Quantitative and qualitative studies of community supervision have empirically

highlighted the key contours of these dynamics. Ethnographic research on both parolees and parole agencies illustrates how individual and institutional char-

acteristics interact in complex, changing ways (Lynch, 1998; Rudes, 2012). The rules of parole—and the ways in which parolees respond to those rules—are

shaped by a multitude of factors such as social ecological conditions, resource constraints, legislative mandates, and officer backgrounds that produce impro- visational on-the-ground strategies that agents and parolees each develop to

navigate the murky reality of supervision (Lynch, 1998; McCleary, 1977; Werth, 2012). These dynamics have also been broadly transformed over time by larger

developments in criminal justice. Simon (1993) describes the critical changes in parole supervision that occurred during the twentieth century, which

reflected a broad philosophical shift from therapeutic to social control ideals, leading to the emergence of managerial techniques to realize modern correc-

tional goals under severe resource constraints (see also Feeley & Simon, 1992). The explosive growth of correctional populations through the economically

lean years of the early twenty-first century has strained parole agencies and this has further contributed to the widespread adoption of corporatized accounting methods and intense institutional attention to the goal of recidi-

vism reduction (see Lin, 2010). The aforementioned ethnographic and historical accounts highlight a multi-

tude of ways in which the nature of supervision can shift to raise or lower the chance of violation detection. Parole personnel monitor some parolees more

SUPERVISION INTENSITY AND PAROLE OUTCOMES 567

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

closely than others. They display less tolerance for the mistakes of some parolees. And they make these strategic decisions, in part, to effectively man-

age their workloads. Statistical studies of community supervision support the idea that individual and institutional conditions interact to produce recidivism

outcomes, and serve to operationalize institutional characteristics that affect these outcomes. Turner, Petersilia, and Deschenes (1992) reported the results of a randomized field experiment comparing the recidivism of drug offenders

on regular community supervision against those on intensive supervision. They found that intensive supervision produced more technical violations, which led

to more reincarcerations (see also Petersilia & Turner, 1993). In other words, watching offenders more closely produced more violations because misbehav-

ior was more easily detected. Sirakaya (2006) analyzed the recidivism patterns of felony probationers, finding that stricter supervision predicted recidivism

likelihood even after controlling for other predictive characteristics. And as mentioned above, Grattet et al. (2011) extended this line of research by devel-

oping the idea of a “supervision regime” to capture institutional characteristics that may raise or lower the risk of parolee recidivism. Specifically, they argued that the supervision regime can be empirically partitioned into three subcate-

gories: intensity of supervision, capacity of the agency to supervise parolees, and the tolerance of parole agents for various forms of parolee deviance.

Intensity refers to the frequency and content of parole supervision, and the authors found that supervision intensity was partially responsible for the viola-

tion risk of parolees in California, and that this effect was most pronounced for serious, violent, and sexual offenders. Capacity was operationalized as

agent workload and weakly predicted violation likelihood; higher caseloads predicted small reductions in the odds of violation. Tolerance was operational- ized as parole agent demographics, tenure, and professional background

(whether they had ever worked in a prison), and the authors found that agent race and tenure affected violation likelihoods. They also tested for differential

tolerance across California’s four different parole regions, finding that one—Los Angeles—was significantly more tolerant than the other three. Taking these

findings together, the authors conclude that parolee deviance is best conceptu- alized as a “joint production” between the parolee and the parole agency.

Each contributes to violation patterns, and therefore, research in this area must account for both.

Existing research on supervision effects has been limited by vagueness in the dependent variable, as many types of violations tend to be lumped together as a single outcome (Grattet et al., 2011). However, parole violations are not all

alike. They differ in their severity and character. Minor technical violations like missing appointments are not the same as serious criminal violations like

alleged homicides and sexual assaults. Thus, in this article, we examine key violation types separately. We model three types of minor violations: abscond-

ing, miscellaneous technical violations, and drug use or possession (which is typically detected through a urine test administered by the parole agency). We

568 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

also model two categories of serious criminal violations: new arrests for violent crimes and new arrests for sexual crimes.

Diversifying outcomes in this way makes sense from a policy perspective. Correctional administrators and policy-makers may find such an analysis more

useful than one in which disparate violation types are combined. However, this strategy also allows us to test the theoretical idea of liberation. The liberation hypothesis argues that decision-makers will display more discretion in cases

with weaker evidence or those involving less serious charges. The rationale is that in such cases, decision-makers are “liberated” to apply extralegal logic

like their own personal values (Chen, 2008; Kalven & Zeisel, 1966; Lin, Grattet, & Petersilia, 2012; Smith & Damphousse, 1998; Spohn & Cederblom, 1991;

Reskin & Visher, 1986; Unnever & Hembroff, 1988). Given that parole agents exercise a fair amount of discretion in terms of violation reporting, we use our

data to test the idea that discretionary factors, especially supervision effects, will have more influence over less serious cases.

The liberation hypothesis focuses on the severity of the crime or violation being evaluated by the sanctioning agency, but official perceptions of offend- ers’ alleged misconduct can also be influenced by official perceptions of the

offenders themselves. To this point, Grattet et al. (2011) found that supervi- sion effects were particularly pronounced for parolees officially designated as

violent or sexual offenders. They posited that parole agents have lower toler- ance for the misbehavior of parolees that occupy these “pivotal categories”

because when such parolees commit high-profile crimes, the parole agency receives intense public criticism. Making cautious decisions about high-profile

parolees thus serves as a sensible form of professional protection. Integrating this idea with the liberation hypothesis suggests that agents will have particu- larly low tolerance for the less serious violation behaviors of parolees who

occupy the aforementioned pivotal categories. Given the reputational conse- quences that parole personnel face when high-profile parolees recidivate, we

expect them to exploit discretionary opportunities to reincarcerate these parolees, including citing them for violations.1

In light of these theoretical considerations, we address the following research questions and hypotheses in this paper:

� For which types of parole violations are supervision effects most pro- nounced? Based on the liberation hypothesis, we would expect these effects to be strongest for the least serious violations.

� Are supervision effects more pronounced for high-profile parolees—those with histories of violent and sexual offending? We also consider whether black parolees and parolees with identified mental health conditions are

more at risk for violations when they are supervised at high levels.

1. Reincarcerating a parolee with a history of serious criminal behavior also removes that parolee from the officer’s caseload—another professional benefit of (and incentive for) low tolerance.

SUPERVISION INTENSITY AND PAROLE OUTCOMES 569

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

� Among high-profile parolees, for which types of parole violations are supervision effects most pronounced? We expect larger supervision

effects for the least serious violations.

Data and Methods

The data for this project are from the California Parole Study (Grattet et al., 2008), an investigation of parole violations and revocations in the California

parole system. The dependent variables are drawn from California Department of Corrections and Rehabilitation (CDCR) administrative databases, which track

parole violation incidents for all felons released from prison to parole between January 1st, 2003 and December 31st, 2004.2 The parolees in the study were

released from prison onto parole after serving a sentence for a new criminal offense, or after serving prison time for a criminal parole revocation in which they were given a new term of imprisonment by the court. Thus, we exclude

two groups included in Grattet, Lin, and Petersilia’s (2011) study. The first are individuals who were already on parole at the beginning of 2003. This removes

roughly 110,000 individuals who were already on parole, although some of those are likely to have been resentenced for new crimes and released during the

two-year observation period and thus re-enter the study population. The second are the roughly 60,000 parolees released from prison after having their parole

revoked by the parole board during 2003 and 2004,3 rather than being released after serving a new sentence delivered by the court. Again, some of these indi-

viduals may re-enter the study population if they were resentenced and released during the observation period. This latter exclusion criterion is empiri- cally appropriate because board-ordered revocation terms are still officially

part of original court-ordered sentences, rather than being new sentences themselves. Board-ordered revocations also involve significantly shorter prison

stays; they are statutorily limited to 12 months. Offenders are commonly returned to prison multiple times before completing their terms of parole,

cycling in and out of custody (Blumstein & Beck, 2005). Release from prison after serving a relatively short revocation term is therefore a fundamentally

different moment in an offender’s correctional trajectory than release from prison after serving a relatively longer “original” sentence. The criminogenic

2. Parole terms in California typically range from three years to life. All study subjects are under observation until they experience a violation event, are discharged from parole, or until December 31st, 2004. 3. Board-ordered revocations can result from non-criminal (technical) violations, and they can also result from new criminal behavior that does not result in a court conviction. When a parolee is accused of a crime, the court (i.e. district attorney) will evaluate the case first. If the district attorney decides not to pursue the case, or if the parolee is found not guilty, the case is then referred to the parole board, which evaluates the case under a more lenient standard of evidence, but can only order a maximum term of reincarceration of 12 months. See Lin et al. (2012) for more detail.

570 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

characteristics of parolees at these two points in time may differ, and institutional orientations toward them may also be fundamentally different. For

these reasons, parolees included in this study are restricted to those being released from prison in 2003 after receiving a new sentence from the court.4

Parolees are followed until their first failure. Because the data are from 2003 and 2004, individuals (N = 79,082) are observed until they experience a violation, are discharged from parole, or until December 31st, 2004—the end of

our observation period. In the vernacular-of-survival analysis, cases that result in discharge or reach the end of the observation period without experiencing a

violation are right censored. We classify failure types into five categories:

(1) Absconding (a technical violation of parole consisting of a failure to report to parole agent).

(2) Other technical offenses (including: access to alcohol or weapons, trav- eling outside of a 50 mile radius of the parolee’s residence, changing

residence without informing the parole division, and failure to attend an outpatient clinic).

(3) Drug use and possession violations.

(4) Violent offenses (including non-sexual violent offenses such as murder, manslaughter, assault, and robbery).

(5) Sexual offenses (including rape, oral copulation, sodomy, and other sex offenses, but does not include violations associated with failing to reg-

ister as a sex offender).

Importantly, separating parole violations into categories means that, for example, a first failure due to a drug use or possession violation excludes or “competes with” the possibility of failure due to a sexual violation.

Although there is a relatively large literature that focuses on single-event hazard models of recidivism (see, for example, Maltz, 1984; Schmidt &

Witte, 1989), only a few studies have employed a competing risks approach (see Bierens and Carvalho (2007) for a review). These latter studies are lim-

ited in their applicability to the present work because they focus on less nuanced measures of recidivism behavior, comparing outcomes such as felo-

nies vs. misdemeanors (Visher, Lattimore, & Linster, 1991), different types of convictions (Copas & Heydari, 1997), different probation revocation types

4. This contrasts with Grattet et al. (2011) who focus on all parolees on parole in California during 2003–2004. They also explicitly allow for repeated events (i.e. multiple instances of violation). Thus, our data represent a subset of their data, restricting attention to first releases onto parole and to first violations. We did so in order to appropriately apply the competing risks approach, as well as for reasons cited in the text above. The reader should also keep in mind that “first release to parole” refers to the first parole release for the current commitment offense. Parolees may have been imprisoned earlier in their lives and completed the parole terms associated with those earlier imprisonments. If these offenders are included in our sample, they have been rearrested and imprisoned for a new offense, and their “first release to parole” actually indicates first release for the current offense.

SUPERVISION INTENSITY AND PAROLE OUTCOMES 571

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

(Rhodes 1986), or sexual vs. non-sexual crimes (Escarela, Francis, & Soothill, 2000; Kruttschnitt et al., 2000; Letourneau Bandyopadhyay, Sinha, &

Armstrong, 2009; Letourneau, Levenson, Bandyopadhyay, Sinha, & Armstrong, 2010). Some of the earlier studies are also limited by small

sample sizes and all are limited in the range of covariates that can be examined, particularly with respect to the supervision factors identified in Grattet et al. (2011).

These studies’ empirical findings do not apply much to our present study, as their findings generally pertain to the effects of individual level-risk factors,

but they provide a road map for the range of methodological approaches that might be brought to bear on our data. Two general approaches for modeling

competing risks are described by Box-Steffensmeier and Jones (2004) and Singer and Willett (2003). The first approach assumes that different outcomes

have different underlying hazard functions and tries to account for this via a stratified hazards model (e.g. a stratified Cox model). This approach is

designed to adjust for heterogeneity in the hazard functions of different fail- ure outcomes. It does not permit an analysis of differential covariate effects across failure outcomes. In other words, it assumes that covariate effects are

constant across event types. Since that assumption does not comport with the theoretical arguments we described above, we opt for the second general

approach—an “event” or “type-specific” hazard model. Event-specific models assume that different outcomes not only have different baseline hazards, but

also that covariates may have different effects on different failure outcomes. Scholarship that informs our above hypotheses about the differential effects of

supervision intensity and the liberation of parole decision-makers suggests that this is likely the case.

The independent variables we examine are taken from CDCR databases and

include measures reflecting parolees’ offending backgrounds and personal characteristics, as well as their assigned supervision levels. Offense history

variables are used to capture the frequency, duration, and type of offending a parolee has engaged in previously. We include variables indicating whether a

parolee had previously been incarcerated in a California prison, whether they had ever been imprisoned for a violent offense, whether they had ever been

imprisoned for a sexual offense, and their age at first imprisonment (along with a quadratic term to capture the expected non-linear effect of age at first

imprisonment). Demographic variables and mental health status are included to measure

other commonly identified dimensions of risk. The demographic variables are

age at prison release (split into three categories: 18–30, 31–44, 45 and older, with the middle category as the reference category), gender, and a dummy-

coded variable indicating black race. The mental health status variable indi- cates whether a parolee had been officially classified by CDCR as having a men-

tal health disorder. We are specifically interested in how, after controlling for these factors, dif-

ferences in supervision affect the hazards of different kinds of violations. We

572 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

measure supervision as a dummy variable indicating whether the parolee was supervised on a high-risk caseload (also known as a “high control” caseload).

Grattet et al. (2011) employed a supervision variable that distinguished between high, medium, and low risk. However, given that the current sample

consists entirely of first releases to parole (excluding those being released from board-ordered revocation terms), and that CDCR almost always places such parolees on either higher or medium-risk caseloads, the contrast between

high supervision intensity and all other types makes the most sense from an empirical standpoint.

The supervision variables are also interacted with the two dummy-coded offense history variables indicating past violent or sexual convictions to exam-

ine whether an offender carrying these labels has an elevated hazard of expe- riencing a particular type of violation. In other words, these interactions are

designed to determine whether the effects of supervision vary according to the type of parolee being supervised. In addition to the “pivotal categories” of

sexual and violent offenders, we also examined whether black parolees and parolees with identified mental health conditions, supervised at a high level of intensity, also have elevated risks of each kind of violation. We do so because

the literature on labeling suggests that race and mental illness are sometimes used as markers of dangerousness and thus offenders who occupy these sta-

tuses might be given “a shorter leash.” In our model, we interact high supervi- sion intensity with black race and mental health status to test for the

presence of these effects. Key to our modeling strategy is that each of the different possible outcomes

represent competing risks. This represents two critical improvements to the statistical design used in Grattet et al. (2011). First, it allows for each viola- tion type to have its own distinct baseline hazard function. For example, it

seems quite likely that absconding from parole supervision might exhibit a distinct temporal patterning compared to violent or sexual offenses. Abscond-

ing from parole is often done soon after release—commonly called “absconding from the gate.” By contrast, the hazard functions for violent and sexual

offenses are likely to be more stable over time. In a Cox model, we need not make any particular assumptions about the hazard function, which is one of

the desirable features of this approach. By using a Cox model in the context of modeling competing risks, we allow for the hazard function to vary by type of

violation. The second advantage of the competing risks design is that it allows for the

investigation of the differential effects of covariates on violation categories of

interest. This possibility is explicitly indicated by the “liberation hypothesis,” which suggests that social control agents exercise more discretion over behav-

iors they view as less serious. The Cox model is again an appropriate choice, given that it can easily be implemented by running a sequence of separate

models examining the differential effects of covariates across different viola- tion outcomes. Each model focuses on one of the violation events described

above and treats the other violation events as censored.

SUPERVISION INTENSITY AND PAROLE OUTCOMES 573

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

Findings

The summary findings are fairly comparable to those found in other research on parolee cohorts released from California prisons. As Table 1 shows, 30% of

parolees had experienced previous spells of incarceration. A quarter of parol- ees had a past or present imprisonment for a violent offense on record. Eight

percent had a past or present imprisonment for a sexual offense. Over 20% of parolees in the sample had an official mental health designation. Twenty-two percent were black, 88% were males, and 44% were aged 30 or under. The key

independent variable of interest—supervision intensity—indicates that about 22% of parolees were placed on “high supervision” at release. The descriptive

details of our interaction terms are also presented, showing that 11% of parol- ees in the population were both under high supervision and had previously

been imprisoned for a violent offense, and six percent were under high super- vision and had previously been imprisoned for a sexual offense.

As expected, most of the covariates representing offending background show consistent effects across different outcome types (Table 2). The prior offend-

ing variables perform as anticipated and are compatible with prior research findings (Grattet et al., 2011). Compared with parolees on their first ever release from prison, parolees with multiple prior terms have a 20% higher

hazards for absconding; 2.5 times the hazards for miscellaneous technical vio- lations, drug use/possession; 2.8 times the hazards for violent violations; and

three times higher hazards for sexual violations. Past history of violent and sexual offenses are both associated with lower violation hazards. This also con-

firms the common finding that individuals with histories of serious offending

Table 1 Summary of data

Variable N Percent

Multiple terms of imprisonment 24,680 30.1%

Past or present violent crime 21,048 25.9%

Past or present sexual crime 6,408 7.9%

Age at first incarceration (median) 32 (mean) 30 (median)

Mental health issue 16,588 20.4%

Black 17,836 22.0%

Male 71,747 88.3%

Age 18–30 35,612 43.8%

Age 31–44 35,200 43.3%

Age 45+ 10,479 12.9%

High supervision 14,694 21.6%

× Past or present violent crime 7,509 11.0% × Past or present sexual crime 4,052 6.0%

574 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

T a b le

2 C a u se -s p e c ifi c c o m p e ti n g ri sk

c o x m o d e ls

o f p a ro le

v io la ti o n s (h a z a rd

ra ti o s w it h st a n d a rd

e rr o rs

in p a re n th e se s)

A b sc o n d in g

M is c e ll a n e o u s

te c h s

D ru g u se /

p o ss e ss io n

V io le n t

S e x u a l

M u lt ip le

te rm

s o f im

p ri so n m e n t

1 .2 0 3 ** *

(. 0 2 9 )

2 .5 4 9 ** *

(. 1 2 4 )

2 .5 8 7 ** *

(. 0 8 6 )

2 .8 2 1 ** *

(. 1 8 0 )

3 .0 2 1 ** *

(. 6 2 2 )

P a st

o r p re se n t v io le n t c ri m e

.8 6 5 ** *

(. 0 2 9 )

.8 2 4 **

(. 0 5 3 )

.7 3 9 ** *

(. 0 3 4 )

.7 1 8 **

(. 0 6 4 )

.9 4 6

(. 2 4 7 )

P a st

o r p re se n t se x u a l c ri m e

.6 1 3 ** *

(. 0 6 4 )

.6 1 1 *

(. 1 2 4 )

.4 4 2 ** *

(. 0 7 1 )

.4 2 5 **

(. 1 3 5 )

1 .7 4 8

(. 9 0 1 )

A g e a t fi rs t in c a rc e ra ti o n

1 .0 4 ** *

(. 0 1 0 )

.9 0 0 ** *

(. 0 1 6 )

.9 2 1 ** *

(. 0 1 1 )

.8 9 7 ** *

(. 0 2 1 )

.9 0 1

(. 0 6 3 )

A g e a t fi rs t in c a rc e ra ti o n sq u a re d

1 .0 0 0 **

(. 0 0 0 )

1 .0 0 1 ** *

(. 0 0 0 )

1 .0 0 1 ** *

(. 0 0 0 )

1 .0 0 1 ** *

(. 0 0 0 )

1 .0 0 1

(. 0 0 0 )

M e n ta l h e a lt h is su e

1 .8 0 0 ** *

(. 0 4 4 )

1 .9 4 8 ** *

(. 1 1 2 )

1 .8 2 6 ** *

(. 0 7 2 )

1 .8 9 3 ** *

(. 1 4 5 )

2 .0 3 9 **

(. 4 9 5 )

B la c k

1 .2 5 2 ** *

(. 0 3 0 )

1 .1 2 0 **

(. 0 6 8 )

1 .1 1 0 *

(. 0 4 3 )

1 .1 6 7 *

(. 0 8 8 )

.9 6 9

(. 2 4 8 )

M a le

1 .0 7 7 *

(. 0 3 6 )

1 .3 6 6 ** *

(. 1 0 4 )

1 .4 2 1 ** *

(. 0 7 4 )

1 .5 3 3 ** *

(. 1 6 0 )

2 .6 2 7 *

(1 .1 1 8 )

A g e 1 8 – 2 9

1 .4 6 3 ** *

(. 0 6 5 )

1 .1 6 6

(. 1 0 2 )

1 .3 7 8 ** *

(. 0 8 3 )

1 .3 0 3 *

(. 1 4 9 )

1 .5 0 3

(. 5 5 2 )

A g e 4 5 +

.7 1 8 ** *

(. 0 3 7 )

.8 9 7

(. 1 0 1 )

.7 2 4 ** *

(. 0 5 9 )

.7 4 5

(. 1 2 0 )

.3 6 6

(. 2 0 1 )

H ig h su p e rv is io n

1 .4 2 9 ** *

(. 0 5 1 )

1 .1 1 4

(. 0 9 8 )

.9 9 2

(. 0 5 9 )

1 .0 3 9

(. 1 2 1 )

.6 0 7

(. 2 5 4 )

H ig h su p e rv is io n × p a st

o r p re se n t v io le n t

.9 9 6

(. 0 5 3 )

.9 1 4

(. 1 0 0 )

1 .0 1 2

(. 0 7 9 )

1 .1 5 6

(. 1 7 0 )

1 .9 8 8

(. 4 6 2 )

H ig h S u p e rv is io n × p a st

o r p re se n t se x u a l

1 .1 3 0

(. 1 3 4 )

2 .0 0 9 **

(. 4 4 8 )

1 .7 2 2 **

(. 3 0 9 )

2 .1 9 6 *

(. 7 6 5 )

.7 4 0

(. 4 9 7 )

H ig h su p e rv is io n × b la c k

.9 6 1

(. 0 4 9 )

.7 8 5 *

(. 0 9 0 )

.9 4 1

(. 0 7 2 )

1 .0 0 8

(. 1 0 6 )

2 .1 3 2

(. 9 7 0 )

H ig h su p e rv is io n × m e n ta l h e a lt h

1 .0 5 7

(. 0 5 3 )

.8 6 3

(. 0 9 0 )

.9 9 6

(. 0 7 4 )

.7 4 1 *

(. 1 0 5 )

1 .0 7 4

.4 7 9

N 7 9 ,0 8 1

7 9 ,8 0 2

7 9 ,0 8 2

7 9 ,0 8 2

7 9 ,0 8 2

E v e n ts

8 ,6 2 6

2 ,1 4 1

4 ,5 5 9

1 ,2 3 4

1 1 8

L o g L ik e li h o o d

− 9 1 ,2 1 7 .5 3

2 2 ,4 0 0 .8 7

− 4 7 ,7 0 1 .2 3

− 1 2 ,7 8 6 4 2 0

− 1 ,2 1 3 .2 3

* p < .0 5 ; ** p < .0 1 ; ** * p

< .0 0 1 .

SUPERVISION INTENSITY AND PAROLE OUTCOMES 575

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

are generally less likely to reoffend in the future (Grattet et al., 2011; Kubrin & Stewart, 2006; Wilson, 2005). Here, however, we must note that our model

predicting sexual violations did not identify many significant covariates. Apart from a couple of personal characteristics, which we discuss below, there are

no covariates that achieve statistical significance in the model of sexual viola- tions. This is largely due to the small number of sexual violations that occurred in the data. The 79,082 parolees in the study population committed only 118

violations in which a sexual offense was the most serious charge. Age at first incarceration has significant curvilinear effects. The form is

slightly different for absconding than the other events. Although there is sup- port for the effects of age of first incarceration on violation hazard, as well as

the idea that those effects change over the life course, we cannot fully account for the difference between its effect on absconding and other viola-

tion types. Our findings about the effects of personal and demographic characteristics

on violation hazards reinforces findings from previous research in this area (Grattet et al., 2011; Kubrin & Stewart, 2006; Petersilia, 2009; Wilson, 2005). Mental health status has large effects across the models, including the sexual

violations model. Having a mental health designation increases the hazard of absconding by 80%, miscellaneous technical violations by 95%, drug use/posses-

sion by 83%, violent violations by 89%, and doubles the hazard of sexual viola- tions. Black parolees consistently have higher hazards for all types of

violations than other demographic groups. Males have higher hazards than females, although the differences with respect to absconding are small—men

have between 8 and 53% higher hazards than women. Gender has a particularly large effect on the hazard of sexual violations; males have over 2.6 times higher hazards of sexual violations than females, which fits with the higher

likelihood of males committing sexual offenses in general. Age effects show that, in general, younger parolees have higher violation risks than older parol-

ees, although the differences are not always statistically significant. If our analyses ended here we might contend that it does not really matter

whether violations are subdivided into categories, and that a type-specific competing risks approach is not warranted because the covariate effects are

largely consistent across outcome types. However, the story begins to change when supervision intensity is considered. As the liberation hypothesis would

suggest, intensity indeed functions differently across various violation types, although not exactly in the ways we expected. Being subject to high supervi- sion increases a parolee’s hazard of absconding by 43%. However, supervision

intensity alone does not appear to increase the hazard of “miscellaneous” technical violations or drug offenses—the other minor violations we modeled.

As expected, intensity of supervision mattered little for violent and sexual offenses, although the low base rate of sexual offending might account some-

what for non-significance in that model. While supervision does not have main effects on any violations aside from

absconding, it does have effects when the parolee subject to the high supervision

576 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

has previously been incarcerated for a sexual crime. Sex offenders subject to high supervision have twice the hazards of miscellaneous technical violations

and violent violations. They have 72% higher hazards of drug use/possession vio- lations. The interaction effect, however, is not significant in the model predict-

ing sexual violations. Although not significant, sexual offenders on high supervision actually seem to have lower risks of sexual violations than others.

We also considered whether race and mental health status might interact

with supervision intensity. We found little support for this idea. While there are significant effects of supervision intensity for black parolees in terms of

the miscellaneous technical violations and for parolees with a mental health designation on violent violations, both coefficients are in the opposite direc-

tion of what would be hypothesized by societal reaction perspective. The alternative interpretation might be that when under high-intensity supervision,

these groups of offenders behave more lawfully or are deterred by supervision. However, given the fact that these findings do not emerge in more than one

model, that they are in the opposite direction of our expectations, and that they are significant only at the .05 level, we would recommend caution in such an interpretation.

These findings offer mixed support for the pivotal categories argument made by Grattet et al. (2011). They found that the effect of being a “Second

Striker” on the risk of violation was lowered when supervision factors were controlled, suggesting that supervision acts to enhance the likelihood of viola-

tion detection for certain kinds of violent offenders. In this article, we focus more generically on indicators of past violence (rather than a specific “strike

count”) and find that higher supervision intensity does not increase the hazard of violation. The difference may result from empirical differences in the con- ceptualization of violent offenders in these two articles—that Second Strikers

carry more stigma than generic violent offenders. Folding Second Strikers into the broader category of “prior violent imprisonment” may be sufficient to

erase the effect. Another possibility is that the finding reflects differences between Grattet

et al. (2011) study population and the one used in this analysis. Grattet et al. (2011) included all offenders on parole in California during 2003 and 2004.

Here, we focused just on individuals first released to parole during 2003. The differences between the two study populations is that the present study

excludes many parolees who have “churned” in and out of prison on parole violations, as well as those who were already on parole prior to 2003. Remov- ing these groups likely changes the composition of the populations in ways that

may account for differences between the findings of Grattet et al. (2011) and those of the present study.

Our finding that sex offenders experience supervision effects over a range of violation types lends support to Grattet et al.’s (2011) argument that these

parolees are on a “shorter leash” than others. However, it also adds some nuance to that finding by showing that controlling for other attributes, the

hazard of violation is contingent upon both who the person is (i.e. they occupy

SUPERVISION INTENSITY AND PAROLE OUTCOMES 577

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

the pivotal category of sex offender) and the kinds of violations involved, as supervision effects are only present when sex offenders are accused of techni-

cal violations, drug use or possession, and violent crimes.5

Theoretical implications

Our study’s hypotheses receive partial support from our findings. Overall, we find that in the first two years following release, supervision intensity raises

the risk of absconding violations but not of other violation types. The more closely parolees are monitored, the more likely they are to abscond (or be

caught for it). Thus, we identify limited support for our hypothesis that super- vision effects would be strongest for the least serious violations, as supervision effects were not identified for technical and drug use/possession violations.

We also identify limited support for our hypothesis that supervision effects would be strongest for high-profile parolees—specifically, sexual and violent

offenders. Supervision effects are present for sexual offenders but not violent offenders. We find that sex offenders under high supervision experience ele-

vated risks of technical, drug, and violent violations—providing conditional sup- port for our hypothesis that among high-profile parolees, supervision effects

would be strongest for less serious violations. As we expected, sex offenders under high supervision do experience supervision effects over miscellaneous

technical violations and drug use/possession violations, but they experience supervision effects over violent violations as well.

Existing research shows that the probability of parole failure is highest

immediately following release and declines over time (Petersilia, 2009; Solomon, Kachnowski, & Bhati, 2005). Supervision also tends to be most

intense during the period immediately following release from prison. Our find- ings suggest that this elevated failure rate in California may be the product of

more absconding violations detected through intensive supervision during the first two years of parole. And for sex offenders, high failure rates during this

period is due to increased attention to technical violations, drug involvement, and violence—which may reflect their inclusion within a particularly stigmatized “pivotal category” of offender that demands intense attention (Grattet et al.,

2011; Lin et al., 2010). We did not identify these effects among violent offend- ers, but this may be the result of creating an expansive category containing

any offender with past or present violence on record, potentially “watering down” the observed supervision effect. Sex offenders in this analysis, on the

other hand, appear to retain their pivotal categorization and experience signif- icant supervision effects over a variety of violation types.

5. Recent research on juvenile sex offenders suggest that registration, which increases the surveil- lance directed at the juvenile, shows that the offenders had higher levels of reported “other” offenses (non-sexual) and a slight increase in sexual offenses, but no increase in the risk of convic- tion for sexual offenses (Letourneau et al., 2009).

578 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

Given that parolees are most intensively supervised during the early parole period, and that parole failure is most likely during this time, our findings

underscore the interactive relationships between parolee characteristics, supervision intensity, and parole outcomes. As correlational data, our findings

cannot be used to definitively identify the mechanisms of the relationship between supervision and failure, but two possible interpretations emerge. First, observed supervision effects may reflect shifts in parolee behaviors. Indi-

viduals subject to heightened supervision may experience greater obstacles to re-entry than comparable individuals. They have more conditions placed upon

them, which creates more opportunities for failure and which intensify the stigmas associated with being an ex-offender. This may increase negative self-

labeling or block other opportunities upon release, which in turn raise the like- lihood of absconding from supervision or committing parole violations (for sex

offenders). Relatedly, parolees subject to high supervision may be at higher risk for absconding because they understand that increased supervision inten-

sity involves more onerous requirements (e.g. drug tests, meetings, mandatory treatment) and thus an elevated likelihood of failure. They may abscond rather than waiting to be cited for violations they are likely to commit and be caught

for. Thus, the observed effects might result from parolees’ behavioral changes as they consider their abilities (or willingness) to comply.

Alternatively, our findings may actually reflect “reporting” effects within the parole agency. More intensive supervision allows for more violation behavior to

become visible to formal social control agents. Once made visible, the agents may be more inclined to respond to the behavior, especially for those offenders

already deemed problematic (e.g. sex offenders). This possibility maps onto earlier work on supervision intensity, which has suggested that the more closely that offenders are supervised in the community, the more likely that their viola-

tion behaviors will be detected by social control agents (Grattet et al., 2011; Sirakaya, 2006; Turner et al., 1992). Higher supervision intensity may also be

associated with shifts in institutional orientations toward parolee behaviors. Agents working with high-supervision parolees may be quicker to violate those

parolees for absconding than parolees on lower supervision caseloads. In other words, parole agents may be more aware and less tolerant of the absconding of

high-supervision parolees—perhaps because agents feel that absconding could sig- nal more serious criminal behavior, which may also be related to the reputa-

tional problems parole agents and agencies experience when parolees reoffend in high-profile ways (see Grattet et al., 2011; Lin, 2010). If this were the case, underlying parolee behaviors might not matter very much, as the observed effect

would emerge from the ways that the parole agency monitors and responds to parolees under more intensive supervision.

This possibility, in conjunction with our finding that sex offenders experi- ence elevated violation risks, also invokes an idea developed by Harcourt

(2007) called the “ratchet effect,” which describes the process by which insti- tutional orientations toward certain groups become less tolerant and more

punitive because of the frequent contact between these groups and social

SUPERVISION INTENSITY AND PAROLE OUTCOMES 579

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

control institutions. High-supervision parolees generally have more extensive criminal histories and more serious risk profiles, thereby signaling the need for

more official attention. They are thus subject to lower tolerance for violation behaviors because of institutional beliefs about ongoing risk. But lower toler-

ance leads to more sanctions, and more sanctions lead to even lower toler- ance, thereby “racheting” up punishment over these groups as their risk profiles are “confirmed” through ongoing system contact. From this perspec-

tive, the risk profiles of parolees are partially created by institutional atten- tion, elevating the likelihood of sanctioning in a discretionary social control

environment. Our findings suggest that ratcheting may occur specifically around absconding violations for high-supervision parolees overall, and around

a wider variety of violations (i.e. technical, drugs, violent) for high-supervision sex offenders—perhaps because of institutional or public fears about ongoing

risk. Future research in this area should focus on empirically disaggregating these

two dynamics (behavioral changes and reporting/ratchet effects). Either or both may be at work but regardless, parole violations are clearly shaped by an interaction between individual behaviors, parolee characteristics, and the

orientation of the parole agency toward those behaviors and characteristics.

Policy implications

In real terms, a continued scholarly focus on the ways that correctional super- vision regimes touch down on the lives of offenders and shape the choices

made by official decision-makers seems warranted in light of dramatic changes recently made to California’s correctional system. In 2010, in an effort to curb

the flow of parole violators into prison, the state legislature adopted Non-revo- cable Parole (NRP). NRP resulted in a large share of non-serious, non-violent,

and non-sexual offenders being placed on “banked” caseloads and given no supervision at all. The only condition of parole placed on these parolees was

the continuation of the policy suspending the requirement of reasonable suspi- cion for search and seizure. This meant that NRP parolees could not be returned to prison for failed drug tests, absconding, or other technical viola-

tions of parole because they were no longer subject to those requirements. Thus, for a significant segment of the parole population, supervision intensity

is virtually non-existent. The criminal and violation behaviors of these parolees are not likely to decline under NRP but official attention to those behaviors

will. So violation rates could drop but it must be understood that a drop in vio- lation rates may not actually indicate a reduction in unwanted parolee behav-

iors. If recidivism is a joint production between what offenders do and institutional orientations toward them, then policy-makers must understand

that changing the structure of a supervision regime can have profound effects on official measures of recidivism, even while parolee behaviors remain largely unchanged.

580 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

A second major change to the supervision regime in California is the passage of Assembly Bill 109, also known as the “2011 Realignment Legislation Address-

ing Public Safety” (AB 109). AB 109 removed almost all non-serious, non-violent, and non-sexual offenders from state prison and parole supervision.

Instead, these offenders are now the responsibility of county jails and proba- tion departments. The result has been a dramatic shrinking of the state parole system, leaving it to deal with only high-risk sex offenders, mentally disor-

dered offenders, Second Strikers, and other serious and violent offenders. This reduces parole caseloads and allows for more intense attention to parolees

who remain under state supervision. Although these changes were undertaken to improve recidivism outcomes, closer monitoring of these high-profile parol-

ees may actually result in higher failure rates—especially for absconding. Pol- icy-makers and scholars should understand that because of the interactional

nature of the parole violation process and the importance of official and unof- ficial institutional orientations towards different types of parolees, recidivism

rates only partially reflect real patterns of criminal behavior.

Acknowledgments

Additional support was provided by the UCI Center for Evidence-Based Correc-

tions and the Institute for Governmental Affairs at the University of California, Davis. The California Department of Corrections and Rehabilitation (CDCR),

and particularly the Division of Adult Parole Operations, the Office of Research, and the Board of Parole Hearings provided access to administrative data. Brad Jones provided early advice on the use of competing risks models.

Julie Siebens provided programming to prepare the data for the analyses.

Funding

Funding for the California Parole Study is from the National Institute of Justice (NIJ Award 2005-U-CX-026).

References

Becker, H. S. (1963). Outsiders: Studies in the sociology of deviance. New York, NY: The Free Press of Glencoe.

Bierens, H. J., & Carvalho, J. R. (2007). Semi-nonparametric competing risks analysis of recidivism. Journal of Applied Econometrics, 22, 971–993.

Blumstein, A., & Beck, A. J. (2005). Reentry as a transient state between liberty and recommitment. In J. Travis, & C. Visher (Eds.), Prisoner reentry and crime in America (pp. 50–79). Cambridge, MA: Cambridge University Press.

Box-Steffensmeier J. M., & Jones B. S. (2004). Event history modeling. New York, NY: Cambridge University Press.

SUPERVISION INTENSITY AND PAROLE OUTCOMES 581

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

Cavender, G., & Knepper, P. (1992). Strange interlude: An analysis of juvenile parole revocation decision making. Social Problems, 39, 387–399.

Chen, E. (2008). The liberation hypothesis and racial and ethnic disparities in the appli- cation of california’s three strikes law. Journal of Ethnicity in Criminal Justice, 6, 83–102.

Cicourel, A. (1976). The social organization of juvenile justice. New York, NY: Wiley. Copas, J. B., & Heydari, F. (1997). Estimating the risk of reoffending by use of exponential

mixture models. Journal of the Royal Statistical Society, Series A, 160, 237–252. Drass, K. A., & Spencer, J. W. (1987). Accounting for pre-sentencing recommendations:

Typologies and probation officers’ theory of office. Social Problems, 34, 277–293. Emerson, R. M. (1983). Holistic effects in social control decision-making. Law & Society

Review, 17, 425–455. Escarela, G., Francis, B., & Soothill, K. (2000). Competing risks, persistence, and desis-

tance in analyzing recidivism. Journal of Quantitative Criminology, 16, 385–414. Feeley, M., & Simon, J. (1992). The new penology: Notes on the emerging strategy of

corrections and its implications. Criminology, 30, 449–474. Grattet, R., Lin, J., & Petersilia, J. (2011). Supervision regimes, risk, and official reac-

tions to parolee deviance. Criminology, 49, 371–399. Grattet, R., Petersilia, J., & Lin, J. (2008). Parole violations and revocations in Califor-

nia (Final Grant Report). National Institute of Justice.Retrieved from http://www. ncjrs.gov/pdffiles1/nij/grants/224521.pdf

Harcourt, B. (2007). Against prediction: Profiling, policing, and punishing in an actuarial age. Chicago, IL: University of Chicago Press.

Kalven, H. Jr., & Zeisel, H. (1966). The American jury. Boston, MA: Little, Brown. Kitsuse, J. I., & Cicourel, A. V. (1963). A note on the uses of official statistics. Social

Problems, 11, 131–139. Kruttschnitt, C., Uggen, C., & Shelton, K. (2000). Predictors of desistance among sex

offenders: The interaction of formal and informal social controls. Justice Quarterly, 17, 61–87.

Kubrin, C. E., & Stewart, E. A. (2006). Predicting who reoffends: The neglected role of neighborhood context in recidivism studies. Criminology, 44, 165–197.

Letourneau, E. J., Bandyopadhyay, D., Sinha, D., & Armstrong, K. S. (2009). The influence of sex offender registration on juvenile sexual recidivism. Criminal Justice Policy Review, 20, 136–153.

Letourneau, E. J., Levenson, J. S., Bandyopadhyay, D., Sinha, D., & Armstrong, K. S. (2010). Effects of South Carolina’s sex offender registration and notification policy on adult recidivism. Criminal Justice Policy Review, 21, 435–458.

Lynch, M. (1998). Waste managers? The new penology, crime fighting, and parole agent identity. Law & Society Review, 32, 839–869.

Lin, J. (2010). Parole revocation in the era of mass incarceration. Sociology Compass, 4, 999–1010.

Lin, J., Grattet, R., & Petersilia, J. (2010). ‘Back-end sentencing’ and reimprisonment: individual, organizational, and community predictors of parole sanctioning decisions. Criminology, 48, 759–795.

Lin, J., Grattet, R., & Petersilia, J. (2012). Justice by other means: Venue sorting in parole revocation. Law and Policy, 34, 349–372.

Maltz, M. (1984). Recidivism. Orlando, FL: Academic Press. McCleary, R. (1977). Dangerous men: The sociology of parole. Beverly Hills: Sage. Petersilia, J. (2009). When prisoners come home: Parole and prisoner reentry. New

York, NY: Oxford University Press. Petersilia, J., & Turner, S. (1993). Intensive probation and parole. In M. Tonry (Ed.),

Crime and justice: A review of research (Vol. 37, pp. 281–335). Chicago, IL: University of Chicago Press.

582 GRATTET AND LIN

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16

Reskin, B. F., & Visher, C. A. (1986). Impacts of evidence and extralegal factors on jurors’ decisions. Law & Society Review, 20, 423–438.

Rhodes, W. (1986). A survival model with dependent competing events and right-hand censoring: Probation and parole as illustration. Journal of Quantitative Criminology, 2, 113–137.

Rudes, D. S. (2012). Getting technical: Parole officers’ continued use of technical viola- tions under California’s parole reform agenda. Journal of Crime & Justice, 35, 249–268.

Schmidt, P., & Witte, A. D. (1989). Predicting criminal recidivism using ‘split population’ survival models. Journal of Econometrics, 40, 141–159.

Silver, E., & Miller, L. L. (2002). A cautionary note on the use of actuarial risk assessment tools for social control. Crime & Delinquency, 48, 138–161.

Simon, J. (1993). Poor discipline: Parole and the social control of the underclass, 1890–1990. Chicago, IL: The University of Chicago Press.

Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change and event occurrence. Oxford: Oxford University Press.

Sirakaya, S. (2006). Recidivism and social interactions. Journal of the American Statistical Association, 101, 863–877.

Smith, B. L., & Damphousse, K. R. (1998). Terrorism, politics, and punishment: A test of structural-contextual theory and the ‘Liberation Hypothesis’. Criminology, 36, 67–92.

Solomon, A. L., Kachnowski, V., & Bhati, A. (2005). Does parole work? Analyzing the impact of postprison supervision on rearrest outcomes. Washington, DC: The Urban Institute.

Spohn, C., & Cederblom, J. (1991). Race and disparities in sentencing: A test of the liberation hypothesis. Justice Quarterly, 8, 305–327.

Travis, J. (2007). Back-end sentencing: A practice in search of a rationale. Social Research, 74, 631–644.

Turner, S., Petersilia, J., & Deschenes, E. P. (1992). Evaluating intensive supervision probation/parole (ISP) for drug offenders. Crime & Delinquency, 38, 539–556.

Unnever, J. D., & Hembroff, L. A. (1988). The prediction of racial/ethnic sentencing disparities: An expectation states approach. Journal of Research in Crime and Delinquency, 25, 53–82.

Visher, C. A., Lattimore, P. K., & Linster, R. L. (1991). Predicting the recidivism of serious youthful offenders using survival models. Criminology, 29, 329–366.

Werth, R. (2012). I do what I’m told, sort of: Reformed subjects, unruly citizens, and parole. Theoretical Criminology, 16, 329–346.

Wilson, J. A. (2005). Bad behavior or bad policy? An examination of Tennessee release cohorts. Criminology and Public Policy, 4, 485–518.

SUPERVISION INTENSITY AND PAROLE OUTCOMES 583

D ow

nl oa

de d

by [

C ap

el la

U ni

ve rs

it y]

a t

17 :0

1 15

J ul

y 20

16