 

 

 

 

 

 

 

 

 

 

. warm-.4, .uwl ,

..‘4U..w
I
. 2 . .1

«um.

L

J
A};

3.5.3... 3:
14‘ it: C
9337...}! 2..

us“:

..a

E
A}...(
‘r‘.

a}

.. .-
315.- rain“.
.13???
(f In?”
.
.5... .
.93“?! art.
.:. 2:»...7.
.. .s...

I:

.a 9
n:
It... 5‘
11 .

. ...e
.11....

.31....

:5.
3.1:...

9

.51. \
i}...i....t£.
\ID" .‘|
I“I'3-!
“um".
I .1 .7) 12¢.
.thliuuil .2 3.)!
a. .. ..
int......-..:..A.u$t.

lilvnu’

.21.?!

I:

.34..) .16... .-
. . . .2
. .. . v ’43.
...

:5.

"THESIS

b/
0 IN

This is to certify that the
dissertation entitled

THE (IN)EFFICACY OF REENTRY BASED PROGRAMS:
EXPLORING THE DIFFERENTIAL EFFECTS OF PRISONER
REENTRY DIMENSIONS ON OUTCOME INDICATORS

presented by

Eric L. Grommon

has been accepted towards fulﬁllment
of the requirements for the

Ph.D degree in Criminal Justice

 

 

2‘14? 5%.-

Major Proféssor’s Signature

05 May 2010

 

Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

LIBRARY
Michigan State
Ul liversity

 

 

 

..-.-.-.-.-.-.—.-.-.-.-.-.-.-.-.-._.-.-.-.-.-.-.-.—.-.-.-.-.-.-._.-._.-.-.-.—.

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5108 KzlProyAchreleIRCIDateDue.indd

 

THE (IN)EFFICACY OF REENTRY BASED PROGRAMS: EXPLORING
THE DIFFERENTIAL EFFECTS OF PRISONER REENTRY DIMENSIONS
ON OUTCOME INDICATORS
By

Eric L. Grommon

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Criminal Justice

2010

ABSTRACT
THE (IN)EFFICACY OF REENTRY BASED PROGRAMS: EXPLORING THE
DIFFERENTIAL EFFECTS OF PRISONER REENTRY DIMENSIONS ON
OUTCOME INDICATORS
By

Eric L. Grommon

Correctional systems continue to develop and implement prisoner reentry
programs to ease the process of transition into the community through intervention
across multiple reentry dimensions. Three of the most pressing dimensions include
housing, employment, and substance abuse treatment. In order to inform future
reentry programs, this study explores how these reentry dimensions interact and affect
correctional outcomes of relapse and recidivism. Data used for this study consists of
a sample of 511 offenders with severe substance dependency histories gleaned from a
larger project that assessed the impact of intensive reentry-based programming.

The results indicate that the stability of housing and employment can directly
inﬂuence substance abuse treatment processes and relapse. Employment stability
directly inﬂuences re-incarceration likelihood and housing stability appears to
directly inﬂuence re-arrest likelihood after controls for substance abuse treatment
processes. Housing and employment stability inﬂuence relapse indirectly by shaping
levels of substance abuse treatment dosage and program violations. Employment
stability also indirectly effects re-incarceration likelihood by inﬂuencing levels of

substance abuse treatment dosage and supervision absconds. The ﬁndings highlight

the need to further explore reentry dimensions and lead to a number of theoretical,

methodological, and policy implications.

ACKNOWLEDGEMENT

I would like to thank my committee members Drs. Timothy Bynum, William
Davidson, Edmund McGarrell, Christopher Melde, and Beth Huebner for their
guidance and support during the dissertation process and throughout my graduate
school experience. I look forward to working with each of you as I continue to
develop and hone a career. I am very grateful to have spent my graduate school
career working with Drs. Timothy Bynum and William Davidson. I have gained a
lifetime of knowledge about the nuances of grant-funded, applied research that range
from “war story” absurdities, to dynamic co-productions with stakeholders and
practitioners, to the development of technical report and assistance products. The
lessons learned were not always easy, but the challenges and necessity of providing
timely resolutions have been one of the most rewarding parts of my graduate
education.

Special thanks go out to my mid- to late-twenties. I appreciate all of your
sacriﬁces and the turmoil you have put yourself through. The amount of stress you
have endured over the years has been remarkable and I sincerely hope that the
correlation between stress and life span is proven unfounded. Your run-ins with the
Olin Death Center have been bizarre and will never be forgotten. Thank you to my
social support network of family, friends, and acquaintances. You all know who you
are and I will thank you over cocktails or a few high-ﬁves. Without you (and

according to criminological theory) I would be a craven criminal without a Ph.D.

iv

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................ vii
CHAPTER I: INTRODUCTION ........................................................................................ 1
Problem Identiﬁcation to Response: The Development of Reentry Knowledge....4
Acknowledging the Problems with Available Research on Prisoner Reentry ......... 9
Current Research .................................................................................................... 14
CHAPTER H: REVIEW OF LITERATURE ................................................................... 16
Where Are We? The Contextualized State of Correctional Discourse ................. 16
Programming Heuristics from the “What Works” Literature .................... 20
The Limitations and Applicability of “What Works” to Reentry .............. 26
Theoretical Frameworks of Prisoner Reentry ........................................................ 31
Social Control Emphasis ............................................................................ 33
Reentry Implications of a Social Control Emphasis .................................. 36
Social Development Emphasis .................................................................. 39
Reentry Implications of a Social Development Emphasis ......................... 42
Supervision Emphasis ................................................................................ 45
Reentry Implications of a Supervision Emphasis ...................................... 49
Transitional Emphasis ................................................................................ 50
Reentry Implications of a Transitional Emphasis ...................................... 54
Theoretical Insights across Emphases ....................................................... 57
From Theory to Practice: Research on Prisoner Reentry ...................................... 60
The Salient Sub-Dimensions of Reentry: Housing and Employment .................. 70
The Sub-Dimension of Housing ................................................................ 71
The Sub-Dimension of Employment ......................................................... 78
The Salient Sub-Dimension of Reentry: Substance Abuse Treatment ................. 87
Questioning the Associations between Treatment and Reentry ............................ 93
Summation and Direction of Research .................................................................. 97
CHAPTER III: DATA AND METHODS ...................................................................... 100
Research Questions .............................................................................................. 102
Dependent Variables ............................................................................................ 105
Independent Variables ......................................................................................... 108
Control Variables ................................................................................................. 1 1 1
Analysis Methods and Modeling Strategy ........................................................... 114
CHAPTER IV: RESULTS .............................................................................................. 117
Descriptive and Bivariate Statistics ..................................................................... 117
Research Question 1: Unconditional, Partial Conditional, and Conditional
Multivariate Analyses of Treatment Dosage and Processes ................................ 120
Summation of Results for Research Question 1: Treatment Dosage and
Processes .............................................................................................................. 138

Research Question 2: Unconditional, Partial Conditional, and Conditional

Multivariate Analyses of Relapse and Recidivism Outcome Measures .............. 141
Summation of Results for Research Question 2: Relapse and Recidivism
Outcome Measures ............................................................................................... 157
Research Question 3: Conditional Multivariate Analyses of Relapse and
Recidivism Outcome Measures Controlling for Treatment Dosage and

Processes .............................................................................................................. 160
Summation of Results for Research Question 3: Relapse and Recidivism
Outcome Measures Controlling for Treatment Dosage and Processes ................ 173
CHAPTER V: DISCUSSION AND CONCLUSION .................................................... 180
Study Limitations ................................................................................................. 194
Research Implications .......................................................................................... 200
Conclusion ........................................................................................................... 208
APPENDICES ................................................................................................................. 21 1
Appendix A: Contextual Information on Broader Program Evaluation ............. 211
Appendix B: Veriﬁcation of Multivariate Regression Analyses of Treatment
Dosage and Processes (Research Question 1) ..................................................... 218
Appendix C: Veriﬁcation of Multivariate Regression Analyses of Relapse and
Recidivism Outcome Measures Research Question 2) ....................................... 225

Appendix D: Veriﬁcation of Multivariate Analyses of Relapse and Recidivism
Outcome Measures Controlling for Treatment Dosage and Processes (Research
Question 3) ........................................................................................................... 230
Appendix E: Sobel Test (1986) of Indirect Effects Without the Initial Presence of
a Non-zero Direct Effect Between Independent and Outcome Variables from

Baron and Kenny (1986) ...................................................................................... 235

Appendix F: Alternative Modeling of Interaction Terms for Housing Stability

and Employment Stability .................................................................................... 238
BIBLIOGRAPHY ............................................................................................................ 244

vi

LIST OF TABLES

Table 1: Control Variable Distributions ......................................................................... 118
Table 2: Independent Variable Distributions .................................................................. 119
Table 3: Dependent Variable Distributions .................................................................... 120

Table 4: Partial Linear Regressions of Substance Abuse Treatment Dosage on
Independent and Control Variables ................................................................................. 121

Table 5: Full Linear Regressions of Substance Abuse Treatment Dosage on Independent
and Control Variables ...................................................................................................... 124

Table 6: Partial Linear Regressions of Treatment Program Violations on Independent
and Control Variables ...................................................................................................... 127

Table 7: Full Linear Regressions of Treatment Program Violations on Independent and
Control Variables ............................................................................................................. 129

Table 8: Partial Logistic Regressions of Supervision Absconding on Independent and
Control Variables ............................................................................................................. 133

Table 9: Full Logistic Regressions of Supervision Absconding on Independent and
Control Variables ............................................................................................................. 137

Table 10: Partial Linear Regressions of Substance Abuse Relapse on Independent and
Control Variables ............................................................................................................. 143

Table 11: Full Linear Regressions of Substance Abuse Relapse on Independent and
Control Variables ............................................................................................................. 145

Table 12: Partial Logistic Regressions of Re-Arrest on Independent and
Control Variables ............................................................................................................. 148

Table 13: Full Logistic Regressions of Re-Arrest on Independent and
Control Variables ............................................................................................................. 15 1

Table 14: Partial Logistic Regressions of Re-Incarceration on Independent and
Control Variables ............................................................................................................. 153

Table 15: Full Logistic Regressions of Re-Incarceration on Independent and
Control Variables ............................................................................................................. 155

vii

Table 16: Full Linear Regressions of Substance Abuse Relapse on Independent,
Treatment Processes, and Control Variables ................................................................... 162

Table 17: Full Logistic Regressions of Re-Arrest on Independent, Treatment Processes,

and Control Variables ...................................................................................................... 166
Table 18: Full Logistic Regressions of Re-Incarceration on Independent, Treatment

Processes, and Control Variables ..................................................................................... 170
Table 19: Summary of Adjusted Sobel’s Tests of Indirect Effects ................................ 177

Table 20: Veriﬁcation of Full Regressions of Substance Abuse Treatment Dosage on
Independent and Control Variables ................................................................................. 221

Table 21: Veriﬁcation of Full Regressions of Treatment Program Violations on
Independent and Control Variables ................................................................................. 222

Table 22: Veriﬁcation of Full Regressions of Supervision Absconding on Independent
and Control Variables ...................................................................................................... 224

Table 23: Veriﬁcation of Full Regressions of Substance Abuse Relapse on Independent
and Control Variables ...................................................................................................... 226

Table 24: Veriﬁcation of Full Regressions of Re-Arrest on Independent and
Control Variables ............................................................................................................. 227

Table 25: Veriﬁcation of Full Regressions of Re-Incarceration on Independent and
Control Variables ............................................................................................................. 229

Table 26: Veriﬁcation of Full Regressions of Substance Abuse Relapse on Independent,
Treatment Processes, and Control Variables ................................................................... 231

Table 27: Veriﬁcation of Full Regressions of Re-Arrest on Independent, Treatment
Processes, and Control Variables ..................................................................................... 232

Table 28: Veriﬁcation of Full Regressions of Re-Incarceration on Independent,
Treatment Processes, and Control Variables ................................................................... 234

Table 29: Summary of Adjusted Sobel’s Test of Indirect Effects Without Baron and
Kenny (1986) Clause ....................................................................................................... 236

Table 30: Summary of Full Regressions of Treatment Processes on Independent and
Control Variables with Alternative Interaction Terms .................................................... 240

Table 31: Summary of Full Regressions of Outcome Indicators on Independent and
Control Variables with Alternative Interaction Terms .................................................... 242

viii

Table 32: Summary of Full Regressions of Outcome Indicators on Independent,
Treatment Processes, and Control Variables with Alternative Interaction Terms ........... 243

ix

CHAPTER I: INTRODUCTION
The reentry movement is largely a response to the challenges of mass
incarceration and its emphasis on individualized accountability, incapacitation, and
surveillance (Clear, 2007; F eely & Simon, 1992; MacKenzie, 2006; Useem & Piehl,
2008). It is the realization that punitive “get tough” crime control efforts have been

unable to produce evidence of sustained effectiveness in curtailing future criminal

behavior at the individual level of analysis]. It is also the acknowledgement that the
pendulum between the philosophical orientations underlying correctional policy and
practice is beginning to shift towards rehabilitative or blended rehabilitative-control
efforts (Byme, 2004; Byme, 2008; Byme & Taxman, 2005). Beyond philosophical
orientations, the logic of the reentry movement is closely interrelated with three
problematic trends that have received a substantial amount of focus from correctional
administrators, practitioners, and academics in recent years: the growing prisoner
population, the relatively stable rate of recidivism, and the individualized process of
transition into the community.

Correctional systems are managing populations of prisoners and individuals under
community supervision whose size has never before been seen (Petersilia, 2003). There
are approximately 1.6 million individuals under state correctional authority in 2007,

which is three times the number of those held twenty years ago (Pew Center, 2008). One

 

1This observation is independent of potential incapacitation effects associated with incarceration. There is
some evidence to suggest that incarceration may impact crime rates at the aggregate or macro level of
analysis (Levitt, 2004; Marvell, 2009; Marvell & Moody, 1994; Spelrnan, 2000; Spelrnan, 2005; Vieraitis
et al., 2007). Periods of decarceration appear to be associated with increases in crime (Clear, 2007;
President’s Commission on Law Enforcement and the Administration of Justice, 1967; Vieraitis et al.,
2007), while periods of mass incarceration appear to be associated with decreases in crime (Levitt, 2004;
Marvell, 2009; Marvell & Moody, 1994; Spelrnan, 2000; Spelman, 2005; Vieraitis et al., 2007). While still
subject to debate, the available evidence suggests that one percent increases in prison populations may

in every 100 adults was incarcerated at least once in 2008 (Pew Center, 2008). At the
same time, over 630,000 individuals will be released from correctional institutions and
returned to local communities each year (Mears et al., 2008; Pew Center, 2008). The
number of individuals currently released annually exceeds the total number of those
under state correctional authority nationwide prior to the 1990s (Hughes & Wilson, 2003;
Pew Center, 2008).

Pressure is being placed on community-based corrections to assist in the transition
process (National Research Council, 2008). It is estimated that nearly one in every 32
adults in the United States is under some form of correctional supervision (Pew Center,
2009). There are approximately ﬁve million individuals under supervision, which is
three times higher than populations in the early 1980s (Lurigio, 2005; National Research
Council, 2008; Useem & Piehl, 2008). Methods of release and supervision have become
increasingly less reintegrative with movement towards the abolition of discretionary
parole board systems and the implementation of mandatory conditional supervision
systems based upon risk management assessments and surveillance (F eeley & Simon,
1992; LaVigne et al., 2003; Nelson et al., 1999; Petersilia, 2003; Simon, 1993). There is
even evidence to suggest that some correctional systems have eliminated conditional
supervision altogether (Glaze, 2003; National Research Council, 2008; Petersilia, 2003;
Travis et al., 2001).

Correctional institutions and community-based ﬁeld agencies have been, and
continue to be, the main components of reentry systems (Glaser, 1964; Irwin, 1970;

National Research Council, 2008; Nelson et al., 1999; Petersilia, 2003; Useem & Piehl,

 

reduce crime rates by approximately one-tenth to one-quarter of a percent (Levitt, 2004; Marvell, 2009;
Marvell & Moody, 1994; Spehnan, 2000; Spelman, 2005; Vieraitis et al., 2007).

2008). Available research has indicated that a prison sentence is a short—term occurrence
that signiﬁcantly affects one’s life trajectory. The average prison sentence is just over
two years (Petersilia, 2003). Approximately 93% of the prison population will eventually
be released and approximately 40% of the prison population will be released within 12
months (Beck, 2000; Petersilia, 2003; Travis et al., 2001). Unfortunately, the well-
documented evidence on recidivism suggests a mediocre level of correctional

effectiveness in the reintegration process, which ultimately challenges the legitimacy of

corrections and the larger criminal justice systemz.

Beck and Shipley (1989) concluded that 39% of individuals released from state
correctional institutions in 1983 were re-arrested within the ﬁrst year of their release.
Sixty-three percent were re-arrested within three years and 41% were re-incarcerated
within three years. More recently, Langan and Levin (2002) replicated the study using a
sample of individuals released in 1994 and found that 44% of the sample were re-arrested
within the ﬁrst year of their release, 68% were re-arrested within three years, and 52%
were re-incarcerated within three years. The annual recidivism rate among those under
correctional community supervision continues to hover around 40% (Glaze & Palla,
2005)

Less than half of community correctional supervision discharges are due to the
successful completion of supervision terms (Hughes & Wilson, 2003). The number of re-

paroles following an initial parole term increased 300% ﬁ'om the early 19805 to the early

 

2 The determination of correctional effectiveness based upon recidivism levels is highly debatable. An
alternative incapacitative-control view can be argued, which suggests that correctional systems should be
used as a last resort for those who have continued to be criminally active after being subjected to less
severe sanctions (Rosenfeld, 2008). The relatively high and stable rate of recidivism would therefore
indicate effectiveness in reaching a target population — those that continue to be criminally active after
criminal justice system intervention (Rosenfeld, 2008).

20005, while the number of new admissions to parole has increased 100% over the same
period (Useem & Piehl, 2008). Those who are non-compliant with supervision terms
represent a large portion of institutional intakes (Taxman, 2008). The revolving door
practice of transitions into and out of institutions has serious implications for correctional
administrators and practitioners (U seem & Piehl, 2008), familial and peer relationships
(Braman & Wood, 2003; Eddy & Reid, 2003; Hairston, 2003; Parke & Clarke-Stewart,
2003; Rose & Clear, 2003; Rossman, 2003; Uggen et al., 2004), and the community at
large (Clear, 2007; Rose & Clear, 2003; Rossman, 2003).
Problem Identiﬁcation to Response: The Development of Reentry Knowledge

The changing dynamic of corrections and the relatively stable rates of recidivism
have led to calls for doing something and correctional academics, practitioners, and
policy makers have heeded the call. The number of empirical and non-empirical reentry
studies on the state of reentry knowledge has increased dramatically since 2000 (Lynch,
2006; Petersilia, 2004). Most of the available literature is comprised of three overlapping
types of information: advocacy information, reentry program models or programming,
and connotations concerning the challenges faced upon the transition into the community.
Each type of literature identiﬁes problems inherent to reentry and suggests solutions. A
small, but growing, body of literature within these categories has focused on the
implementation and assessment of programs that can be used to ease the process of
transition (Jacobs & Western, 2007; National Research Council, 2008; Lattimore et al.,
2004; Lattimore et al., 2005; Lattimore & Visher, 2009; Wilson & Davis, 2006;

Winterﬁeld et al., 2006).

Advocacy literature oﬁen exposes a reader to the dynamics of the reentry problem
and attempts to make the difﬁculties associated with the reintegration of prisoners into
local communities visible across multiple points of view (Lynch, 2006). Emphasis is on
the need to shift the focus from piecemeal practices toward continuum of care that blends
any and all components that may assist in the reintegration process (Seiter & Kadela,
2003; Taxman et al., 2002; Visher & Travis, 2003). Pre-release and post-release
planning is viewed as being the most vital component of the process of reentry. Most of
the planning efforts discussed in the literature argue for the need to meet immediate
transitional needs of housing, employment, the provision of treatment/social services, and
the enhancement of generalized social supports (LaVigne & Cowan, 2005).

Informed by the advocacy line of literature is a secondary focus on program
models and programming. Reentry models have been proposed and designed to allow
correctional systems to shift their existing reintegration policies and practices towards a
reentry focus (National Institute of Corrections, 2008; Taxman et al., 2002). A common
structure exists across reentry models and consists of three distinct stages (see Taxman et
al., 2002). The ﬁrst stage, institutional, consists of the assessment and classiﬁcation of
individuals upon admission to an institution and the development of individualized
treatment plans that begin within the institution and are matched to services in the
community. The second stage, structured reentry, begins at least six months prior to
release and continues into the ﬁrst post-release month. This stage consists of intense
preparation for release, the development and maintenance of a reentry plan, the
establishment of connections within the community to meet immediate needs, and

continued treatment or referrals as necessary. By attempting to meet the needs of

prisoners released into communities and streamline the process of transition, this phase is
perceived as being vital for ensuring stability, which may lessen the risk of recidivism
(Lynch, 2006). The third and ﬁnal stage, community reintegration, begins the second
post-release month and continues until community—based correctional supervision is
successfully completed. The focus of the ﬁnal stage is to sustain gains made in other
stages, the maintenance of reentry plans as changes arise, and the achievement of
independence through cooperation with the community.

The three-stage process is assumed to achieve a number of goals. The process
seeks to maintain continuity of care between institutional and community based
programming in an effort to increase structure and stability that can ease the transition
process. Structure and stability are vital to the existing reentry models, but the treatment
and intervention processes inherent to each stage of the model are critical to the transition
process (Listwan et al., 2006). Progression through the successive stages also transitions
from reliance on formal control mechanisms (e. g., police, court, and correctional
operations) to informal social control mechanisms (e. g., family, peers, community and
community groups, and treatment or social service providers). In totality, staged reentry
programs seek to achieve a goal of breaking old criminogenic habits and establishing new
civic roles that enhance the social capital of the community (Taxman et al., 2003; Uggen
et al., 2004).

Federal funding is available to help state correctional systems design and
implement system-wide reentry infrastructures (US Department of Justice, 2008). State
correctional systems receiving funds are generally left to their own discretion in the

design of reentry programs depending on the speciﬁc correctional sub-population they

wish to target (Listwan et al., 2006; Multi-Site Evaluation of SVORI, 2008). The
programs are not entirely standardized, but commonly include intake assessments,
classiﬁcations, programming, reassessments and reclassiﬁcations, in—reach opportunities,
pre—release planning, post-release programming, and structured phase transitions (Multi-
Site Evaluation of SVORI, 2008).

Increased efforts have been made to determine the characteristics of correctional
programming that may produce reliable reductions in recidivism and behaviors associated
with recidivism. The emphasis is not on the existence of programming, but rather upon
the quality of the programming provided (Andrews et al., 1990; Antonowicz & Ross,
1994; Cullen & Gendreau, 2000; Listwan et al., 2006; Lowenkamp & Latessa, 2005;
MacKenzie, 2006; MacKenzie, 2005; MacKenzie, 2000; Palmer, 1983). Numerous
principles and heuristics of efﬁcacious programs have developed and a plethora of “what

works” literature on select components of correctional systems can be found in most

academic and trade journals3 (Cullen & Gendreau, 2000; MacKenzie, 2006; MacKenzie,
2005; MacKenzie, 2000; Visher, 2006). The “what works” terminology has been
identiﬁed as a contemporary paradigm shift within the ﬁeld of corrections (Byme, 2008;
Cullen & Gendreau, 2000; Listwan et al., 2006; Palmer, 1983). Available principles and
heuristics are not without criticism (Antonowicz & Ross, 1994; Byme, 2008; Rhine et al.,
2006; Wilson & Davis, 2006). Research is needed to solidify effective correctional

practices, especially those related to the process of transition into the community.

 

3 “What works,” “best practices,” and “evidence-based” titles have been used interchangeably to explain
synonymous concepts in correctional literature and within the ﬁeld of criminal justice and criminology as a
whole. For the sake of this discussion, the term “what wor ” will be used as an overarching term that
includes best practices and evidence-based designations.

While the quality of intervention is paramount within the growing paradigm of
“what works,” an overarching consideration is the cost effectiveness of programming.
Taxpayers are estimated to spend more than 60 billion dollars annually on corrections
(Gibbons & Katzenback, 2006; Lattimore & Visher, 2009). Considering the additional
“front end” costs of law enforcement efforts and court processing, taxpayers spend
approximately 215 billion dollars on the criminal justice system (Lattimore & Visher,
2009). While most of the correctional population is under some form of community
supervision, one in every 10 dollars spent by correctional systems is for community
corrections while the vast majority is used simply for institutional corrections (Carver,
2004). This equates to approximately $2,000 to $4,000 per parolee per year (National
Research Council, 2008), which is substantially lower than the estimated $25,000 a year
to incarcerate one individual (Petersilia, 2003; Stephan, 2004). Even with the substantial
reductions in cost associated with community based supervision, it is reasonable to
question whether such a small proportion of correctional funding is adequate to transition
individuals into the community.

In combination, advocacy and programming efforts are being developed and
further elaborated based upon the realization of the disadvantages faced by prisoners
returning to communities. This tertiary focus has been a concern in the literature, but has
become more pronounced in the era of mass incarceration. Most individuals enter
prisons with an assortment of needs that are not being met in prison are even more
difﬁcult to meet once released into the community (Useem & Piehl, 2008). Poverty, low
educational background, physical or mental illness, and substance abuse problems are

common hindrances (Petersilia, 2003; Useem & Piehl, 2008). Reductions in funding for

prison programming, community social services, and community correctional supervision
services have corresponded with increased public perceptions of such services being
unnecessary expenditures (Harrison, 2001; LaVigne et al., 2003; Listwan et al., 2006;
National Research Council, 2008; Petersilia, 2003; Seiter & Kadela, 2003; Taxman et al.,
2002; Travis et al., 2001).

The lack of available funding for and attitudes against the provision of services
for individuals stigmatized as being “less than the average citizen” (U ggen et al., 2004, p.
261), has marginalized many systematic attempts to provide reentry assistance. The
current situation has left prisoners facing dilemmas of returning to local communities
having “less treatment, fewer skills, less exposure to the world of work, and less focused
attention on planning for a smooth transition” (Travis & Petersilia, 2001, p. 300). By
implication, prisoners released to communities are more at risk for continued criminal
behavior and recidivism since they are released from a correctional system whose
reintegration mission has deteriorated to such an extent that fewer and fewer transitional
services are available at release relative to services available during admission to prison
(Gibbons & Katzenback, 2006).
Acknowledging the Problems with Available Research on Prisoner Reentry

The focus on reentry brings to the forefront questions of how the ﬂow of prisoners
into and out of institutions affects public safety, how correctional institutions should
manage release, and what communities can do to absorb and reintegrate released
prisoners (Lynch & Sabol, 2001). Equally important are concerns of correctional
efﬁcacy, the function of rehabilitative programs, and considerations of governmental

responsibility in assisting those released from correctional institutions in the acquisition

of resources such as housing and employment (Petersilia, 2003). While there has been
much discussion of the issues surrounding reentry, there is a relative dearth of
contemporary empirical evidence available that can inform theories of prisoner reentry or
the future design of reentry based programs. The reentry movement is growing, but it is
not a well-informed movement.

There are a number of problems that have stunted the development of knowledge
on reentry. The inherent variability of reentry-based programming limits abilities to
deduce conclusions across studies. Some state and local governments have implemented
programming and services speciﬁcally for reentry purposes, while others consider reentry
to be synonymous with community-based corrections or participation in community-
based services (Petersilia, 2003; Solomon et al., 2004b). Moreover, some reentry
services are offered within the institution, others are associated with community-based
supervision, and some rely upon community-based partnerships between correctional
ﬁeld ofﬁces and local service agencies (Wilson & Davis, 2006).

Reentry pro grarnming seeks to ameliorate an assortment of dimensions that affect
transitional experiences. Dimensions of reentry that can be targeted and incorporated
into programming include intra-individual cognitive change, housing assistance,
employment assistance, educational assistance, familial relationships, peer relationships,
and broader relations with the community or community based institutions. Often the
programming focus is placed on select dimensions to the exclusion of remaining
dimensions. Variability in the structure, strategy, and content of reentry programs does

limit the production of knowledge, but it is also necessary in order to meet the needs of

10

an overall correctional population or speciﬁc sub-population (Andrews et al., 1990;
Cullen & Gendreau, 2000; Lipton et al., 1975; Lurigio, 2005; Palmer, 1995).

The funds distributed to local jurisdictions for reentry programming purposes
have become a double-edged sword. There has been an increase in the number of reentry
programs, strategies, and initiatives that may ease the process of transition. Yet, one of
the main criticisms of the study of reentry is the lack of available initiatives and/or
programs that are reentry focused (Travis et al., 2001). The increased focus and
implementation of reentry based programs without a larger understanding of the meaning
of prisoner reentry limits the ability to determine what programs may be the most
effective. Useem and Piehl (2008) noted that research on the process of prisoner reentry
is still in the formative stages and trial and error should be expected for a number of years
before model programs, strategies, or content start to distance themselves from less
effective programs.

One of most pressing problems is the relative lack of available knowledge on the
process of reentry (Visher & Travis, 2003). All too often the discussion of reentry is tied
to measures of recidivism such as re-arrest or re-incarceration after a prolonged follow-up
period. Reductions in crime and increases in public protection are primary concerns of
correctional policy, but the emphasis on recidivism can mask reentry gains or other
improvements in pro-social ﬁinctioning that inﬂuence recidivism outcomes (Listwan et
al., 2006; MacKenzie, 2006; Palmer, 1983; Rosenfeld, 2008; Visher & Travis, 2003).
Gains in social functioning may include the obtainment of housing, employment, and
treatment for medical or mental health needs — all of which have been identiﬁed as key

dimensions to the reentry process (Petersilia, 2003; Rosenfeld, 2008; Taxman et al.,

11

2002; Travis et al., 2001; Visher & Travis, 2003). There is a need to further develop the
understanding that recidivism is a part of the process of reentry and explore the
relationship between reentry dimensions and traditional recidivism outcomes (Rosenfeld,
2008; Visher & Travis, 2003).

There is also a need to explore the differential reentry effects experienced by
speciﬁc correctional sub-populations. An overwhelming majority of prisoners will
eventually be released and reentry is commonly used to describe a reintegrative
experience that is expected to be similar for all types of offenders. What becomes lost in
such a generalized approach is the complex mix of interrelated reentry dimensions that
affect transitions into the community for speciﬁc types of offenders. One of the most
pressing correctional sub-populations with their own unique reintegrative needs are those
with severe substance abuse dependencies.

The enormous grth of state correctional populations has been associated with
drug-involved offenders (Blumstein & Beck, 1999; Carver, 2004; Mumola & Karberg,
2006; Tonry, 1995; Useem & Piehl, 2008) and a majority of prisoners have used or been
under the inﬂuence of drugs preceding their offense leading to incarceration (Mumola,
1999; Mumola & Karberg, 2006). Those with substance abuse and dependency histories
often require comprehensive substance abuse intervention and many will also require
additional mental health services (Compton et al., 2003). Unfortunately, many of those
most in need of treatment will leave prison without it, which places populations of
prisoners with substance abuse and dependency histories at risk for recidivism (Byme,
2008; Byme & Taxman, 2005; Carver, 2004; National Research Council, 2008; Urban

Institute, 2008).

12

There is a reliance on the local community to meet immediate substance abuse
treatment needs (Taxman, 2008; Visher & Farrell, 2005). Many offenders are referred to
community-based treatment, but there is relatively little evidence available on the
effectiveness of this type of treatment intervention (Chanhatasilpa et al., 2000; Cullen &
Gendreau, 2000; Dowden et al., 2003; Lipton et al., 1975; MacKenzie, 2006). It is not
clear how community-based substance abuse treatment affects recidivism. More
importantly, it is not clear if the treatment intervention can directly or indirectly affect the
transition back into the community. Many evaluative undertakings are unable to capture
and control for treatment processes that contribute to and affect observed programmatic
outcomes such as recidivism (Onifade et al., 2008).

This gap in the knowledge — how treatment interventions affect and are affected
by reentry dimensions — also plagues efforts to determine the efﬁcacy of reentry based
programs. There is not a clear understanding as to what circumstances, programs, or
treatment interventions lead to successful reentry outcomes (National Research Council,
2008; Visher, 2006). There is also no clear understanding if reentry based programs can
reduce recidivism or reduce behaviors associated with recidivism (V isher, 2006). Fewer
than 5% of all correctional programs and less than 1% of reentry based correctional
programs are ever subjected to any type of formal evaluation (Listwan et al., 2006;
Petersilia, 2003; Petersilia, 2004; Visher & Travis, 2003). Even among the few reentry
based pro grams that are evaluated, only a handﬁil utilize experimental or quasi-
experimental research designs that allow for valid inferences to be made (F arrington,
2003a; Farrington, 2003b; Lipsey & Cordray, 2000; Onifade et al., 2008; Petrosino et al.,

2000; Rossi et al., 2004; Shadish et al., 2002; Weisburd, 2000). The “black box” effect is

13

also very apparent in the evaluation of reentry based programs. The treatment
intervention processes that affect outcomes are often overlooked (Onifade et al., 2008).
In most cases, there is not enough quality evidence to state that a program does or does
not work as well as the type of participants who would or would not beneﬁt from
programming (Boruch, 1997; Petrosino et al., 2000).

Current Research

The study reported here sought to remedy some of the issues associated with the
current state of knowledge on prisoner reentry. The goal was to examine how speciﬁc
reentry dimensions interact and affect the transitional process. It is still not clear if a
programmatic focus on speciﬁc reentry dimensions can ease the process of transition into
the community. It is also unclear if speciﬁc reentry dimensions can interact with one
another to impede the process of transition and produce unintended consequences for
programs.

The primary purpose of the study was to explore the process of reentry and
determine how speciﬁc reentry dimensions inﬂuenced and were inﬂuenced by other
reentry dimensions. Most pressing were the reentry dimensions of housing, employment,
and substance abuse treatment, which have been identiﬁed as key obstacles to
reintegration that may place individuals at substantial risk for recidivism (Brooks et al.,
2006; Helfgott, 1997; McGarrell etal., 2005; National Research Council, 2008;
Petersilia, 2003; Uggen, 2000; Visher & Farrell, 2005). Borrowing largely from an
overarching evaluative theory framework, this research attempts to un-package the “black
box” of reentry processes and explore the effect of reentry dimensions on treatment

intervention processes and common program outcome indicators. With the grth of

14

reentry programs and the implementation of such models without rigorous evaluation,
this research seeks to inform ﬁrture models by identifying issues that shape program

OUtCOITlCS.

15

CHAPTER H: REVIEW OF LITERATURE

The purpose of this chapter is to review and assess the literature on prisoner
reentry. The chapter begins with a review of the state of correctional discourse on
treatment intervention efﬁcacy. Next will be a discussion of the theoretical frameworks
that underlie prisoner reentry. While limited and piecemeal, the empirical validity of
reentry will follow. The discussion will transition into the salience of three reentry
dimensions: housing, employment, and treatment in the form of substance abuse
services. Finally, the chapter will provide information on the theoretical frameworks that
are associated with substance abuse treatment and highlight the empirical validity of such
efforts. An assessment of the overlap between prisoner reentry and substance abuse
treatment will be made based upon the discussion of theoretical frameworks and
empirical evidence.
Where Are We? The Contextualized State of Correctional Discourse

Almost any discussion that considers the efﬁcacy of correctional interventions in
modifying behavior begins with a mention of the “Scarlett M” (Marlowe, 2006). The
“M” refers to the research conducted by Martinson (1974), which implied that
correctional rehabilitative programs fail to reduce future recidivism. As such,
correctional efforts to rehabilitate were viewed as being ﬁitile and ushered in the era of
“nothing works” correctional philosophies. The work of Martinson (1974) and the
widespread acceptance of his policy implications became an important watershed in the
ﬁeld of correctional programming. In essence, the conclusions shifted correctional
discourse from a generalized acceptance of rehabilitation and individualized arnendability

to intervention towards the dissolution of programs and programming.

16

Yet, the work of Martinson was later anecdotally prefaced ﬁve years later
(Martinson, 1979). Rather than being completely ﬁitile, there was little evidence to
suggest that a speciﬁed type of correctional programming or intervention could work.
Palmer’s (1975; 1983; 1992) replication and re-analysis of Martinson’s ﬁndings
suggested that a number of methodological and program implementation problems biased
ﬁndings toward the null hypothesis that correctional programming would have no effect
on recidivism. The main issue with the research conducted by Martinson, according to
Palmer (1975), was the overarching assessment of correctional treatment interventions as
a whole. The analytic strategy used by Martinson for determining successful correctional
interventions required all of the reviewed studies to produce reductions in recidivism.
This is an insurmountable task, especially in consideration of the heterogeneity of
correctional interventions within speciﬁc subgroups (e. g., individual treatment, group
treatment, community based treatment, etc.) and across an overarching intervention
framework (e. g., substance abuse treatment, mental health treatment, sex offender
treatment, etc.). An alternative analytical strategy focused on intervention components
with some degree of success in reducing recidivism may have revealed that programming
that is innovative or new, with participants amenable to treatment, and with trained and
quality staff may be the most efﬁcacious (Palmer, 1975). The question to be asked is not
what works as a whole, but what works for whom under what circumstances (Palmer,
1975; Sherman et al., 1997).

Palmer (1983) has argued for the necessity of relativity. His differential
intervention position suggests that some interventions may work, some of the time, under

some circumstances. The position arose in the 19805 as a challenge to the nothing works

17

notion and continues to expand today. It is important to note that the nothing works
philosophy has not been replaced. A modiﬁed version of the philosophy has been called
the treatment amenability position (Palmer, 1983).

The treatment amenability perspective maintains that speciﬁc participants will
respond to treatment interventions while others will not. The implications of this
perspective are controversial because they can lead to conclusions that individuals cannot
change or be “ﬁxed” and can lead to calls for selective incapacitation (Pahner, 1983).
Moreover, the treatment amenability position suggests a certain degree of futility in
treatment interventions, since only a few participants will change by self-selection
because they are willing to respond to treatment. The differential intervention position
contrasts this perspective with the suggestion that treatment interventions can produce
positive program effects if the conditions of treatment are correctly suited for those who
are and who are not amenable to treatment. According to Palmer (1983), the reality is
that the efﬁcacy of correctional interventions likely resides somewhere between these
two modes of philosophical thought.

Even with contemporary movement towards rehabilitative ideals, there remain
strong arguments from correctional policy makers and practitioners that treatment
intervention efforts still do not work (F arabee, 2005). Commonly cited evidence in
support of this notion is usually attributed to the problems of inducing individualized
behavioral changes via the criminal justice system (Bean & Nemitz, 2004; Brown et al.,
2004; F arabee, 2005; Prendergast et al., 2002; Sherman, 2007). Additionally, the
observance of stable overall recidivism rates and highly variable rates among speciﬁc

sub-populations that have received some form of correctional programming have also led

18

to implications that correctional interventions may not be effective and should be
curtailed (Farabee, 2005). Rebuttals to these claims commonly argue that correctional
programs and programming are poor on theoretical and practical levels (Marlowe, 2006;
Wilson & Davis, 2006). A “garbage in, garbage out” mentality exists where correctional
policy makers, program designers, and evaluators should expect poor outcomes based
upon poor programs. Feel good programs or politically conscious programs that are
perceived to work are common, despite the fact that many of these types of programs
often do not work (Wilson & Davis, 2006).

The Martinson effect is resilient and continues to shape contemporary correctional
research implications. Marlowe (2006) has suggested that many correctional researchers
selectively look for any explanation that can suggest evidence of a positive program
effect in an effort to combat “nothing works” orientations. At the same time, there are an
assortment of additional correctional researchers who write off a program entirely with a
safe conclusion that poor program implementation explains any and all marginal or
negative program effects in an effort to curtail attempts to make an overall “nothing
works” implication.

Some researchers continue to advocate the claim that nothing works, but preface
the claim by suggesting that such ﬁndings are beneﬁcial and support the replication and
expansion of correctional programs and programming. Toch (1997; 2002) suggests that
many of the null ﬁndings found in research on correctional efﬁcacy imply that the control
or comparison groups involved are doing just as well, if not better, than program
participants in terms of subsequent reformation or behavioral change. This lends some

credence to the notion that correctional intervention in any form (including simplistic

l9

incapacitation) has the ability to alter future behavior. The observation also suggests that
there is an unknown base rate of behavioral change that can be expected with samples of
correctional populations.

Palmer’s (1983) differential intervention perspective has given rise to the
continued development and reﬁnement of “what works” literature in contemporary
correctional research. The focus on contextualized interventions - speciﬁc correctional
sub-populations, in a speciﬁc setting, utilizing a speciﬁc treatment modality consistent to
individualized need — remains the main focus (Andrews et al., 1990; Maruna, 2001;
Sherman et al., 1997). This line of research is an attempt to make corrections more
effective, scientiﬁc, and is in direct response to past notions that nothing works and the
competitive treatment amenability perspective (Corbett, 2008; Harris, 2005; Maruna,
2001; Rosenfeld, 2008). The “what works” perspective relies heavily upon efforts to
enhance program design and implementation. Relevant program outcomes are assumed
to be largely determined by programs and programming. Individuals are assumed to be
amendable and programs must be set up in a speciﬁc fashion in order to induce a given
effect. The underlying philosophies and assumptions of rehabilitative change may still be
important to some degree, but they are secondary to concerns of design and
implementation (Andrews et al., 1990; Byme & Taxman, 2005).

Programming Heuristics from the “What Works ” Literature

Most of the available evidence on correctional program or programming efﬁcacy
suggests that the best programs can be expected to produce 20% to 30% reductions in
recidivism (Aos et al., 1999). Most of the typical programs have slightly less

effectiveness but can still expect to produce 5% to 10% reductions (Aos et al., 1999;

20

Cullen & Gendreau, 2000). This programmatic variability between “best” and “typical”
programs is also manifest with assorted offender populations as speciﬁc program and
individual effects can alter the expected reductions in recidivism (Lowenkamp & Latessa,
2005)

Currently, there is a preference and empirical support for service based programs
that can affect individual level change (MacKenzie, 2005; National Research Council,
2008). Cognitive, behavioral, skill-orientated, or multimodal programming interventions
continue to be the most promising framework from which to design and implement
programs (Andrews et al., 1990; Landenberger & Lipsey, 2006; Lipsey et al., 2001;
MacKenzie, 2005). Programs and programming whose service provision increases social
opportunities among pro-social others and the broader local community also have
support, but to a lesser extent of individual change program or programming models.

Deterrence, incapacitation, surveillance, and other assorted control-based

strategies have been unable to demonstrate sustained effectiveness in reducing

recidivism, even though they are still widely used4 (Byme & Taxman, 2005; MacKenzie,
2005; National Research Council, 2008). There are exceptions. McGarrell and colleges
(2003; Corsaro & McGarrell, 2009; McGarrell et al., 2009) have suggested that strategies
utilizing deterrence-based offender notiﬁcation meetings and social capital-based
referrals to community-based social services upon release can prolong time to recidivism,

but not necessarily reduce levels of recidivism. F arabee (2005) argues for increases in

 

4 This ﬁnding appears to be generalizable to many crime prevention efforts in the ﬁeld of criminology and
criminal justice. Community based efforts in crime control and prevention have documented the relative
ineffectiveness of heavy-handed deterrence and enforcement efforts (Bursik & Grasmick, 1993; Decker,
2003). A preference for service delivery and community building continues to be made as an overarching
intervention model or, at the very least, a component of an intervention strategy (Bursik & Grasmick, 1993;
Decker, 2003; McGarrell et al., 2009).

21

control and surveillance mechanisms within programs in order to produce long-term
behavioral changes. Some research has indicated that combinations of treatment
interventions with enhanced control and surveillance mechanisms can reduce recidivism
in the short term, but the long term erosion of effectiveness is expected (Byme &
Taxman, 2005; McGarrell et al., 2003; McGarrell et al., 2009; Petersilia & Turner, 1991;
Tumer et al., 1992).

Additional generalities of program effectiveness suggest that interventions should
utilize more than one treatment modality and attempt to intervene across multiple deﬁcits
based upon need (Andrews et al., 1990; Cullen & Gendreau, 2000; Listwan et al., 2006;
Palmer, 1983). Treatment should be intense and encompass a substantial period of time
over the ﬁrst year of release (National Research Council, 2008). Cullen and Gendreau
(2000) suggest that services should occupy at least 40% to 70% of an individual’s daily
time over the course of three to nine months. Increasingly a focus has been placed on the
competency and effectiveness of staff delivering services (Palmer, 1983; Wilson &
Davis, 2006). The establishment of working relationships between service providers and
participants and the formation of a generalized supportive atmosphere may prove to be
beneﬁcial. Palmer (1983) was one of the ﬁrst to advocate that the provision of fair and
humane treatment services, in addition to staff competencies, can foster individual
change.

MacKenzie (2006) has recently provided a comprehensive review of correctional
programs, policies, and interventions. The overall ﬁndings suggest that rehabilitative
treatment interventions do work and reinforce the ﬁndings of correctional efﬁcacy ﬁom

prior meta-analyses (see Gaes et al., 1999; Petersilia, 2003). In terms of speciﬁc service

22

frameworks, twelve programs were found to be effective in reducing recidivism. They
include academic and vocational education; Moral Reconation Therapys; Reasoning and

Rehabilitation6; cognitive restructuring, cognitive behavior and behavioral treatment for

sex offenders; hormonal/surgical treatment of sex offenders; Multi-Systemic Therapy for

juveniles7; drug courts; drug treatment in the community, and incarceration based drug
treatment (MacKenzie, 2006, pp. 331-333). An additional twelve programs were found
to be ineffective in reducing recidivism. They include life skills education; correctional
industries; multi-component work programs; psychosocial sex offender treatment;
residential treatment and community supervision for juveniles; domestic violence
treatment using a feminist perspective; domestic violence programs using cognitive
behavioral treatment or arrest interventions; boot camps for adults and juveniles;
intensive supervision; and electronic monitoring (MacKenzie, 2006, p. 333).
MacKenzie’s (2006) identiﬁcation of speciﬁc treatment frameworks is helpful,
but does not capture the variability of program content or service delivery. The
variability of programming effectiveness is largely inﬂuenced by the types of services
delivered to particular offenders in speciﬁc settings (Andrews et al., 1990; Antonowicz &
Ross, 1994; Cullen & Gendreau, 2000; Listwan et al., 2006; Palmer, 1983). One of the
more ever increasing dynamics examined relative to program effectiveness and service

delivery are the principles of risk, need, and responsivity (Andrews et al., 1990; Cullen &

 

5 The therapy is based on cognitive development. The overall goal of the therapy is to improve the
participant’s social, moral, and behavioral deﬁcits (MacKenzie, 2006).

6 The therapy is based upon the modiﬁcation of thought processes by enhancing problem solving and
coping skills (MacKenzie, 2006).

23

Gendreau, 2000; Listwan et al., 2006). The risk, need, and responsivity model of
programming suggests that the most appropriate and effective treatment should target
high-risk offenders, the speciﬁc criminogenic needs of the offenders, and be matched
with the leaming styles of offenders (Andrews et al., 1990; Cullen & Gendreau, 2000;
Listwan et al., 2006). Use of inappropriate treatments -— those that do not target high risk
offenders and cannot cater to the needs or modes of learning for an offender population —
are not likely to produce substantial reductions in recidivism. The key is to design and
implement programs that target predictors of recidivism that can be changed or affected.
One of the prevailing arguments is that services should be reserved for those of
highest risk8 (Lowenkamp & Latessa, 2005). It is often assumed that the delivery of
services to those of highest risk will lead to greater outcome beneﬁts. Low-risk offenders

will likely complete correctional supervision terms without the need for programming or

 

7 The therapy is focused upon the social network an individual possesses (family, peers, school, and
neighborhood). Family members are integrated into the treatment therapy (MacKenzie, 2006).

8 Any discussion of risk — whether that risk be at-risk, high risk, or low risk — begs the question of how risk
assessment determinations are made. The determinations are controversial and can rely heavily on
psychometric properties on one hand and convenience on the other (Byme & Taxman, 2005; Cheliotis,
2006; Corbett, 2008; Feeley & Simon, 1993; Foucault, 1977; Harris, 2005; Simon, 1993; Wacquant, 2001;
Wacquant, 2000). The reliability and validity of most assessment instruments have been well established,
but in practice it continues to be very difﬁcult to predict future behavior (Gottfredson & Gottfredson,
1994). Most risk assessments are associated with static factors derived from an immediate context or an
individual’s history of behavior (Taxman, 2008). Common factors include, but are not limited to, age of
ﬁrst arrest, number of prior arrests, and the number of prior convictions (Taxman, 2008). Need
assessments have arisen recently and are based upon dynamic or ﬂuid factors (Taxman, 2008). Common
factors include, but are not limited to, substance abuse, negative peer associations, dysfunctional families,
and criminal values (Andrews & Bonta, 2003).

Ideally, dynamic assessment procedures should be in place and continually maintained to ensure that
potential participants who ﬁt eligibility criteria or are amendable to treatment are actually placed into
services. Unfortunately, this notion may be nothing more than wishful thinking rather than a policy of most
state correctional systems. Most assessments are made upon intake to an institution and are rarely
replicated with the passage of time. Post-release assessments are not built into most state correctional
systems or shared across state agencies internal or external to the criminal justice system Instead, they are
a function of correctional ﬁeld agencies (personal communication 10 September 2008, Doug Kosinski).
The growing reentry movement is bringing the issue of risk determinations to the forefront and will
continue to modify and enhance determinations with the passage of time.

24

interventions. In fact, the use of programming and interventions among low risk
populations may actually produce undesirable results by draining correctional system
resources and increasing the likelihood for recidivism and criminal behavior (Andrews &
Bonta, 2003; Lowenkamp & Latessa, 2005; Taxman, 2008). Commonly termed an
“iatrogenic effect,” this unintended consequence occurs when low risk offenders are
exposed to intense interventions and high risk offenders partaking in such interventions
(Dishion et al., 1999; Lowenkamp & Latessa, 2005). The causal mechanism that
contributes to this effect is not well understood. Lowenkamp and Latessa (2005) suggest
that the effect pertains to poorly implemented programs with subpar service delivery to
high risk populations. Dishion and colleagues (1999) argue that the effect has more to do
with breakdowns in programming content and goals. The exposure to intense
programming with high risk populations can disrupt the pro-social associations that low
risk offenders can form with individuals and local institutions. The building and
maintenance of such associations are necessary for the prevention of ﬁrture criminal
behavior.

There is some empirical evidence to support the differential effects of high and
low risk offenders. Lowenkamp and Latessa (2005) conducted a review of 10,000

offenders in 53 community-based residential service facilities to determine whether the
effectiveness of a program differed according to determinations of risk9. Overall, the
results indicated that programs were more or less effective depending on the risk proﬁles

of the offenders admitted to the programs. Programs with low risk offender populations

 

9 It is important to note that programming content was not a concern of the study. Each program included
in the sample was considered to be similar and lumped under the umbrella categorization of halfway house
or community-based services.

25

had higher recidivism rates relative to a comparison group who was not subjected to
services, while high risk offender populations had lower recidivism rates relative to the
comparison group. These differential outcomes must be balanced against some of the
participation outcomes. High risk offender populations were found to be less likely to
complete programs compared to low risk offender populations. As such, there is some
indication that recidivism, program participation, and completion outcomes are shaped by
risk determinations.

The Limitations and Applicability of “What Works” to Reentry

While there is growing body of knowledge on the principles and structure of
programs that can reduce recidivism among correctional researchers and evaluators, it is
important to recognize that these principles will not ensure success. Antonowicz and
Ross (1994) have argued that the principles of effective programs are not widely or
consistently found in a review of “what works” literature. Even when the principles are
found, the effects are not clear. Listwan et a1. (2006) suggested that the lack of evidence
and conﬂicting results is due to the fact that the principles have yet to reach the status of
common knowledge or conventional wisdom in the ﬁeld.

The focus on hi gh-risk offenders provides an illustrative example of the difﬁculty
of relying upon the growing principles of correctional programs and programming.
Offenders with the greatest need for treatment intervention are often the least motivated
to partake in programming. The notion of matching learning styles of offenders to
programming implicitly assumes that motivation for treatment can be gained, which may
not be the case for many offenders and especially those considered high risk. Low risk

offenders have often been the preferred population to work with among program

26

providers since the population is easier to manage with few problems and established
connections to pro-social individuals and institutions in the community (Byme &
Taxman, 2005; Lowenkamp & Latessa, 2005).

Programs and programming that target high risk populations become precarious
interventions. It is likely that these programs will be judged negatively by program
administrators and key stakeholders due to relatively high recidivism rates in the short
term (Byme & Taxman, 2005; Listwan et al., 2006). Programs that target high risk
populations will also require longer lengths of programming duration to overcome
deﬁcits in motivation, resistance, and resentment associated with participation placement
before any rehabilitative effect can occur (Wilson & Davis, 2006). The amount of time
and resources needed for high risk populations can lead to costly programs and
unrealistic expectations of timely results that prove to beneﬁt participants.

Added concerns are ﬁscal and organizational pressures among the consumers of
research on correctional programs and programming (Byme, 2008; Rhine et al., 2006).
Correctional policy makers and administrators need immediate information, which
systematically overlooks longitudinal programs that may possess the most valid and
reliable outcomes (Byme, 2008; Zhang & Zhang, 2005). At best, the slightly marginal to
positive effects that could be expected with interventions targeting high risk populations
can fulﬁll a utilitarian notion of success for program administrators and research partners.
That is, the success of a few of the most difﬁcult participants is likely to reduce far more
subsequent criminal behaviors than a low overall recidivism rate with an easier, more
amendable low risk population. Unfortunately, the recidivism rate continues to be the

bottom line for nearly all correctional programs (Maltz, 1984; Maruna, 2001). Evidence

27

of high recidivism rates, even in the short term, often leads to claims by correctional
policy makers and administrators that a speciﬁc program for high-risk participants does
not work. This conclusion may subsequently lead to the discontinuation of federal and/or
state funding and the dissolution of a program.

The public at large also has a stake in correctional programs and programming,
which is a signiﬁcant source of organizational pressure. There is a necessary balance of
providing services while protecting the public and retaining organizational legitimacy
(Byme & Taxman, 2005). Programs and programming can ﬂuctuate depending upon
public sentiment with intensive services being provided up until a well-publicized fall out
that requires an immediate change in correctional operations (see DiIulio & Piehl, 1991;
Newburn & Jones, 2005).

The “what works” discourse continues to expand and become integrated into the
language of correctional policy makers, administrators, and practitioners. At the same
time, there is no real consensus on how resources, programs, and programming efforts
should be structured to ease transition into the community immediately following release
from an institution (Wilson & Davis, 2006). A deﬁnitive a priori model of reentry does
not exist (Lattimore et al., 2005). One must be critical of any reliance on “what works”
literature to understand reentry (Maruna, 2001). Byme (2008; 2009) even goes as far to
suggest that no one can claim that there is a “what works” foundation for reentry since
there are no systematic reviews of reentry programs to date. All that exists to date are

reentry program models, implementation reviews of select reentry programs, and

28

borrowed “what works” reviews that may or may not be applicable to reentry10 (Byme,
2009)

There are a number of concerns related to the development of “what works”
literature that are problematic for research on reentry. First, there are issues of
deﬁnitional and conceptual ambiguity. Existing reviews often lump correctional
programming into one overarching construct often considered to be rehabilitation (or
some variant of rehabilitation) (see Lipton et al., 1975). This generalized classiﬁcation
ignores the fact that programs may have conﬂicting theoretical assumptions that cancel
out one another when included in analyses to decipher effective correctional efforts from
ineffective correctional efforts. For instance, a correctional program based on vocational
training is essentially pro-social since the program enhances the human capital of the
participant. Another correctional program may be an intervention in which a participant
is required to attend in-service training as a result of a positive drug screen. The
intervention can enhance human capital, but it is required as part of the punishment

process. When the program and the intervention are included in a rehabilitation review

 

10 Seiter and Kadela (2003) ﬁt within the notion of borrowed “what wor ” reviews. In their review of
what works, what does not, and what is promising for prisoner reentry, the researchers do not focus on
reentry programs or programming. Rather, the focus is placed on an assortment of programs or
programming that has been used in corrections that may or may not be consistent with prevailing reentry
models. Vocational training and/or work programs were found be effective in reducing recidivism and
increasing job readiness. Drug treatment, halfway house, and pre-release programs were also found to be
promising in reducing recidivism Educational programs were found to increase standardized achievement
scores. Each of these categories of programming type has some empirical support individually, but not in
an overarching reentry context that would combine all of the programming types in one model.

In some respects, MacKenzie (2006) is guilty of a similar extrapolation to reentry programs with a brief
discussion of how speciﬁc programs — notably cognitive-behavioral, drug treatment, vocational, and
employment programs — were found to be effective in reducing recidivism and should be considered as
components of reentry models. What differentiates MacKenzie (2006) from Seiter and Kadela (2003) is
the preface made by MacKenzie (2006) that additional research is needed into reentry models to determine
how speciﬁc programming types inﬂuence and are inﬂuenced by reentry concerns of housing options,
employment opportunities, and familial reuniﬁcation. Reentry model developers should consider available
“what works” evidence, but further research is necessary before making “what works in reentry” claims.

29

the underlying assumptions of each (i.e., skill building and punishment) are masked, and
the effect sizes can cancel one another out and produce unreliable outcomes.

In direct relation to the ﬁrst concern, the “what works” literature cannot inform
researchers of the underlying “black box” dynamics of programs that are determined to
be effective. The ﬁndings generally provide the exact name (e. g., Multi-Systemic
Theory) or categorize (e. g., academic programs) a program and mention the basic
theoretical framework used. Implementation information is not provided and there is an
underlying assumption that the rehabilitation programs have been implemented in accord
to the program model. A second problem with simpliﬁed titles or categories of programs
is the notion that most rehabilitative programs have multiple dimensions (Palmer, 1995).
The focus on academic programming, for instance, may also include drug treatment and
family reuniﬁcation dynamics. These sub-dimensions are not apparent with a focus on
the primary rehabilitation mechanism.

Third, the dominant focus in the “what works” literature is on recidivism. The
prediction of recidivism has been considered to be one of the most widely studied
phenomenona in criminology and criminal justice (Maltz, 1984; Mauma, 2001). In
practice, recidivism is generally a bottom line measure used by correctional
administrators, practitioners, and policy makers to determine if taxpayer monies should
continue to fund speciﬁc programs (Maltz, 1984). While widely examined and used,
measures of recidivism often cannot identify if direct or indirect societal gains were made
while in the program. Beneﬁcial results such as mentorship, expanded social ties,
education, employment, cognition, and drug abstinence can be overlooked with a strict

focus on recidivism.

30

These concerns reinforce the notion that prisoner reentry does not (and cannot) ﬁt
into the ﬁndings derived ﬁ'om the “what works” literature. One cannot assume that what
will work for rehabilitative programs and programming will work for reentry. At the
same time, these concerns provide a foundation for generating knowledge on the
processes and events that surround prisoner reentry. There is a need to critically assess,
specify, and understand the dynamics of reentry programs how such dynamics shape
relevant programmatic outcomes. There is also the need to reconsider the relevance of
recidivism in a reentry context to understand how speciﬁc dimensions of reentry can
affect program processes and subsequently affect levels of recidivism. The following
section seeks to provide some indication of what is known about prisoner reentry.
Theoretical Frameworks of Prisoner Reentry

Much of the available research on reentry is atheoretical and relies upon post hoc
interpretations of reentry events and processes that may or may not involve correctional
programming (Lattimore et al., 2005; Lynch, 2006; Maruna eta1., 2004a). Reentry
research often relies on a patchwork of existing criminological theories that seek to
explain an assortment of criminal behaviors and include a variety of assumptions about
human nature and the underlying conditions that may foster crime. It is not clear how
applicable existing criminological theories are related to the issue of reentry. It is also
not clear if the theoretical frameworks used to inform reentry processes and events are
integrated or if they can be integrated.

Often, there is an implicit assumption that reentry programming will positively
beneﬁt participants regardless of the underlying theoretical framework that informs

program design and content. The logic behind this assumption suggests that the

31

provision of any type of programming will assist participants more than no programming
at all. It is equally possible that the existing theoretical frameworks, program principles,
and program content used in reentry programming diverge from one another to such an
extent that the anticipated beneﬁts for participants will be negated and produce null
program effects. There is also the possibility of unintended consequences that may
subject program participants to more harm than good (McCord, 2003).

Without a distinct theoretical emphasis for reentry guiding the development of
policies and practices it is difﬁcult to interpret results and identify the components of
programming that contributed to the observed results (Harris, 2005; Lattimore et al.,
2005; Taxman, 2004). It is even more difﬁcult to develop policy and practice
implications from the observed results to inform future programs without a theoretical
foundation (Harris, 2005; Lattimore etal., 2005; Taxman, 2004). This situation is
especially problematic in today’s correctional environment where state correctional
systems are looking for immediate methods to manage the ever-increasing population of
offenders reentering society.

Since there is a relative lack of reentry speciﬁc theories draw upon (and an
equally important dearth of information on the empirical validity of reentry
programming), it is likely that many correctional systems formulate and implement
reentry programming in a reactionary (e. g., response to growing populations, to be
released populations, and community correctional populations), mimetic (e. g., some
states correctional systems have developed reentry-based infrastructures, the current state

should as well), or evolutionary (e. g., existing policies, practices, or programs include a

32

new reentry dimension or component) manner1 I. The real danger in proceeding in this
manner is the likelihood that reentry programs will be viewed as being ineffective and
inefﬁcient. In turn, the evidenced ineffectiveness and inefﬁciency has the potential to
cease reentry movements altogether and shift the correctional continuum towards policies
and practices that are less reintegrative and more incapacitative.

Lynch (2006) suggested that in order to develop knowledge on reentry, the
criminological ﬁeld needs to deconstruct individual studies of speciﬁc reentry programs
and attempt to synthesize and extrapolate ﬁndings across studies. In order to accomplish
this goal, Lynch (2006) argued for the need to identify the driving (implicit or explicit)
theoretical frameworks that are associated with reentry events, processes, and programs.
He suggests that reentry programs are largely shaped by four speciﬁc emphases. They
include those that emphasize social control, social development, methods of supervision,
and the immediate transitional experience from prison into the community. It is
important to note that these categorizations are not mutually exclusive. Instead, there is
likely to be a substantial amount of overlap and interaction between the categories and
their theoretical claims. While purported to be “crude” (Lynch, 2006, p. 405), these
categorizations serve as a useful heuristic ﬁom which to develop a foundation of
systematic knowledge on reentry. Each of the respective emphases will be discussed
separately.

Social Control Emphasis

Social control theory suggests that individuals partake in criminal behavior

because their social bonds to conventional others and social institutions are weak, broken,

 

H Similar lines of reasoning can be found within generalized organizational literature (Donaldson, 1999)
and organizational literature that is focused upon the operations of criminal justice agencies (Katz, 2001;

33

or non-existent (Hirschi, 1969; Reiss, 1951). It is the lack of restraint (or guardianship)
in one’s social environment that fosters criminal behavior, rather than underlying forces
that motivate criminal behavior. By implication, the social control perspective maintains
a relatively negative view of human nature. Individuals are assumed to possess an
inherent drive to participate in criminal behaviors and activities. If left uncontrolled by
formal or informal social controls, criminal behaviors and activities are a likely product.
Social bonds are shaped by social ties, networks, and institutions based in one’s
social environment and posited to impart the normative values and goals of conventional
society through a process of internalization (Hirschi, 1969; Reiss, 1951). Each individual
is assumed to possess a variable quantity of existing social bonds and opportunities to
formulate future bonds. The potential for bonds to be more or less available are
contingent upon one’s prior behaviors, current situation, or stage in life (Laub &
Sampson, 1993; Matza, 1964; Sampson & Laub, 1993). It is not clear how the causal
process of internalization is developed, but there is some evidence to suggest that it is the
quality of social bonds (rather than the quantity of social bonds) that can affect behavior.
Hirschi (1969) suggests that the strongest and most effective social bonds are
those that foster individualized belief in the norms and values of conventional society and
consist of elevated levels of attachment to social ties and networks, commitment to
conventional society, and involvement in conventional activities. Family members are
often relied upon as being vital social bonds for offenders transitioning into the
community after release (Berg & Huebner, 2009; LaVigne et al., 2004; Mallik-Kane &
Visher, 2008; National Research Council, 2008; Nelson et al., 1999; Petersilia, 2003).

Increasing evidence also suggests that speciﬁc social bonds, such as marriage and

 

Katz et al., 2002).

34

employment, can signiﬁcantly reduce levels of criminal behavior and even terminate
future criminal behavior independent of prior criminogenic involvement (Giordano et al.,
2002; Homey et al., 1995; Huebner, 2005; Laub & Sampson, 1993; Sampson & Laub,
1993; Shover & Thompson, 1992; Warr, 1998). While growing in popularity, the
bonding efﬁcacy of marriage and employment continues to be questioned. There are
counter instances in which individuals remain single and unemployed and fail to re-
offend, particularly when analyses include female samples (Giordano et al., 2002).

Social bonds can provide structure, stability, and accountability to daily life
(Maruna et al., 2004a). Bonds can also foster lifestyle changes (Shover & Thompson,
1992). To achieve this level of functioning requires ongoing management of inter-
relationships and meeting relational expectations. In this sense, social bonds are a
commodity that can be gained or lost.

Sampson and Laub (1993) follow a similar logic in their discussion of the age-
graded theory of informal social control. Borrowing heavily from Coleman’s (1988)
conceptualization of social capital, investments in pro-social relationships and
participation in conventional activities can promote one’s own conventional goals and
provide exposure to conventional others that can build one’s skill set and resource
potential. The more commitment made to conventional behavior, the more likely one
will gain the upward mobility to participate and compete in society (Girodano et al.,
2002)

Participation in criminal behavior attenuates the development of pro-social bonds
and the resource potential that can be derived from bonds. Criminality comes at a cost as

conventional social bonds negatively react to such behavior (via shaming, informal

35

punishment, or withdrawal/distancing) (Homey et al., 1995; Shover & Thompson, 1992).
In turn, an individual is likely to continue involvement in criminal behavior as outlets for
the development of pro-social bonds become less available (Sampson & Laub, 1993).
This notion of the cumulative loss of bonds is hypothesized to explain why criminal
behavior can often become stable over time.

Reentry Implications of a Social Control Emphasis

Reentry programs and programming recognize the importance of building pro-
social bonds and inducing participation in conventional activities to curtail ﬁrture
criminal behavior. The task for reentry programs and programming is to affect bonds and
build social capital. Many of the existing reentry program models include planning and
programming that will assist in the reparation of prior social bonds or the building of new
social bonds (see Multi-Site Evaluation of SVORI, 2008; National Institute of
Corrections, 2008; National Research Council, 2008; Taxman et al., 2002). Much of the
effort is placed on the assessment of individualized need and the identiﬁcation of speciﬁc
strengths or deﬁcits from which to build from. Irnportantly, however, the focus of
intervention is only indirectly placed on the individual. Much more of the intervention
emphasis is directed towards the individual’s external environment — one’s social ties,
networks, and institutions.

It is not clear if interventions can or should reach external environments. For
example, there is mixed evidence to suggest that family reuniﬁcation efforts can ease the
process of reentry (Braman & Wood, 2003; LaVigne et al., 2004; Naser & Visher, 2006;
National Research Council, 2008; Nelson et al., 1999; Parke & Clarke-Stewart, 2003;

Petersilia, 2003; Western et al., 2004), even though families can provide immediate

36

avenues for the development of social capital (Berg & Huebner, 2009). Some families
may be supportive and involved, other families may refuse to be involved, and some
individuals may not want their families subjected to programming that is tied to their past
criminal behavior (Braman & Wood, 2003; LaVigne et al., 2004; Naser & Visher, 2006;
Nelson et al., 1999; Parke & Clarke—Stewart, 2003; Petersilia, 2003; Western et al.,
2004). Efforts to involve one’s external environment are also controversial. Post-modem

criminological theorists have argued that such efforts extend the reach of punishment

beyond the individual who was convicted of a crime to law abiding citizens12 (Foucault,
1977; Simon, 1993; Wacquant, 2001; Wacquant, 2000).

Reentry programs and programming that utilize a social control emphasis assume
that intervention can affect social bonds. The effect is largely driven by the assumption
that interventions can provide exposure to sources of social bonds that may increase the
quantity of subsequent bonds one possesses. The intervention may in and of itself serve
as a social bond that can alter prior criminal behaviors (Laub & Sampson, 2003;
MacKenzie & Brame, 2001; Palmer, 1994; Pahner, 1995; Sampson & Laub, 1993).
There is some evidence to suggest that the bonding experience between participants and
their service providers can enhance pro-social behavior modiﬁcations (MacKenzie &
Brame, 2001; Palmer, 1994; Palmer, 1995), especially when paired with a quality social
bond with one’s correctional supervision agent (MacKenzie & Brame, 2001; McCleary,

1978; Nelson et al., 1999; Palmer, 1994; Palmer, 1995). These insights form a relatively

 

12 . . . . . . .
The very notron of rnforrnal socral control 15 F oucauldran. Informal socral control assumes that socral
others will provide surveillance and inforrml punishments for non-normative behavior. These mechanisms

of control become ingrained and are posited to control ﬁrture behavior.

37

positive view of programming, which contrasts with the negative view of human nature
inherently held by the theoretical perspective.

The feasibility of affecting social bonds through programming is not without
question. Interventions may provide exposure to bonding sources and increase the
quantity of bonds, but the development of quality social bonds is inherently subjective
process and is contingent upon the interpersonal skills of an individual. By the
theoretical tenants of the social control perspective, prisoners reentering society are

expected to lack available social bonds that may have controlled prior criminal

behavior”. Without a foundation from which to build, it is likely to be difficult to
formulate a pro-social network.

The incarceration experience itself is also likely to constrain and damage the
utility of one’s existing bonds even if they are available prior to prison (Braman & Wood,
2003; LaVigne et al., 2004; Nelson et al., 1999; Parke & Clarke-Stewart, 2003; Petersilia,
2003; Western et al., 2004). Upon release, the pathway to building pro-social
relationships does not become much easier. At the individual level, prisoners returning to
local communities are likely to be poor, lack education, lack conventional social
networks and ties, are disconnected from their families, have suffered abuses, and have
substance abuse dependencies or histories (Maruna, 2001; Maruna et al., 2004a; Nelson
et al., 1999; Petersilia, 2003; Travis et al., 2001). The communities to which they will
return generally have few employment and conventional opportunities, easy access to

drugs, high amounts of crime, increased law enforcement presence, and unforgiving

 

13 One of the many criticisms about the social control perspective is its tautological nature: an individual
engages in criminal activity because they lack social bonds and an individual is criminal because they lack
social bonds. The perspective can be used to explain future criminal behavior (i.e., onset) and explain
continued criminal behavior (i.e., continuity or stability of behavior).

38

community members (Maruna, 2001; Maruna et al., 2004a; Nelson et al., 1999; Petersilia,
2003; Travis et al., 2001). Equally problematic is the stigma associated with a criminal
record. Once in the community, individuals are systematically excluded from many
activities that would provide access to conventional society and increase levels of social
capital. The presence of criminal history records can hamper one’s ability to obtain
housing (Brooks et al., 2006; Grommon & Devitt, 2003; Helfgott, 1997; Maruna, 2001;
Metraux & Culhane, 2004; Nelson et al., 1999; Pager, 2003; Pager, 2007; Petersilia,
2003; Roman, 2004; Travis et al., 2001; Visher & Farrell, 2005; Visher & Travis, 2003)
or employment (Grommon & Devitt, 2003; Holzer et al., 2004; Nelson et al., 1999;
Pager, 2003; Pager, 2007; Petersilia, 2003; Uggen, 2000; Visher et al., 2005; Western et
al., 2001; Western & Pettit, 2005), limit domestic and international travel (Grommon &
Devitt, 2003), and prohibit participation in civic duties (Nelson et al., 1999; Petersilia,
2003; Uggen et al., 2004; Uggen & Manza, 2004).
Social Development Emphasis

The social development emphasis of reentry programs and programming overlaps

with the social control emphasis to an extent and there have been calls to integrate the

two perspectives” (Bottoms et al., 2004; Burnett & Maruna, 2004; Farrall & Bowling,

 

14 Calls for integration and the acknowledgment of the interaction between these two perspectives have
been made in the growing literature on desistance. Desistance has been and continues to be an important
consideration within corrections, criminal justice, and criminology as a whole. As a concept, desistance
refers to the observation that most offenders will stop offending at some point in time (National Research
Council, 2008; Laub & Sampson, 2001). In practice, there is relatively little knowledge on the mechanisms
that can induce desistance, whether that desistance is deﬁned as complete termination of criminal offending
or short-term cessation of offending (National Research Council, 2008).

What is known about desistance is that it is an encompassing, longitudinal process that leads to a lessened
frequency of offending (Laub & Sampson, 2003; Maruna, 2001 ; National Research Council, 2008; Loeber
& LeBlanc, 1990). The focus is generally retrospective or bibliographic with assessments of offending
frequency and variability across age-crime curves. Arrest, conviction, institutionalization, and the

39

1999; Giordano et al., 2002; Maruna, 2001; McNeil], 2006). Similar to social control, the
social development emphasis is concerned with trajectories of behavioral change. The
emphasis under the social control perspective is how the quantity, quality, and timing of
social bonding opportunities and experiences insulate one from ﬁrrther criminal
behaviors. By contrast, the social development perspective focuses on one’s own self
narrative, cognitive belief system, or identity. One will cease or reduce levels of
participation in criminal behavior once one has cognitively convinced themselves that a
criminal lifestyle is no longer the type of life they would like to live. Maruna et al.
(2004a) suggested that this cognitive intra-individual change must occur prior to the
development of pro-social bonds. As such, the social development perspective
establishes some degree of temporal ordering by suggesting that intra-individual change
must precede any efforts to induce pro-social change via program or programming
intervention.

The social development perspective is largely informed by the work of Maruna
(2001; Maruna et al., 2004a) and Giordano, Cemkovich, and Rudolph (2002). Maruna
(2001) suggested that individual offenders must make the choice to change and accept a

new crime-free lifestyle. The major difference between those who return to crime and

 

transition into the community are components of the broader desistance process and may or may not
contribute to future offending.

At the same time, research on desistance involves more than just the actions and processing of the criminal
justice system. Often the focus is on pro-social relationships and normative behavior. Immediate and
extended family (Berg & Huebner, 2009; Haggard et al., 2001; Nelson et al., 1999), peer networks
(Giordano et al., 2003; Warr, 1998), marriage (Giordano etal., 2003; Homey et al., 1995; Huebner, 2005;
Laub et al., 1998; Laub & Sampson, 2001; Piquero et al., 2002; Sampson & Laub, 1993; Sampson et al.,
2006), and employment (Berg & Huebner, 2009; Huebner, 2005; Sampson & Laub, 1993; Uggen, 2000)
have all been identiﬁed as components of the broader desistance process that may lessen subsequent
criminal activity. These components are directly related to the theoretical perspectives of social control and
social development in addition to the literature on the age-crime curve. There is merit to theoretical
integration, however, the current discussion seeks to highlight the inherent issues to any solitary focus

40

those who desist from future criminality are the scripts and sense of self that one creates.
Borrowing heavily from the social-psychological concept of syrnbolic-interactionism
(Blumer, 1969; Morris, 1962), the social development perspective assumes that
individuals can manage their perceptions of self and their identity. Often this
management of self perceptions is accomplished via the reciprocal relationship between
an individual who affects and is affected by their larger social environment (Blumer,
1969). Those who are ready to desist develop and operate upon a sense of self that is
good and conventional. The old criminal sense of self is shed. This process of active

internal self change is opposed to perspectives that assume that individuals are passively

molded by external social mechanisms”. Maruna (2001) does not argue that social
mechanisms are unimportant. Those individuals who are ready to desist from offending
identify with a particular community, group, or cause and attempt to contribute the most
they can to their community, family, and group. The process of selective social
identiﬁcation can reinforce the new self and inﬂuence future behavior. What is important
is the ordering of the process. Returning prisoners must ﬁrst convince themselves that

they are ready to desist before becoming integrated to the community.

 

(prior to and independent of integration). Moreover, the focal concern of the research is the reentry
experience, which is a microcosm that may or may not shape the broader desistance process.

15 One area where literature on the social development perspective is particularly silent concerns the issue
of reinforcement of criminal or offender status. Conditions of community based correctional supervision
often include periodic meetings with supervisory agents, assorted tasks (e.g., drug testing, treatment
referrals, etc.), and graduated sanctions for non-compliance (e.g., temporary conﬁnement, tether placement,
revocation center placement, etc.) that may reinforce one’s status as being an offender. The social stigma
of being an offender also serves as an important agent of reinforcement. The extent to which these external
social mechanisms of status reinforcement inﬂuence the creation of a conventional identity needs to be
developed and explored further, especially within a reentry context. Some of the available evidence
suggests that criminal and conventional identities can be managed depending on the situation (Goffrnan,
1959; Goffrnan, 1963; Harding, 2003; Maruna et al., 2004b), but it is unclear how identity management
interacts with reentry programs or programming to ease or hinder the transition process.

41

Giordano et a1. (2002) elaborated the work of Maruna (2001) in their development
of the theory of cognitive transformations. The focus is placed once again on the
individual, with speciﬁc emphasis on human agency, and the reciprocal relationships one
has with their larger social environment. Paralleling Maruna (2001), individuals specify
relationships to form in their social environment to form a new identity. In seeking to
change behavior, individuals become opportunistic by latching onto speciﬁc catalysts for
change (conceptualized by the term “hooks for change”) (Giordano et al., 2002, p. 1000).
Once these catalysts for change are identiﬁed, one must proceed through a process of
internal cognitive change before behavioral change can follow.

Four elements of cognitive change are identiﬁed. First, an individual must be
open and ready to change their identity and behavior. Second, an individual must
perceive the opportunity for change as being worthwhile, salient, and meaningful.
Giordano et a1. (2002) argue that this element is one of the most important aspects of
cognitive change. Openness to change needs to be followed by an inherent connection to
an agent of change. Third, an individual must be able to shift identities by viewing
oneself in a manner that will allow the old self to be disregarded. The ﬁnal element of a
cognitive transformation concerns the meaningfulness of criminal behavior. Once
criminal behavior is viewed as being less positive, meaningful, or salient, the process of
cognitive change may occur. In combination, these elements are assumed to produce
cognitive changes that can affect behavior and future involvement in crime.

Reentry Implications of a Social Development Emphasis
In terms of implications for reentry programs and programming, the social

development perspective provides a continuum of optimistic and pessimistic views on the

42

utility of intervention. For the former, the perspective suggests that interventions can
serve as supplements in the process of change. Efﬁcacious interventions should target
the individual and their cognitive processes. The focus must be placed directly on
identity change and only indirectly on behavioral change. Once cognitive transformation
occurs, one can formulate a new identity or self, abide by pro-social self-narratives, shed
old identities, and embrace a non-criminal lifestyle.

The task for reentry programs and planning is to develop methods of intervention
that will modify cognitive capabilities, foster conventional self-narratives, and provide
numerous role models in the immediate community that can be used as catalysts for
cognitive change. lrnportantly, these efforts must be balanced against the active degrees
of human agency and decision-making that individuals possess. Luckily, there is a
foundation for cognitive change within the ﬁeld of corrections. Numerous correctional
interventions utilize cognitive-behavioral curriculums or variations of cognitive-
behavioral approaches (Landenberger & Lipsey, 2006; Lipsey et al., 2001). A number of
reviews have also found cognitive-behavioral approaches to be efﬁcacious in reducing
levels of relapse and recidivism (Landenberger & Lipsey, 2006; Lipsey et al., 2001).

The optimism in the utility of intervention is countered by the focus on human
agency and its inherent relativity. In a pessimistic sense, the social development
perspective can lead to implications of the relative futility of interventions. The decision
to change is largely left to the individual independent of any external sources such as
programming, life circumstances, or social environment. This decision is a personal and
self-reﬂective question of existential being. Once again, the timing of change becomes a

focal concern. If the decision to change is not made and ingrained into one’s cognitive

43

thought prior to release and participation in subsequent interventions or programming, it
is unlikely that an individual will beneﬁt from such efforts. Alternatively, if an
individual has made a cognitive effort to change their own self narrative, it is likely that
the individual would desist or lessen their criminal behaviors by themselves even if
intervention or programming is offered.

The social development emphasis assumes that interventions will largely have
chance effects. Many of the issues contributing to these chance effects concern treatment
amenability and self-selection that are conditioned by the relativity of human agency. To
combat these issues, recent efforts have attempted to focus on readiness to change
identiﬁcation instruments, which have long been used for substance abuse treatment
(Miller & Rollnick, 1991; Miller & Tonigan, 1996). These instruments can be used to
identify individuals who may not be cognitively ready to make behavioral changes and
refer the individuals to motivational interventions to increase treatment amenability prior
to participation in treatment interventions (Checinski & Ghodse, 2004; Longshore et al.,
2004; Miller & Tonigan, 1996; Taxman, 2008). These efforts to induce motivation may
increase openness to treatment programming (Miller & Rollnick, 1991), but are not likely
to produce comparable levels of motivation for individuals who desire to make cognitive
changes without motivational interventions.

The social development emphasis challenges prevailing assumptions of equal and
available opportunities for behavioral change via programming and intervention with the
suggestion that change is much more variable due to the unpredictability of human
agency (Giordano et al., 2002). The importance of the individual is brought to the

forefront. Unfortunately, the emphasis on social development is not without limitation.

44

Giordano et a1. (2002) have suggested that there may be differential effects of behavioral
change inﬂuenced by cognitive transformations. Using qualitative interviews with
female offenders, the researchers found that female offenders appeared to be more likely

to change their behaviors via cognitive transformations, while a comparison group of

males were more likely to change their behaviors via formal control mechanisms”. This
ﬁnding also leads to questions of the assumptions of the social development perspective.
Individuals may not be as individualistic as the perspective seems to suggest if ﬁndings
can be interpreted by socio-demographic groupings.

Part of the problem and a cause for confusion is the infancy of the perspective.
There is relatively little empirical evidence to support the perspective, which is likely due
to the resources needed to longitudinally follow individuals and measure their personal
self-narratives. The perspective has spawned and provided support for the
methodological utility of life history or event history narratives (see Hepburn & Grifﬁn,
2004; Homey et al., 1995), but it is still unclear if this approach is empirical or
bibliographic (see Maruna et al., 2004a). If the narratives are more closely associated
with the latter, concerns of ad hoc rationalizations become problematic.
Supervision Emphasis

The supervision emphasis of reentry programs is based upon the complex
interrelationship between surveillance, deterrence, and accountability in community
corrections. Supervision effects cannot be understood without consideration of

surveillance mechanisms. Surveillance effects cannot be understood without

 

16 It is important to note that this ﬁnding in addition to the conceptualization of cognitive transformation
led to the development of the theory of cognitive transformation that is often used by those who focus on
the social development perspective.

45

consideration of deterrence mechanisms. In some instances the threat of punishment
associated with being under surveillance is enough to modify behavior. In other
instances, surveillance allows for the identiﬁcation of undesirable behavior for which
sanctions of punishments can be used to deter future manifestations of undesirable
behavior.

Taxman (2005) argued that any discussion of correctional programming in the
community must consider the generalized accountability model that underlies
correctional supervision. This model is related to conditional releases that hold
individuals to speciﬁc standards of behavior while under supervision. The supervisory
term functions as a source of formal social control that constrains everyday liberties,
structures daily activities, and holds an individual to a speciﬁc standard of conduct that is
assumed to shape future behavior (Taxman, 2008). Akin to the perspective of routine
activities theory (Cohen & FelSon, 1979), the perspective suggests that involvement in
crime can be shaped by conditions that change daily role and responsibility activities and
affect opportunities for crime. It is hoped that once daily activities are constrained and
structured, opportunities to partake in crime will be marginalized or greatly diminished.

The goal of behavioral accountability must also be balanced against the provision
of services and ensuring public protection (Corbett, 2008). Supervisory terms can expose
an individual to a number of agents of formal social control even if the terms are
orientated toward the provision of needed services. For instance, substance abuse
treatment conditions generally include drug testing conditions. Both of these services
attempt to constrain daily activity and available free time, but are also thought to build

pro-social skills and functioning through participation in treatment services. At the same

46

time, these supervisory term conditions can increase the risk for technical violation and
re-incarceration if compliance with treatment and testing conditions are not met. The
service providers responsible for treatment and testing can become sources of
surveillance that can inﬂuence an individual’s community supervision status (Foucault,
1977; Mobley, 2005; Simon, 1993).

Surveillance is a controversial issue within the ﬁeld of corrections and the
discipline of criminology and criminal justice as a whole. There are opposing viewpoints
on the utility of surveillance policies and processes. On one end, some have argued that
there are beneﬁcial connections between programming that utilizes surveillance
mechanisms in conjunction with the provision of social services (National Research
Council, 2008; Petersilia & Turner, 1991; Taxman, 2008; Useem & Piehl, 2008). For
instance, Corbett (2008) suggests that the ﬁeld of corrections as a whole must move
beyond the notion that individual change is contradictory to accountability through
surveillance. The logic of the claims supporting the connection between the two
correctional issues rests upon the presumption that the threat of surveillance can increase
compliance with programming because it increases accountability for actions. Moreover,
services can structure daily activities, provide meaning to daily activities, and act as a
new source of informal surveillance that can reduce reliance on formal surveillance.
Taxman (2008) argues that the blending of these features can make it possible to lower
recidivism rates and most importantly, reduce levels of technical violations.

Others are far more pessimistic in their discussion of the marriage between
surveillance and the provision of services. Increased attention to compliance of

supervision terms can have the unintended effect of constraining the effectiveness of

47

treatment or intervention services (Lipton et al., 1975; National Research Council, 2008;
Petersilia & Turner, 1991 ). Rosenfeld (2008) argued that rehabilitation through services
possesses distinct policy objectives and is qualitatively different from surveillance and/or
control through supervision. It is essential to keep services and supervision separate and
observe their individual effects. When services and supervision are intertwined diverse
outcomes may be observed.

In one of the most widely cited studies on the impact of supervision policies,
Petersilia and Turner (1991; Turner et al., 1992) suggested that intensive correctional
supervision efforts absent of treatment can lead to increases in technical violations. The
ﬁnding was unexpected and suggested that intensive supervision increases levels of
surveillance and re-incarceration (rather than decreasing levels of re-incarceration as
originally expected). Despite these overall ﬁndings, Petersilia and Turner (1991)
suggested there may be some degree of differential effects when considering the
partnership between enhanced supervision and treatment services. The researchers found
that enhanced supervision paired with treatment reduced levels of recidivism. However,
this partnership had relatively no effect on ameliorating increased levels of technical
violations associated with enhanced supervision.

The supervision orientations of community correctional agents may also affect the
transition process. Orientations adopted by agents may determine how surveillance is
used. Research suggests that philosophical orientations of service can affect the progress
of individuals under correctional supervision (Clear & Latessa, 1993; Glaser, 1969;

McCleary, 1978). Paparozzi and Gendreau (2005) found that parole agents who

48

identiﬁed themselves as punishment orientated were more likely to issue technical
violations relative to those agents who identiﬁed themselves as service orientated.

The relative value of deterrence mechanisms associated with correctional
supervision is also controversial and is equally split into opposing viewpoints.
Correctional interventions are largely grounded under the philosophies of deterrence and
incapacitation (MacKenzie, 2006). The threat of punishment is thought to induce
compliance with supervision terms and lead to normalized behavior through formal
agents of control. Punishment is most often assumed to be effective when viewed as
legitimate and possessing some degree of certainty, swiftness, and commensurability. If
these dimensions of punishment are not present, it is likely that the deterrent value of
punishment will be lost.

Many of the arguments for or against the value of deterrence are based upon the
calculation of risk. Deterrence overtly or inadvertently assumes that individuals will be
deterred by the rational calculation of risk where the punishment can outweigh the
perceived beneﬁts of criminality. However, the notion of pure rationality is largely
mythical. The literature suggests that individuals function under bounded or constrained
rationality that is inﬂuenced by one’s immediate socio-structural situation (Hechter &
Kanazawa, 1997). As such, some degree of relativity must be considered when
determining the merits of rationality.

Reentry Implications of a Supervision Emphasis

Reentry programs and programming that utilize a supervision emphasis shift the

focus away ﬁom the individual or the individual’s immediate social context. Instead, the

focus of the intervention is placed on the role of the criminal justice system in general

49

and the functionality of correctional agencies speciﬁcally. It is assumed that the
modiﬁcation of supervision terms and the intensity of supervision can induce individuals
into compliance, whether that compliance involves meeting supervision term orders or
meeting the requirements of service provision. It is important to note that the supervision
determinations are out of the control of the individual and are driven largely by
organizational decision-making.

Reentry programs and programming recognizes the importance of supervision,
since supervision and the generalized accountability model is a foundation of corrections.
The task for reentry programs and programming is to determine the balance that is
needed between agency supervision needs and liabilities versus the provision of services.
Reentry programs and programming that relies too heavily on deterrence based
punishment orientations are likely to have marginal or unintended consequences that can
hinder the process of transition and constrain attempts to adopt pro-social lifestyles.
Ideally, supervision policies and intensity in reentry programs and programming should
focus on service provision, partnered assistance, and advocacy with an understanding that
non-compliance with supervisory terms is a part of the transitional process, not a ﬁnal
event (National Research Council, 2008). The difﬁculty in meeting this ideal is the
organizational need of ensuring public safety for the broader community and maintaining
individual accountability.

Transitional Emphasis

The transitional emphasis of reentry programs and programming is the foundation

of nearly every available model of reentry. The emphasis is largely gleaned from

propositions of strain theory and the consistent ﬁnding of elevated risks of recidivism

50

during the ﬁrst 6 to 12 months after release from an institution. It is the transition from
structured and conﬁned daily life within an institution to unstructured life in the
community that involves a period of stressﬁrl adj ustrnent that can place individuals at risk
for recidivism. The stress of transition becomes salient, which can lead to reversions to
past criminal behavior.

The strain theory perspective implicitly centers on the dissonance between social
goals and the means to achieve such goals (Agnew, 1992; Bernard, 1984; Cloward, 1959;
Cloward & Ohlin, 1960; Cohen, 1955; Kornhauser, 1978; Merton, 1968). This
dissonance may be at the societal level, where the structures within society fail to provide
the legitimate means to reach conventional goals, or at the individual level, where
individualized feelings of stress (or variants of stress such as frustration, anxiety, worry,
depression, and anger) due to societal strain can be associated with criminal activity
(Cullen, 1988). In attempting to reach a predeﬁned societal goal, individuals can utilize a
number of methods based largely upon the available opportunities one possesses as well
as those available within their immediate social environment (Cloward & Ohlin, 1960).
Some of these methods will be consistent with dominant ideologies and norms, while
other methods will challenge the status quo and utilize non-conformist means to achieve
socio-cultural goals (Agnew, 1992; Cloward, 1959; Merton, 1968). The factor that can
guide an individual towards one method of achieving goals over another is the degree of
stress or strain an individual is subjected to in their attempt to reach societal goals.

Merton (1968) suggested that individuals commonly utilize ﬁve methods of
adaptation that can be used to minimize the amount of stress or strain experienced in

attempts to reach goals. Adaptations describe individualized decisions on how to react to

51

stress and strain (Merton, 1968; Vold et al., 2002). They do not describe typologies of
individualized personality. Individuals may also choose to partake in a single adaptation
or they may simultaneously utilize a number of adaptation strategies. The ﬁve methods
of adaptation include conformity, innovation, ritualism, retreatism, and rebellion.

Conforrnists accept the goals of society and use dominant ideologies and norms to
meet the goals of society. Innovators accept the goals of society, but develop their own
methods to meet the goals of society. Merton (1968) suggests that most crime that occurs
in society is due to innovation, where the end goal justiﬁes the means of attainment.
Ritualists accept the fact that the obtainment of societal goals is generally unattainable,
but continue to use dominant ideologies and norms as methods to strive for the
obtainment of goals. Retreatists chose not to participate in attempts to achieve goals nor
do they participate in the conventional ideologies and norms used by conventional
members of society. Merton (1968) identiﬁes individuals with severe substance abuse
problems as retreatists since their drug use is an adaptation to and an escape from
participation in society. Rebellionists reject the conventional goals of society altogether
and substitute their own goals and means of attaining those goals. By rejecting any and
all conventional goals, individuals who adapt to society by rebellion function as a
subculture with their own conventional goals, ideologies, and norms (Cloward & Ohlin,
l 960).

When faced with stressful situations or life circumstances there is an increased
likelihood of participation in criminal activities (Agnew, 1992; Cloward & Ohlin, 1960;
Merton, 1968), especially for those who have been previously processed through the

Criminal justice system and institutionalized (Irwin, 1970; Mobley, 2005; Travis et al.,

52

2001). The danger among those who will be transitioning into the community is the
reversion to past behavioral short cuts that may have contributed to entry in the criminal
justice system. The initial optimism of being transitioned into the community is quickly
met with the realities of everyday life (Nelson et al., 1999). Stress, strain, and unmet
expectations can come from multiple sources all at one time immediately following
release. The necessary adjustment and relative shock of community life upon transition
from an institution has been associated with the elevated rates of recidivism that occur
within the ﬁrst year after release (Zamble & Quinsey, 1997).

Gleaning from Maslow’s (1943) hierarchy of needs, physiological needs of food
and water are necessary and are followed by the safety needs of clothing, shelter,
employment, and generalized mental and physical well-being with or without the
assistance of social services. On top of the procurement of these needs, is the
management of social and interpersonal relationships. Most pressing are the relationships
with family members (Braman & Wood, 2003; Irwin, 1970; LaVigne et al., 2004; Naser
& Visher, 2006; Nelson et al., 1999; Parke & Clarke-Stewart, 2003; Petersilia, 2003;
Travis et al., 2001; Western et al., 2004; Zamble & Quinsey, 1997), peers (Irwin, 1970;
Taxman et al., 2002), treatment or service providers (Nelson et al., 1999; Pahner 1994;
Palmer, 1995; Petersilia, 2003), correctional supervision agents (Clear & Latessa, 1993;
McCleary, 1978; Glaser, 1969; Nelson et al., 1999; Palmer, 1994; Palmer, 1995;
Petersilia, 2003), employers (Brooks et al., 2006; Helfgott, 1994; Nelson et al., 1999;
Pager, 2007; Travis et al., 2001; Visher & Farrell, 2005), roommates or landlords (Brooks
et al., 2006; Travis et al., 2001; Visher & Farrell, 2005), community institutions (Brooks

et al., 2006; Nelson et al., 1999; Travis et al., 2001; Uggen, 2000; Visher & Farrell,

53

2005), and members of the community at large (Brooks et al., 2006; Nelson et al., 1999;
Travis et al., 2001; Visher & Farrell, 2005).
Reentry Implications of a Transitional Emphasis

Reentry programs and programming have recognized the importance of lessening
the compounded stressors an individual faces upon release into the community. There is
an attempt to minimize the stigma associated with incarceration and reconnect an
individual with their larger social institutional relationships in the community.
Interventions that utilize the transitional emphasis tend to focus on comprehensive pre-
release planning and the continual maintenance and adjustment of reentry plans in an
effort to minimize the strains of what will be faced during the transition process. Service
programs and programming are subcomponents of the transitional emphasis and are
overshadowed by efforts to plan for the future and adjust plans in accordance with
progression in meeting the pre-deterrnined goals.

The transitional emphasis recognizes that all of the other theoretical emphases do
not occur in a vacuum. All of the emphases overlap and interact with one another
immediately following release and across prolonged periods of adjustment to community
life. There is a presumed variability with how each emphasis will affect individuals.
Planning efforts must consider the strengths and weaknesses each individual has across
the theoretical emphases of social control, social development, and supervision.

Key to the process of planning is ensuring the continuity of services if and when
services are needed. It is expected that plarming efforts coupled with service continuity
will increase levels of stability by structuring daily activities and also hold individuals

accountable to meeting the goals outlined in reentry plans. Reentry programs and

54

programming have begun to restructure the process of transition by placing more
emphasis on planning for release upon admission to an institution, rather than providing
preparatory sessions on what to expect during release weeks prior to release.

Risk and need assessment instruments are used to develop individualized plans.
These plans are intended to be comprehensive and can cover deﬁcits in housing options,
employability background, familial relationships, peer relationships, and the possession
of necessary identiﬁcation documentation (e. g., birth certiﬁcates, social security cards,
driver’s licenses, or state identiﬁcation cards). Moreover, these plans attempt to identify
those who will need intensive treatment services within the institution and, depending on
progress within the institution, those who will need similar intensive treatment services in
the community upon release.

Individualized planning once again places the focus back on the individual and
their immediate social environment. Since each individual has a unique background with
deﬁcits and assets to build from, comprehensive planning should aid the transition
process. There are some correctional sub-populations — such as those with severe
substance abuse histories and dependencies — that have an assortment of deﬁcits that can
hinder the transitional process due to the potential for stress induced relapse, which can
contribute to recidivism (National Research Council, 2008). One size ﬁts all planning for
this population is likely to cause more harm than good. As such, there is an expectation
that differential effects on relevant correctional indicators such as recidivism as well as
reentry-based indicators such as housing and employment should be expected since

reentry plans can vary on a case by case basis.

55

However, the individuality of planning is often relative. In an effort to
standardize across sub-populations and increase efﬁciency in forrnalizing reentry plans,

many state correctional systems are developing new methods or modifying old methods

of actuarial reentry assessment”. The use of and reliance upon actuarial instrumentation
is not without controversy (Feeley & Simon, 1992; Mears et al., 2008; Petersilia, 2003;
Simon, 1993). The assessments are designed to minimize the potential for differential
effects between individuals who are identiﬁed to receive speciﬁc forms of intervention.
As such, the assessments can be of assistance to offenders who need intervention. At the
same time, however, the assessments have the unintended ability to prevent those who
may be most in need of intervention from gaining eligibility for participation by design,
external circumstances, and conditional overrides by service providers (Feeley & Simon,
1992; Useem & Piehl, 2008).

One of the largest hurdles for reentry programs and programming that attempt to
emphasize the process of transition is following through with the implementation of
comprehensive transitional services. The current discussions on the provision of
transitional services are more ideal than practical. Very few, if any, state correctional
systems have established reentry based services despite the wide acceptance of the

available reentry models by correctional administrators and policy makers (National

 

17 For instance, the Correctional Offender Management Proﬁling for Alternative Sanctions (COMPAS) has
been increasingly used for assessments of reentry needs and risks, although the assessment was not
originally designed speciﬁcally for reentry determinations. COMPAS has been used to manage risk, offer
suggestions for intervention targets and types, and provide assessment and feedback on progress in
behavioral change (Brennan & Oliver, 2000). The assessment has been validated (Brennan et al., 2009;
Brennan & Oliver, 2000; Pass et al., 2008), but there is some question as to whether the instrument is
applicable to racial and ethnic sub-populations (F ass et al., 2008). The Stages of Change Readiness and
Treatment Eagemess Scale (SOCRATES) has been validated with participants in substance abuse treatment
(Miller & Tonigan, 1996) and is being extended for use with offenders nearing release in some states. The
scale has yet to be validated speciﬁcally for reentry purposes. It is likely that these instruments, and similar

56

Research Council, 2008). Many have suggested that the degree of organizational change
— both in terms of structure and prevailing philosophies — required to truly shift focal
concerns towards reentry will take a number of years and may not even be feasible
(Clear, 2007; Lynch, 2006).

Adding to these limitations are evaluative and empirical concerns. The
transitional emphasis of reentry programs and planning attempt to be as comprehensive
as possible and include a number of dimensions directly or indirectly related to the
process of transition. The comprehensiveness of the transitional emphasis makes it
difficult to parcel out and determine speciﬁc effects that inﬂuence relevant programmatic
outcomes and inform future reentry programs and programming. These characteristics
have also contributed to the overall lack of empirical evidence on the efﬁcacy of reentry.
An emphasis on transition suffers from black box deﬁcits in understanding since the

process between the identiﬁcation of program participants and the production of relevant

program outcomes cannot be made with much conﬁdence1 8. One can often claim that
something happened and speculate to a number of potential explanations.
Theoretical Insights across Emphases

Each theoretical emphasis suggests points of programmatic intervention that are
hypothesized to reduce or eliminate criminal behavior — the quantity and quality of social

bonds, cognitive amenability and transformation, supervision experience and quality, and

 

applied instruments, will be used as additional sources of information that can aid in the reentry planning
process rather than being stand-alone reentry assessments.

8 The “black box” issue of program evaluation is especially problematic within the ﬁeld of criminal
justice (see McGarrell et al., 1999; Weisburd, 2000). Many evaluative efforts are made for large-scale
interventions involving numerous components and dimensions, which makes it difﬁcult to determine the
speciﬁc program processes that contributed to program outcomes (Lipsey & Cordray, 2000; McGarrell et
al., 1999; Rossi et al., 2004).

57

the minimization of transitory strain. In turn, a focus on these various intervention points
is assumed to ease the process of transition and reentry programs that incorporate these
intervention points may produce beneﬁcial reentry effects for participants. The key
question that has yet to be answered (and the key assumption that has yet to be tested) is
whether programmatic responses can reach and affect the points of intervention
suggested by the theoretical emphases.

The positivist assumptions that underlie social control, social development, and
transitional theoretical emphases support the use of interventions to produce behavioral
change (Byme & Taxman, 2005). Each of the emphases places a focus on individual
amenability and provides direction to areas of program intervention (Harris, 2005). The
supervision emphasis also rests largely upon positivist assumptions, but it is not in line
with the use of traditional intervention methods (via programs and programming) to
modify behaviors. Instead, the focus is placed upon the efﬁciency and effectiveness of
organizational supervision policy and practices.

The social development emphasis also conﬂicts with positivist assumptions.
Maruna (2001) has argued that the social development approach should be viewed as
being a critical response to the “what works” perspective whose focus concerns the
design and use of programming. In fact, the social development emphasis has been built
upon observations that individuals can change and terminate their trajectories of criminal
behaviors without interventions or programming.

It may be too simplistic to focus on the merits of one theoretical emphasis over
another. Prisoner reentry is complex and is likely to be affected by an assortment of

theoretical combinations. Effort must be placed on determining the value of integrating

58

the theoretical emphases since there is a necessary overlap and interaction between all of
the emphases. Currently, the transitional emphasis appears to be the preferred reentry
program model structure since the emphasis encompasses all of the other theoretical
perspectives. Due to its model preference in the ﬁeld, the most promising avenue for the
development of theoretical knowledge on reentry and the application of such knowledge
should begin with the transitional emphasis. The integrative framework inherent to the
emphasis suggests that programmatic interventions should minimize levels of stress or
strain for participants, expose participants to formal and informal sources of social bonds,
and foster participant self-development through cognitive change. All of these points of
intervention would be subsumed under the auspices of supervision.

The presumed integration and interaction of theoretical approaches across
emphases is made regardless of contradictory focuses and underlying assumptions.
Social control and social development approaches are compatible with one another; the
development perspective is an elaboration of the control perspective (Bottoms et al.,
2004; Burnett & Maruna, 2004; F arrall & Bowling, 1999; Giordano et al., 2002; Maruna,
2001; McNeill, 2006). The supervision emphasis is based upon efforts to control through
surveillance and deterrence mechanisms by agents of formal social control. As such,
arguments can be made that the supervision emphasis is also congruent with social
control and social development emphases with a focus on control mechanisms.

However, a complete picture of how social control mechanisms shape identity
formulation and how identity management affects social bonding during the transition
process has yet to be established. Additionally, it is still not clear how the strain focus of

the transitional emphasis can be integrated into the remaining emphases or if the focus

59

can be integrated at all. It may be possible that the strain emphasis ﬁrnctions as a
mediator or moderator that shapes social control, social development, and supervision
emphases and exacerbates or minimizes the difﬁculties faced upon the transition into the
community.

There is promise to the existing theoretical emphases and the potential integration
of the emphases. According to Petersilia (2004), reentry should work in theory. There is
little doubt that nearly all of the theoretical emphases suggested by Lynch (2006) have
some degree of empirical support in affecting criminogenic behavior. What is not clear is
whether these theoretical perspectives can be directly applied to reentry programs. There
is simply too little empirical evidence to suggest, at this point in time, that any one of the
theoretical perspectives associated with reentry would be any more effective than any
other effort. This observation is damaging since reentry programs will continue to be
developed and implemented without a ﬁrm theoretical foundation.

From Theory to Practice: Research on Prisoner Reentry

What is empirically known about the process of reentry is marginal at best.
Maruna (2001) argued that reentry is the least understood phenomenon in corrections,
adding that the “enormous difﬁculties faced by ex-convicts after release have been
consistently and extensively documented for the past 100 years, [but] the mechanisms for
ex-offender reintegration have not improved greatly in that time” (p. 70). In some
regards, this should be expected. It is difﬁcult to discuss the average transition when
there is such a wide degree of heterogeneity to the reentry experience (National Research
Council, 2008). However, with millions of taxpayer funds being invested into efforts to

assist in the transition into the community, the continued progression of uninformed

60

programs may not be the best use of available funds and can produce more harm than
good.

The most direct avenues for the development of theory and research on reentry
have been qualitative in nature. For instance, Nelson, Deess, and Allen (1999) followed a
small group of released prisoners in an effort to observe ﬁrst month post-incarceration
experiences and identify some of the dimensions that can affect the process of transition.
The researchers concluded that the most important determinate of reentry success was
family and community support. Nearly all of the released prisoners chose to reside with
their families and were perceived to be welcomed to stay there. Simple bivariate
analyses suggested that those who resided at shelters were far more likely to abscond
during their ﬁrst month on parole. Employment was difﬁcult to obtain, but many found
employment opportunities through family, ﬁiends, or the reconnection with old
employers. By far the most important challenge faced during the ﬁrst month was the
obtainment of identiﬁcation and insurance (needed for substance abuse treatment). One
of the more interesting ﬁndings was the relative shock the parolees faced during their ﬁrst
few days out. The released prisoners expected their parole ofﬁcers to assist them in
ﬁnding employment or with employment referrals, however were disappointed to ﬁnd out
that parole ofﬁcers did not provide such services.

The Urban Institute has also provided a great deal of information regarding
reentry. Research conducted by the Urban Institute generally falls into three frameworks.
First are reentry focus groups. Brooks et al. (2006) arranged a focus group of residents
from communities with high concentrations of returning prisoners to assess their

perceptions of reentry. Residents acknowledged that returning prisoners had a number of

61

social deﬁcits that require supportive families and competent parole ofﬁcers. The
residents also suggested that prisoners were ill prepared for reintegration into the
community. A number of services were available in the community, but the process of
providing information to begin participation in services was not provided to the prisoners.
Some services, such as job training and placement, are not offered. The residents also
seemed to realize that their community was undergoing substantial change independent
of the movement of prisoners into the community. Economic downturn, the loss of
community values and responsibility, drugs, and violence were identiﬁed as some of the
most recent changes to the local community. Elected officials were perceived as being
unconcerned with prisoner reentry. In all, this type of information contributes to a social-
psychological understanding of reentry by placing the focus on the reactions of the
broader social environment to transitional issues.

Second and third lines of research conducted by the Urban Institute were city and
state proﬁles of returning prisoner populations (LaVigne et al., 2003; Solomon et al.,
2004b). The research generally discussed demographic and criminal history information
for the entire population of returning prisoners and the concentrations of areas to which
they return. Census data generally accompanied the information to highlight the
structural disadvantage that exists in the areas to which prisoners return (LaVigne et al.,
2003; Solomon et al., 2004b). The city proﬁles also included additional information that
is often not found in the state proﬁles. Visher and Farrell (2005) surveyed returning
prisoners, residents from communities with high concentrations of returning prisoners,
and reentry administrators and practitioners to develop insights on the reentry process in

one major city. All of the respondents emphasized the important role of the local

62

community in their discussion of reentry. Returning prisoners suggested that some
neighborhoods had more opportunities (e. g., housing, employment, social services) than
others. Overall, the returning prisoners were content in their neighborhood, but were
having a difﬁcult time ﬁnding employment. Additionally, their neighborhood of
residence after prison was often not the same neighborhood prior to prison admission.
Residents and reentry stakeholders noted that the local neighborhood provides important
social control functions. Families were again seen as being the key to a successful
transition. Additionally, there was a continued emphasis on the fact that communities are
unprepared for the population of returning prisoners and lack essential transitional
services.

Knowledge on reentry dimensions serves an important purpose by identifying the
challenges to reentry that can affect the process of transition. Unfortunately, the focus on
identifying dimensions has yet to be translated into practical application. It is still not
clear how the miscellaneous reentry dimensions affect reentry programs or programming
and how the interactions between the two can affect the overall transition experience.
Equally important, it is still not clear how speciﬁc reentry dimensions affect other
relevant reentry dimensions to shape transitional experiences. Many of these issues have
not been resolved (or even attempted). The available evidence that could inform these
issues — largely from reentry program or programming evaluation — is virtually non-
existent.

There is some evidence to suggest that research is beginning to move toward
considerations of the interrelationships between reentry dimensions and programming.

The Urban Institute has partnered with the Research Triangle Institute (RTI) Intemational

63

to conduct a multi-site evaluation of the Serious and Violent Offender Reentry Initiative
(SVORI). SVORI established a pool of federal funds (approximately $110 to $150
million) to be dispersed to state and local correctional agencies to assist with the
development and/or expansion of reentry programs in the early 20005 (Lattimore et al.,
2004; Lattimore et al., 2005). Each state received at least one award disbursement, with
awards ranging in value from $500,000 to $2 million across three years (Lattimore et al.,
2004). States and localities were largely left to their own accord in how the funding
should be used, but there were some requirements associated with the receipt of ﬁrnding.
All of the grantees were required to form partnerships between criminal justice and
community agencies, establish continuity of services within the institution that extends to
the local community, and specify service deﬁcits the initiative may ameliorate (Lattimore
et al., 2004). Additionally, grantees were encouraged to focus upon their most serious
populations under the age of 35 and were able to use discretion to select those most at
risk (Lattimore et al., 2005; Lattimore & Visher, 2009).

Early process evaluation ﬁndings impart the heterogeneity of participants and
programming involved in SVORI and reentry in general. A majority of the established
programs did not include exclusionary criterion for participation by offense type or
service need and most serviced less than 100 participants (Lattimore et al., 2004;
Lattimore et al., 2005). Additional substantive problems with eliciting participants were
the establishment of a variety of voluntary reentry programs and operational deﬁcits
related to the accuracy of release information and dates (Lattimore et al., 2005). Nearly
all of the established programs included multiple components including, but not limited

to, assessment, plan development, housing assistance, employment or educational

64

training assistance, substance abuse treatment, mental health counseling, medical or
dental services, and faith-based services (Lattimore et al., 2004). Most of the established
programs also utilized some form of case management and service coordination to
enhance continuity of care (Lattimore et al., 2005). The established programs generally
included similar components, but the components themselves varied from program to
program (Lattimore et al., 2005).

In terms of process evaluation results on service delivery, Winterﬁeld and
colleagues (2006) suggested that few SVORI participants received a full dose of services
within their reentry program. Moreover, comparison groups of individuals were just as
likely to receive similar services through institutional and community supervisory
services as SVORI participants. However, the preliminary ﬁndings on service delivery
suggested that SVORI participants were more likely to receive more types of services in
an overall package of services than a comparable group of individuals. On average,
SVORI participants received substantially more coordination services and employment,
education, and skill development services. These differences persisted when the analyses
were constrained to comparisons during institutional and community-based service
deliverables. Considering the ﬁndings regarding dosage, there appeared to be differential
effects between SVORI and non-SVORI participants that may be largely due to increased
exposure to speciﬁc types of services.

Unfortunately, at this time, the outcome evaluation of SVORI has yet to be
completed (Lattimore & Visher, 2009). Lattimore and Visher (2009) have provided
preliminary results that suggested positive, but small, differences in employment

outcomes for SVORI participants across 3, 9, and 15 month follow-up intervals.

65

Participants were slightly more likely to support themselves with a job that was
permanent, and included formalized pay periods with beneﬁts. It is noteworthy that the
most dramatic positive differences in employment considerations for participants were
for the obtainment of a job with beneﬁts across all follow-up periods and the supporting
of oneself with a job 15 months post-release. By implication, these preliminary ﬁndings
suggest important employment differences that have the potential to lead to gains in
health and social capital over time.

What is still lacking from the outcome evaluation of SVORI is information on
program processes. Speciﬁcally, there is still a need to determine how program processes
inﬂuence and are inﬂuenced by dimensions of reentry, such as the gain or loss of housing
and/or employment. Once these determinations are made, there is still a need to
determine how the interrelationship of program processes and reentry dimensions
contribute to program outcomes. Simple exposure to a number of services may produce
beneﬁcial outcomes for participants relative to non-participants, but this narrowed
viewpoint fails to provide insights about speciﬁc services that may be more efﬁcacious
than others in easing the transition process and reducing levels of recidivism, which can
be used to inform future reentry programs or programming. Unfortunately, nearly all of
the available research on reentry programs and programming suffers from similar deﬁcits
in acknowledging program processes.

Wilson and Davis (2006) discussed a randomized evaluation of a reentry program
in New York. The program followed the structure of many posited reentry models and
has received the title of being the only empirically assessed contemporary prisoner

reentry program to date according to the National Research Council (2008). The

66

comprehensive program included a phased transition process that began upon entry to an
institution and continued through release and post-release supervision. Initial
assessments were made, programming on responsibility, reconnection with family and
friends, and life skills training were offered, and attempts to establish connections with
family, ﬁiends, employers, community organizations, and parole staff were conducted
eight weeks prior to release. Programming was largely cognitive behavioral,
emphasizing the modiﬁcation of thinking and behavior, and was multimodal. Additional
components of the program provided job training and placement for interviews, made
efforts to keep individuals away from shelters, provided relapse prevention programming
provided for those with substance abuse problems, and offered family counseling when
necessary. Life skills training included lessons on practical considerations such as public
transportation, budgeting, banking, and time management.

Accredited clinicians provided most of the programming services. A community
coordinator was utilized to network individuals with community service providers. A
case manager was also assigned to each individual. The case manager had the
responsibility to schedule release plans for individuals to follow, documented progress
and adjusted plans as needed. All of the information gained from program staff was
shared with an individual’s supervision agent.

One year post-release outcomes indicated that program participants had more
referrals and service contacts than comparison groups. However, program participants
were more likely to be arrested and more likely to be arrested for more serious crimes
relative to comparison groups. There were some indications of individual and program

effects. Younger individuals with lengthier criminal histories convicted of drug or

67

property instant offenses had increased probabilities of re-arrest. Certain case managers
who provided programming services were also associated with increased risk of failure.

In explanation of the unexpected ﬁndings, the researchers suggested difﬁculties in
the haphazard and subsequent random assignment process, issues of program design, and
problems in program implementation as factors leading to negative outcomes for program
participants. Marlowe’s (2006) assessment of the ﬁndings suggested that the program
relied on cognitive-behavioral, family reuniﬁcation, and employment service
programming that was found to be ineffective or lacked evidence of effectiveness across
a number of independent studies and meta-analyses. In turn, the transitory “kitchen sink”
approach most likely interacted with a population of offenders who have already been
exposed to a number of superﬁcial programs, which then produced outcomes more
negative than those to be expected with no programming. Rhine et al’s (2006)
assessment of the ﬁndings placed emphasis on the failure to implement the program
model and the inability to maintain program integrity.

Nearly all of the critiques of the work and ﬁndings of Wilson and Davis (2006)
emphasized the realization that there is little theoretical or practical knowledge available
from which reentry programs or pro grannning can build upon. The current state of
knowledge appears to be more of selective choosing from prior correctional discourses
that may or may not be consistent with the prevailing deﬁnitions of reentry or the reentry
process (including but not limited to such discourses as rehabilitative ideal/amendable to
intervention, nothing works, intensive supervision, what works, and
reintegrative/desistance). Even with the marginal and largely negative outcomes, there

are lessons to be learned concerning program processes that can be used to inform ﬁrture

68

programs and programming (Rhine et al., 2006). Of critical importance is the
identiﬁcation of key obstacles that may have contributed to negative program outcomes.
While yet to be examined in the literature, it is likely that salient sub-dimensions of
reentry such as housing, employment, and substance abuse treatment can directly and
indirectly affect programming, which also shapes program outcomes.

It is also worthwhile to note that it is equally important to develop and utilize
stringent research designs and evaluation methods for research on reentry programs or
programming. Reliable and valid outcomes are needed in the current era of reentry
research that is still in its infancy (U seem & Piehl, 2008). Very few correctional
programs or programming have been subjected to formal evaluation (Lattimore & Visher,
2009; Petersilia, 2003; Petersilia, 2004; Wilson & Davis, 2006). Even fewer of those
programs or pro gramming that have been subjected to formal evaluation have been
designed using quasi-experimental or experimental designs (Lattimore & Visher, 2009;
Petersilia, 2003; Petersilia, 2004; Wilson & Davis, 2006). This is an unfortunate
situation that continues to add confusion to the design and implementation of reentry
efforts as well as the interpretation of program processes and outcomes.

Even though knowledge on reentry programs and programming is still in its
infancy, there are valuable implications that can be made to inform future reentry
programs and research. First, the reentry experience is heterogeneous. An average
pathway of transition does not exist across individuals or within offender types and one
should expect differential effects when considering the reentry process. Second, there
may be beneﬁcial program effects with regard to speciﬁc reentry dimensions despite

observations of null or negative overall program outcomes on recidivism. The limited

69

evaluations of reentry programs seem to suggest that participants make favorable gains in
the reentry dimensions of housing, employment, and treatment exposure or participation
relative to individuals experiencing traditional transitional services (Lattimore et al.,
2004; Lattimore et al., 2005; Lattimore & Visher, 2009; National Research Council,
2008; Wilson & Davis, 2006; Winterﬁeld et al., 2006). There is a need to determine why
positive gains in reentry dimensions somehow lead to marginal program outcomes.
Speciﬁcally, there is a need to explore how speciﬁc reentry dimensions can interact,
shape program process, and contribute to overall program outcomes. In order to further
this research agenda, it is necessary to explore some of the most salient dimensions of
reentry — housing, employment, and treatment (speciﬁcally substance abuse treatment).
The Salient Sub-Dimensions of Reentry: Housing and Employment

Two of the most important dimensions found in the literature and one of the most
pressing needs immediately following release into the community are those of housing
and employment (Brooks et al., 2006; Helfgott, 1997; Lattimore et al., 2004; Mallik-
Kane & Visher, 2008; National Research Council, 2008; Petersilia, 2003; Rosenfeld,
2008; Taxman et al., 2002; Visher & Farrell, 2005; Visher & Travis, 2003). Housing and
employment are generalized needs, often applicable to all offender types. Both of these
reentry dimensions can provide structure to and enhance stability of everyday life, while
the inability to obtain either of the dimensions places one at risk for continued criminal
behavior (Andrews & Bonta, 2003; Petersilia, 2003; Visher & Travis, 2003). Moreover,
both of these dimensions necessitate a reciprocal and interactive relationship with one
another. The obtainment of housing can stabilize and direct employment options, while

the obtainment of a job can stabilize and direct housing options. Once these two

70

dimensions are in place, the obtaimnent and development of additional reentry
dimensions (e.g., personal documentation, medical coverage, food stamps and other
assorted beneﬁts, enrollment and admission to treatment programs, participation in
community based services) can occur.

Considerations of housing and employment introduce a degree of humanism into
correctional policy and practice. As Rosenfeld (2008) acknowledged, housing and
employment issues surrounding reentry programs and policies should not be considered
enhancements of public safety or control efforts. Rather, these issues should be
considered because individuals experiencing the transition from prison to the community
lack housing and employment. Efforts to enhance housing and employment options or
opportunities should be made to produce gains in social capital (Clear, 2007; McGarrell
et al., 2005). These gains can be re-invested into the local community and improve the
overall worth of a geographic area and society as a whole.

The Sub-Dimension of Housing

The obtainment of suitable housing is a key concern that can assist those
transitioning into the community from staying out of prison (Brooks et al., 2006;
Helfgott, 1997; Metraux & Culhane, 2004; Petersilia, 2003; Visher & Farrell, 2005;
Travis et al., 2001; Visher & Travis, 2003). The relevance of housing is particularly
acute since it is the very ﬁrst issue faced prior to and immediately following entry into
the community. In general, conditions of community-based correctional supervision
require pre-deterrnination assessment and approval of residence placement prior to
release into the community. The approval of placement is a relatively simplistic process,

with most placements being accepted as long as generalized conditions are met (e.g.,

71

homeowner or tenant approves of residence, residence is drug free, weapons are not
easily accessible, individuals with felony convictions do not reside within the residence,
etc.). The leniency that surrounds placement approvals is largely due to the notion that
an identiﬁed placement is preferred over placements to homeless shelters or the streets.

All of those returning to local communities have been detached, for some period
of time, from their local neighborhoods, which makes it that much more difﬁcult to
establish connections in the housing market (Roman, 2004). Many of those transitioning
into the community lack the monetary and social resources necessary to compete in the
housing market and obtain immediate housing (Metraux & Culhane, 2004; Roman, 2004;
Wilson & Davis, 2006). The stigma of having a criminal record exacerbates housing
difﬁculties (Pager, 2003). There are a number of laws that prevent individuals with
felony records from residing in public housing or assisted/subsidized housing units as
well as with family members or peers with criminal records (Helfgott, 1997; Lattimore et
al., 2004; National Research Council, 2008; Pager, 2003; Roman & Travis, 2004;
Solomon et al., 2004a).

Transitional housing services and programs are not widely available (Latessa,
2004; Roman, 2004; Roman & Travis, 2004). Many of those transitioning to the
community expect such services and become frustrated when they are not available or
when the knowledge of such services is not provided by correctional supervisory staff
(Nelson et al., 1999; Roman & Travis, 2004). If available, the housing services and
programs are not always safe and are commonly located in crime ridden areas devoid of
social capital (Metraux & Culhane, 2004; Wilson & Davis, 2006). Even if relatively

affordable housing can be obtained, there is evidence to suggest that the established

72

residence will likely be located in areas similarly situated with high levels of crime and
marginalized social structures (Brooks et al., 2006; Clear et al., 2005; Visher & Farrell,
2005)

There is a relatively common “not in my backyard” attitude held by community
residents towards the potential or actual residence of offenders in their neighborhoods
(V isher & Farrell, 2005). It is likely that returning offenders will be relatively
concentrated in a speciﬁc area and will return to neighborhoods that are structurally
similar to areas resided in prior to incarceration (Clear et al., 2005; Visher & Farrell,
2005). In combination, these issues systematically push and keep offenders in low-rent
and socially disorganized neighborhoods on the periphery (National Research Council,
2008). These areas commonly suffer from prominent and entrenched drug markets,
which increase risks for recidivism or re-incarceration (V isher & Farrell, 2005). The low
capacity for social control, collective efﬁcacy, and resource potential in these areas
drastically limits the availability of services and marginalizes opportunities for upward
mobility and potential gains in social capital (Wilson, 1996; Wilson, 1990).

The limited housing options generally lead to a reliance on family members or
peers to assist with living accommodations. Very few individuals live by themselves
(Jacobs & Western, 2007). Most live with relatives, parents, spouses, or partners (Jacobs
& Western, 2007; LaVigne & Kachnowski, 2003; Nelson et al., 1999; Roman & Travis,
2004). These members are often the ﬁrst to provide assistance and serve as a reliable
source of support for initial monetary and housing needs (Brooks et al., 2006; Nelson et
al., 1999; Roman & Travis, 2004). There is a wide degree of variability in the stability of

housing with family members. For some, familial housing will be long term and enhance

73

stability while under correctional supervision, while for others familial housing will be
short term due to past and current stressful family circumstances or relational strains
(Mallik-Kane & Visher, 2008; Roman & Travis, 2004; Solomon et al., 2001a). The same
thought holds true for attempts at procuring housing with peers. At the ends of the
continuum are options for housing stability or instability, which can affect future criminal
behavior as familial and peer networks and bonds ﬂourish or breakdown (Mallik-Kane &
Visher, 2008; National Research Council, 2008; Roman & Travis, 2004; Solomon et al.,
2004b)

Social supports may be available to some, but at the extreme are instances where
an individual needs to be placed in a homeless shelter or the street. These options are not
always conducive to the establishment of a crime free lifestyle, but are often used when
release placements cannot be speciﬁed (Metraux & Culhane, 2004; Roman & Travis,
2004; Visher & Farrell, 2005). The exact proportion of individuals released to local
communities and residing in homeless shelters or the streets is not known. Langan and
Levin (2002) suggest that approximately 12% of soon to be released prisoners reported
being homeless prior to incarceration and Roman and Travis (2004) have suggested that
approximately 10% of releasees will be homeless (and an equal proportion will enter an'
institution after being homeless for a period of time). Using samples of homeless
individuals, Schlay and Rossi (1992) suggest that the proportion of homeless individuals
who have served prison time is somewhere between 4% to 49%. More recent evidence
suggests that the proportion is near 23% (Metraux & Culhane, 2006).

The risk for homeless shelter use or life on the street is most apparent

immediately following release (even with efforts to gain approved placements) and

74

quickly erodes over the ﬁrst few months post-release (Metraux & Culhane, 2004). There
is evidence to suggest that placements at shelters or on the streets lead to an increased
risk of recidivism. Metraux and Culhane (2004) tracked a cohort of nearly 50,000
individuals released from prison and found that approximately 11% of the releasees
resided in a homeless shelter at some point in time within two years post-release. Of
those who were admitted to a homeless shelter, approximately 33% were returned to
prison within two years post-release.

With or without social supports, offenders transitioning into the community are
likely to face unstable housing opportunities, which lead to multiple housing movements
and a reliance on temporary accommodations (Metraux & Culhane, 2006; Mallik-Kane &
Visher, 2008; Visher & Farrell, 2005). The average number of residential moves made
by ex-offenders has yet to be deﬁned in general or among sub-populations (e. g., by
offender type, by geographic area, etc.), but there seems to be a consensus that the
frequency of moves made by ex-offenders is substantially higher than individuals without
criminal histories (V isher & Farrell, 2005). The effect of continued mobility for ex-
offenders is not clear. Frequent residential movement can disrupt the broader social
community by reducing the social capital capacity and increasing the levels of social
disorganization in an area (Bursik & Grasmick, 1993; Clear, 2007; Shaw & McKay,
1969). At the individual level, it is reasonable to assume that the frequency of
movements made can negatively affect participation in treatment or community services
and increase the risk for recidivism. Unfortunately, there is a dearth of available
empirical information to support such a claim. Additional research is needed to

determine the extent to which residential stability affects the process of reentry.

75

Housing considerations and placements have been and continue to be a constant
focus of the pre-release process. As such, the issue of housing is one of the few
forerunner concerns that existed prior to the developing reentry movement. Reentry
programs and programming commonly consider housing options, but it is not clear if the
considerations are anything beyond the traditional focus on approving residential
placements during the pre-release process. Some have suggested that the amount and
quality of pre-release planning made is more in line with the latter thought (Metraux &
Culhane, 2006; Roman, 2004).

Despite being relatively ingrained into correctional policy and practice, there is
little knowledge of how housing issues can shape reentry experiences. Much of the
available information on housing is ad hoc and consists of descriptive information on
location of residence, type of residence, and information on the other residents an
individual resides with. It is likely that the dearth of information on the issue of housing
is simply overlooked due to the fact that housing should be a need that is met with or
without correctional intervention. Moreover, there are a number of practical limitations
that constrain research concerning housing issues. It is difﬁcult (if not impossible) to
systematically evaluate the differential effects of housing options through a rigorous
research design due to issues of assignment to speciﬁc housing conditions (see Metraux
& Culhane, 2004). Even if such a research proposal could be established, it would be far
too costly to conduct. It is also likely to be difﬁcult to gain information on housing issues
since such research generally involves family members or peers who are external to the

criminal justice system and who often do not want to be included in any efforts related to

76

or emphasizing issues directly or indirectly related to the system (see Solomon et al.,
2004a)

Latessa (2004) suggested that there is a lack of transitional housing and effective
programs or programming that include transitional housing components available to
manage ex-offenders returning to the community with a plethora of overlapping needs.

One of the few areas in which research has been conducted on issues of housing has been

on halfway houses”. Research on halfway houses has indicated that the scope of
halfway house programming is very narrow and therefore only available to a handful of
individuals (Roman, 2004). There is some evidence to suggest that halfway house
placement of speciﬁc offender types can lead to reductions in recidivism (Lowenkamp &
Latessa, 2002; Lowenkamp & Latessa, 2005). However, it is unclear if the step-wise
progression from an institution, to a temporary halfway house stay, to an approved
residential placement contributes to the reduction in recidivism or if it is the combination
of individual effects, halfway house programming effects, and the progression to
residential placement that contributes to the reduction. Most of the available research on
halfway houses relies on poor evaluative designs leading to questionable validity of
ﬁndings (Roman & Travis, 2004). The “black box” surrounding halfway house processes
has not been adequately addressed in the available research and it is not clear if the

positive ﬁndings could translate to reentry programs.

 

19 Halfway houses have been categorized with community based residential centers and conceptualized as
partial physical custody centers by Lipton et al. (1975). This type of intervention generally includes
substance abuse treatment services, life skills training, and referrals to community based services. As such,
partial physical custody centers attempt to intervene across speciﬁc needs (often contingent upon eligibility
and admission criteria). These centers characteristically represent an additional period of conﬁnement,
even when the location and function of the center is not associated with correctional authorities.
Participants must abide by center rules in addition to community supervision mles or risk the possibility of
returning to an institution.

77

Even with the provision of housing services under a reentry program framework,
it is probable that reductions in levels of recidivism will not come to fruition (McGarrell
et al., 2005). It is more likely that gains in housing will lead to gains in social capital,
which may foster reductions in levels of criminality and desistance (McGarrell et al.,
2005). With the grth of reentry programs and programming, it is likely that housing
services and placement will continue to be of vital importance for beginning the process
of reentry. Additional research is needed to explore how the gain or loss of housing
affects recidivism, how the gain or loss of housing affects participation in reentry
programs or programming, and how the gain or loss of employment affects program
participation, which in turn affects recidivism outcomes. A determination of the impact
of housing concerns is lacking, but is necessary to inform future efforts to develop and
implement reentry programs and programming.

The Sub-Dimension of Employment

The obtainment of employment has traditionally been identiﬁed as a key
component of community-based correctional supervision since the implementation of
parole and probation as a correctional policy (Listwan et al., 2006; National Research
Council, 2008; Simon, 1993). Generally, a condition of supervision is to seek and obtain
employment (National Research Council, 2008). Failure to do so can produce violations
of supervision term charges that can lead to graduated sanctions or re-incarceration.
Among annual populations of re-incarcerated parolees, approximately 20% were returned
due to failures to obtain employment or other miscellaneous orders associated with

employment conditions (Hughes et al., 2001).

78

The overarching assumption that underlies the focus on employment and its
linkage to correctional supervision terms is the inherent degree of daily structure that is
created through employer supervision (Listwan et al., 2006; Simon, 1993). The
interrelationship between employer, employee, and other employees serves as a
mechanism of social control and the exposure to conventional others can enhance a sense
of being a productive member of society (Berg & Huebner, 2009; National Research
Council, 2008; Sampson & Laub, 1993; Uggen, 2000). In essence, employment has the
potential to create a tie and necessitate investment into conventional society. The
difﬁculty with this assumption is the fact that offenders experiencing the transition into
local communities are likely to reside in areas with few employment opportunities and
where other residents without criminal records are struggling to obtain employment
(Clear, 2007).

A majority of former prisoners were working in some capacity prior to
incarceration (Beck et al., 1993; Western & Beckett, 1999). Recent evidence has
suggested that while most were working at some point prior to incarceration, few had
been working 6 months prior to incarceration (Lattimore & Visher, 2009). Once
institutionalized, nearly all correctional facilities offer work programs or employability
programming, but it is unclear how institutional opportunities translate to real world
employment opportunities (Stephan, 2008; Stephan, 1997). Institutional programs and
programming can simulate real world experiences by enhancing work habits, gaining
experience in interactions with supervisors and co-workers, and balancing budgets
(Stephan, 2008; Stephan, 1997). This is a start, but the experiences are often not

sufﬁcient enough to be competitive in the labor market. Moreover, these experiences are

79

often not viable enough to be listed on a resume as a potential candidate for an
employment position (Stephan, 2008; Stephan, 1997).

Most ex-offenders are unemployed immediately following release, but gain some
type of employment well after release (Jacobs & Western, 2007; Mallik-Kane & Visher,
2008; Nelson et al., 1999; Steurer et al., 2002). The unemployment rate for ex-prisoners
is estimated to be between 25% and 40% (Petersilia, 2003). One of the most apparent
problems that can lead to behavioral regressions back to criminal activity surround
difﬁculties with low educational levels, few prior work experiences, and few skills
(Andrews & Bonta, 1994; Brooks et al., 2006; Lattimore & Visher, 2009; Petersilia,

2003; Solomon et al., 2004a). Even if working, only a few were able to obtain a

sustainable or livable wage20 (Harlow, 2003). The stigma of a felony record exacerbates
problems by making it difﬁcult to obtain and maintain livable wage employment (Holzer
et al., 2002).

Criminal history record information is often used by employers, higher education
institutions, and occupational licensure boards to reject applicants (Helgott, 1997; Pager,
2003; Solomon et al., 2004a). Service industry or manual labor employment is attainable
for many with criminal history records, but it is difﬁcult to transition from these positions
to open labor market positions with a generalized reluctance on the part of employers to
hire individuals with criminal records (Pager, 2003). Moreover, conditions of community
based correctional supervision can be a hindrance to maintaining employment once

gained after release (Solomon et al., 2004a). In all, these issues are likely to produce

 

20 Mallik-Kane and Visher (2008) have reported that among their sample of recently released prisoners,
men averaged a median monthly salary of $950 and women averaged a median monthly salary of $700.
The differences between these gendered sub-samples were not statistically signiﬁcant.

80

numerous short-lived employment opportunities instead of long-term careers (Solomon et
al., 2004a; Uggen, 2000).

There is a growing consensus that employment is one of the more important
turning points for ex-offenders (Huebner, 2005; National Research Council, 2008;
Sampson & Laub, 1993). Recent research has suggested that the turning point
effectiveness of employment interacts with age, with employment having more of an
effect on reducing the likelihood for future criminality among those over the age of 26
(U ggen, 2000). At the same time, there is mixed evidence on the effects of imprisonment
on subsequent employment opportunities. Much of the evidence suggests a negative
effect on future employability, employment options, length of time employed, and overall
wages (Holzer, 1996; Huebner, 2005; Pager, 2003; Urban Institute, 2008; Western, 2002;
Western & Pettit, 2005; Western et al., 2001). There has been some suggestion that there
is an initial negative effect that gradually erodes with the passage of time (Jacobs &
Western, 2007; Western & Beckett, 1999), and there is a potential for differential effects
by race with the erosion effect lingering longer for black individuals relative to white
individuals (Western & Pettit, 2005).

Others have suggested that imprisonment has no effect on subsequent
employment (Kling, 2006; Useem & Piehl, 2008). Instead, those who have been
incarcerated were likely to have had poor prospects for employment and were unable to
attain sustainable wages prior to institutionalization and will continue to do so upon
release. Useem and Piehl (2008) suggest that high levels of unemployment is more of a
product of “who they are and their immediate social environment, rather than a product

of their incarceration experience” (p. 157).

81

Employment opportunities are contingent upon the willingness of employers to
hire ex-offenders. Employers have a great deal of anxiety about the liabilities of hiring
an ex-offender and most are unwilling to do so (Pager, 2003; Solomon et al., 2004a). In
one of the more intriguing studies on employment and ex-offenders, Pager (2003)
matched groups of white and black males by appearance and work history and randomly
assigned individuals to an embellished criminal record condition in an effort to determine
which individuals would receive employment. The ﬁndings suggested differential and
interactive effects for race and criminal history records. Black males were less likely to
receive an employment offer than whites and those within the criminal record condition
were less likely than those without a record of criminal history to receive an employment
offer. Those most discriminated against for employment opportunities were black males
with criminal history records.

There is practical value in programming focused on employability training,
placement, and skill building since such programming is relatively easy to develop and
implement (National Research Council, 2008). Relatively speaking, employability
programs and programming utilize fewer and less complex factors than the content
needed for personal change, peer group, and familial relationship programming (National
Research Council, 2008). Yet, there are still few employability programs or
programming tied to reentry programs (Bloom, 2006). Instead, there is anecdotal
evidence to suggest that local service providers rely upon established referral based
networks (Justice Research Associates, 2005). These networks are often developed by

local service providers and require a referred individual to be motivated and competent

82

enough to follow through with many legwork duties (e. g., meeting referral times,
development of resume, interview skills, transportation to and from interviews, etc.)
The one bright spot for employability programs and programming is that there is
some evidence to suggest that such efforts are promising and can beneﬁt participants
(National Research Council, 2008; Visher et al., 2005). Unfortunately, much of the
evidence should be considered with caution since the observables generated from
research are typically based upon poor evaluative research designs (National Research
Council, 2008). Bloom et al. (2007) evaluated a transitional employment program that
provided immediate full-time employment for one to two months, employment assistance
and readiness courses once a week, as well as transportation and supermarket vouchers.
The provided employment opportunities were primarily grounds keeping and other forms
of manual labor for minimum wages. The ﬁndings suggested that immediate full-time
employment upon transition into the community was associated with reduced re-arrest
rates relative to a comparison group that only received employment search assistance.
The researchers suggested that the immediacy of employment appeared to be the key to
reduce re-arrest rates. Further analyses suggested that those who entered the transitional
employment program months after release did not have comparable reductions in re-
arrest relative to the comparison group.

Jacobs and Western (2007) evaluated a transitional program that combined
substance abuse treatment, housing, and employment services. The goal of the program
was to assess and place participants into programming within 48 hours of release into the
community. Substance abuse treatment involved bi-monthly drug screens in addition to

individual and group counseling for three to six months depending on progress made.

83

Transitional housing and employment services included residence in small shared
apartments with ﬁrll-time minimum wage manual labor jobs for approximately nine
months. Employment search assistance, service referrals, life skills classes, 12 step
programming, and vocational programming were available, but were not a requirement of
participation. Overall, the researchers found that participants were signiﬁcantly less

likely to recidivate relative to a comparison group matched on criminal history and
demographic indictors. Participants were three times more likely to be employed and
were also likely to have lower self-reported drug and alcohol use relative to the
comparison group.

The evaluations conducted by Bloom et a1. (2007) and Jacobs and Western (2007)
suggest that transitional employment services can reduce recidivism and potentially
increase employability as long as the employment services are bundled with a variety of
other services and the services begin immediately after release into the community. The
question of adequate “dosage” of employment services is not answered, but there is some
suggestion that the mere exposure to employment services can lead to the obtainment of
future employment and increased gains in social capital (Lattimore & Visher, 2009;
McGarrell et al., 2005). There is also some evidence to suggest that employment can
serve as an important stabilizer for other efforts to increase social capital.

Unemployment is likely to affect and lead to reductions in program participation
(Jacobs & Western, 2007). Rossman and Rossman (2003) suggest that full—time
employment can increase the number of interactions made with treatment providers and
increase participation in treatment programs, which can lead to reductions in levels of

criminal behavior. These effects are especially apparent with substance abuse treatment

84

programs. Employment services within comprehensive drug treatment programs appear
to lead to increases in full time employment and less drug use, even though there may not
be substantive reductions in recidivism (Rossman & Rossman, 2003).

The effect of employment on future recidivism is not clear. At best, employment
training and placement assistance will be associated with very small reductions in
recidivism (Aos et al., 2006; Homey et al., 1995; Listwan et al., 2006; MacKenzie, 2006).
Unemployment is often associated with recidivism at various post-release follow-up
intervals (Berg & Huebner, 2009; F inn, 1998; LaVigne et al., 2004). However, the risk
of recidivism associated with employment can be affected by the presence or absence of
quality social bonds with immediate relatives (Berg & Huebner, 2009). The Urban
Institute (2008) suggests that ex-offender employment decreases recidivism and higher
wages are negatively associated with the probability of recidivism. Work-release
employment programs that mimic the release transition into the community and promote
stable employment fail to produce consistent beneﬁcial effects for program participants
relative to comparison groups who do not participate in work-release programming
(MacKenzie, 2006). Additional research has suggested that employment placement and
income supplements were unable to reduce rates of recidivism or unemployment (Berk et
al., 1980; Piliavin & Gartner, 1981).

The stability of employment can also affect subsequent criminal behavior, with
some longitudinal research indicating that the more stable the employment, the less likely
one will participate in criminal activities (Sampson & Laub, 1990; Sampson & Laub,
1993). Additional evidence suggests that among youthﬁrl populations, higher rates of

crime are observed during phases of unemployment relative to phases of employment

85

(Farrington et al., 1986). Homey et al. (1995) found that the likelihood of desisting from
criminal behavior is greater after employment is gained relative to when employment is
either lost or not obtained. Similarly, the likelihood of committing a criminal offense is
twice as likely after employment is lost or not obtained relative to when employment is
gained. Many of the ﬁndings with regard to longitudinal efforts must be viewed with
caution. The studies commonly involve a number of problems surrounding issues of
small sample sizes and self-selection (Homey et al., 1995).

A number of unintended consequences have been associated with employability
programs and programming or employment in general. Berk et al. (1980) suggested that
an employment program that was based off of income supports (via monthly stipends)
was shown to reduce rates of property crime, but at the same time, created less incentive
to ﬁnd full-time employment. This suggests a need to explore the content of
employability components in reentry programs and programming to determine how to
gain and maintain employment, reduce crime, and ease the transition into the community.
Homey et a1. (1995) used a longitudinal sample of ex-offenders and found that
employment had a weak association with future criminality. One of the most interesting
secondary ﬁndings suggested that employment increased the likelihood of committing a
property crime. The researchers suggested that this effect may be a remnant of routine
activity or opportunity theoretical perspectives, with gains in employment providing new
chances to commit crimes.

Overall, the ﬁndings surrounding employment are not well understood and lack
consistency. Very few employment programs or programming have been subjected to

rigorous evaluation. Far fewer have been included as components of reentry programs or

86

programming and been subject to evaluation. Despite the marginal gains for participants
of employability programs and programming, there is still a common call that some form
of employment is better than no form of employment in reducing recidivism (U ggen,
2000). It is likely that employability programs and programming will continue to expand
and become integrated into reentry programs and programming. Additional research is
needed to explore how the gain or loss of employment affects recidivism, how the gain or
loss of employment affects participation in reentry programs or programming, and how
the gain or loss of employment affects program participation, which in turn affects
recidivism outcomes.
The Salient Sub-Dimension of Reentry: Substance Abuse Treatment

Existing reentry program model frameworks are designed to be as generalizable
as possible across an assortment of correctional sub-populations. It is assumed that the
processes of intake assessment and continued re-assessment will lead to the development
of individualized reentry plans that will be revised with progression in meeting plan
goals. As such, there is a necessary balance between generalizable reentry program
frameworks and individualized reentry program delivery. One must consider the
complex mix of interrelated reentry dimensions that can affect transitions into the
community within speciﬁc offender types to determine the balance between
generalizability and individualization.

One of the most prevalent correctional sub-populations with its own unique
reintegrative needs is those with severe substance abuse or dependency histories (Mallik-
Kane & Visher, 2008). The arrest and conviction of individuals for drug-involved crimes

has been one of the most dominant factors associated with the growth of state

87

correctional populations (Blumstein & Beck, 1999; Carver, 2004; Sevigny & Caulkins,
2004; Tonry, 1995; Useem & Piehl, 2008). Crime control philosophies and practices
associated with enhanced deterrence and incapacitative efforts (i.e., the “war on drugs”)
were utilized in effort to curb drug use and sales, which has contributed to a lag of
steadily increasing non-violent prison admissions for drug possession, manufacturing,
and trafﬁcking offenses (Useem & Piehl, 2008). Over half of state prisoners have used
illegal substances a month prior to their instant offense leading to incarceration and a
third were under the inﬂuence during the commission of their offense (Mumola &
Karberg, 2006). A majority of institutionalized persons report a history of substance use,
abuse, or dependence (Lattimore & Visher, 2009; Mumola, 1999; Mumola & Karberg,
2006)

Many of those in need of substance abuse and dependency treatment leave prison
with unresolved substance abuse issues (Mallik-Kane & Visher, 2008; National Research
Council, 2008). Few have received any form of substance abuse treatment or have been
exposed to treatment (Karburg & James, 2005; Lattimore & Visher, 2009; Mallik-Kane
& Visher, 2008; Winterﬁeld & Castro, 2005). One-third of state prisoners report past
participation in drug or alcohol abuse programming, but only 12% to 28% reported
participation while in prison (Mumola, 1999). Increasing budget constraints will likely
reduce the availability of correctional-based treatment programs for years to come
(Travis et al., 2001). Those with substance abuse and dependency histories also suffer
from high rates of physical and mental illnesses, which requires comprehensive substance
abuse, physical and mental illness intervention (Compton et al., 2003; Mallik—Kane &

Visher, 2008). Add these deﬁcits to the stigma and consequences of prison culture, the

88

lengthy isolation from the outside world, and the difﬁculty in meeting immediate
humanistic needs in the community and one can see how complex the reentry process is
for drug involved offenders.

Not surprisingly, those with severe substance abuse and dependency histories are
at higher risk of recidivism and tend to recidivate relatively soon after release relative to
other offender populations (Byme, 2008; Byme & Taxman, 2005; Carver, 2004; Mallik-
Kane & Visher, 2008; National Research Council, 2008; Urban Institute, 2008). Drug
use has been identiﬁed as the most powerful predictors of future involvement in crime
(Homey et al., 1995). The causal mechanism producing the differential recidivism rates
among those with substance abuse and dependency histories is not well understood.

Wilson and Davis (2006) suggest that ex-offenders with substance abuse
problems are primarily stigmatized for having a criminal record and secondarily
stigmatized due to drug consumption patterns, which further distances these types of
offenders from pro-social others in the community. There may also be systemic
inﬂuences, with conditions of community based correctional supervision affecting the
transition into the community (Solomon et al., 2004a). In general, those with such
histories generally have drug treatment referrals or placements and drug testing ordered
as conditions of community supervision (National Research Council, 2008). Positive
drug tests, failure to show for a drug test, and the failure to make treatment referrals or
attend treatment can lead to technical violations of supervision terms, which increase the
risk for recidivism. Among annual re-incarcerated parolee populations, approximately

16% were returned for violations related to drug testing (Hughes et al., 2001).

89

The combination of unresolved substance abuse problems and physical and/or
mental health issues often interact and contribute to difﬁculties in obtaining residential
placements and employment (Mallik-Kane & Visher, 2008; National Research Council,
2008). Individuals with substance abuse and dependency histories are less likely than
other correctional sub-populations to have secured housing prior to release (Mallik-Kane
& Visher, 2008). Additionally, the degree of housing instability is accentuated for those
with substance abuse and dependency issues (Mallik-Kane & Visher, 2008). Drug use is
associated with homelessness among recently released individuals (Mallik-Kane &
Visher, 2008; Nelson et al., 1999). Periods of abstinence or decreased use have been
associated with increases in the obtainment of employment (Hser et al., 2006). At the
same time, periods of use can exacerbate existing physical or mental health conditions
that condition the ability to work on a day-to-day basis (Mallik-Kane & Visher, 2008).
Relative to other correctional sub-populations, individuals with substance abuse or
dependency histories are more likely to earn wages through non-traditional means, often
associated with criminal activity (Mallik-Kane & Visher, 2008). Additional research has
suggested that relapsed use does not prevent the obtainment of employment, but does
reduce the likelihood of following through with treatment services that may prevent
relapse (Nelson et al., 1999).

Those with severe substance abuse and dependency histories represent a difﬁcult
population - one that is often viewed as being far less compliant than other types of
offenders and more reliant on their local community to meet immediate needs (Mallik-
Kane & Visher, 2008; Taxman, 2008; Visher & Farrell, 2005). Community based

substance abuse treatment is often perceived to be a viable and more effective alternative

90

to institutional based intervention (Aos et al., 2006; Chanhatasilpa et al., 2000). The
interventions are substantially less expensive than institution-based treatments and can
generate substantial cost savings for state correctional systems (Carver, 2004).
Supervision paired with treatment in the community is approximately one-third the cost
of incarceration (Urban Institute, 2008). Additionally, community based substance abuse
treatment programs are available in most jurisdictions although there are common
problems with treatment accessibility among those lacking private or public
transportation as well as those who reside in economically distressed areas (Visher &
Farrell, 2005).

The problem with community based substance abuse treatment programs is that
despite their general acceptance and perceived utility, their effectiveness is not well
established and relatively little is known with regard to their effectiveness when
compartmentalized within broader reentry programs or programming (Chanhatasilpa et
al., 2000; Cullen & Gendreau, 2000; Dowden et al., 2003; Lipton et al., 1975;
MacKenzie, 2006). The literature on community based substance abuse treatment is
expansive and there are a wide variety of programs that have been designed and used to
serve offender populations with generalized or speciﬁc abuses and dependencies. By
implication, the heterogeneity of service delivery, participant population serviced,
treatment modality utilized, and overall ﬁndings makes it difﬁcult to assess the overall
efﬁcacy of substance abuse treatment programming.

Confounding problems further are the research designs utilized to develop
empirical ﬁndings. Many studies of community-based substance abuse treatment

programs are poorly designed, leading to inﬂated and inaccurate results. Very few are

91

experimental or quasi-experimental, while the majority are descriptive or bivariate
(Onifade et al., 2008). Often ﬁndings are truncated with a focus on speciﬁc subsets of
participants or services provided in a broader treatment intervention (Palmer, 1983;
Palmer, 1994; Palmer, 1995). Successful program completions are usually the focus, to
the exclusion of the remaining pool of program participants who did not complete the
program for various reasons (e. g., drop-outs, unsuccessful discharges, program no longer
operational, etc.) (Palmer, 1983; Palmer, 1994; Palmer, 1995). One of the largest
problems is the lack of information on programming intensity and integrity. Measures of
program processes are rarely reported or captured at all, which leads to difﬁculties in
determining the effectiveness of programs and interpreting relevant program outcomes
(Onifade et a1, 2008).

With these caveats in mind, there are some promising indications that substance
abuse treatment and testing can reduce use and criminal behaviors that are assumed to
ease the process of transition back into the community (Andrews & Bonta, 2003; Anglin
& Hser, 1992; A05 et al., 1999; A05 et al., 2006; Brown et al., 2001; Butzin et al., 2002;
Carver, 2004; Chanhatasilpa et al., 2000; Hiller et al., 1999; Homey et al., 1995; Inciardi
et al., 2004; MacKenzie, 2006; Mallik-Kane & Visher, 2008; Martin etal., 1999;
National Research Council, 2008; Wexler, 1995). The effect sizes among these programs
are relatively small or marginal, with single digit reductions in recidivism for participants
relative to non-participants and wide discrepancies in overall program ﬁndings (Andrews
et al., 1990; A05 et al., 2006). It is not clear if treatment intervention directly or

indirectly affects the transition back into the community. Many evaluative undertakings

92

are unable to capture and control for treatment processes that contribute to and affect the
observed outcomes (Onifade et al., 2008).

There is a need to explore the differential reentry effects experienced by
correctional sub-populations with substance abuse histories and dependencies. It is
necessary to determine how speciﬁc dimensions of reentry interact and affect one another
to shape the process of transition. Research seeking to understand the process of reentry
is lacking. It would be useful to explore how participation in substance abuse treatment
is affected by the loss or gain of housing and employment. It would be equally useful to
determine how the relationships between treatment and the reentry concerns of housing
and employment interact and contribute to overall program outcomes such as recidivism.
Questioning the Associations between Treatment and Reentry

Reentry programs and programming form an overarching framework of which
substance abuse treatment and testing is one of many salient components. Existing
reentry program models assume that each and every component included within a reentry
model will ease the process of transition into the community and lessen the risk for
recidivism. It is not clear if this assumption is accurate. Available research has yet to
“rm-package” and analyze a reentry model to explore the dynamics of speciﬁc sub-
components to determine their effects and interactions on other sub-components, and how
these dynamics will in turn affect program outcomes. Many reentry programs have
become operational and will continue to expand without knowledge of how a
comprehensive set of services will affect participants. There is a danger that reentry

programs will continue to proceed blindly and simply utilize “kitchen sink” approaches,

93

which often lead to marginal, negative, or unintended program outcomes that do not
beneﬁt participants (Marlowe, 2006).

The association between substance abuse treatment and the reentry dimensions of
housing and employment is not clear. In terms of reducing recidivism, the effect sizes for
substance abuse treatment are small, but positive, suggesting that participants beneﬁt
from such programs (Aos et al., 2006; MacKenzie, 2006). The effect sizes for housing
programs and services and employment programs and services are not well established
and tend to be marginal at best (Andrews et al., 1990; A05 et al., 2006; MacKenzie,
2006). When combined in an overall program framework, it is not clear if these effect
sizes build off of the positive gains of substance abuse treatment to produce additional
gains, cancel one another out, or produce a multiplier effect contributing to marginal or
negative program outcomes. What is clear, however, is the acknowledgement that the
base rates of change produced by correctional programming will be rather small, often
producing single digit reductions in relevant program outcomes (Aos et al., 2006).

There is an established body of literature on substance abuse treatment, but there
is little knowledge on how the loss or gain of housing and employment affects treatment
participation and compliance with substance abuse treatment. There is some evidence to
suggest that stable participation in treatment is associated with stable housing and
employment (National Research Council, 2008). At the same time, there is evidence to
suggest that unstable participation in treatment is associated with stable housing and
employment (National Research Council, 2008; Nelson et al., 1999). It is likely that
unstable housing and/or employment will negatively affect participation in treatment, but

this may not always be the case.

94

At a fundamental level, it is still not clear how the philosophical orientations of
services offered in substance abuse treatment, housing, and employment programs mesh.
In some respects, there is a marriage between the services. Each of the services attempts
to provide structure and stability to daily life, while also holding an individual
accountable for their actions and conventional responsibilities. The services also seek to
increase pro-social functioning by means of increasing the social capital of individuals.
Moreover, the services seem to be speaking the same language with an emphasis on
ameliorating deﬁcits and are moving in the same direction. Substance abuse treatment is
undergoing a transformation and movement towards a broader focus on recovery
(Checinski & Ghodse, 2004; Maruna et al., 2004a). Relapse events are becoming more
and more tolerated as part of the process of recovery and encompassed into
comprehensive substance abuse programming (Checinski & Ghodse, 2004). Recovery is
therefore considered to be an overarching framework of which treatment is a sub-
component. A parallel process is occurring with reentry, where recidivism is being
viewed as an isolated event in the broader focus on reintegration into the community.
There is a growing recognition that reversions to past behavior will occur and these
reversions should form a foundation from which one can learn.

In other respects, there are conﬂicting and contradictory issues inherent to the
inclusion of substance abuse treatment, housing, and employment services within a single
comprehensive reentry program. For as much emphasis that is put into the development
of social capital for participants, there is an equal amount of punishment-based
accountability that underlies substance abuse treatment and can affect progress in

obtaining and maintaining housing and/or employment (Byme & Taxman, 2005).

95

Substance abuse treatment requires a variable degree of mobility to and ﬁom treatment

and testing center52]. Since treatment and drug testing is often included as a condition of
community-based supervision, there is a need for offenders to obtain relatively stable and
nearby housing that will maintain compliance with treatment orders. Failure to maintain
compliance with treatment can lead to graduated sanctions or violations of supervision
terms, which can indirectly or directly affect housing tenure. Given the increased
mobility and frequency of short-term residences (Clear, 2007; Metraux & Culhane, 2006;
Visher & Farrell, 2005), it is not clear if offenders can meet the demands of treatment and.
sustain housing. Treatment, in this instance, may have the unintended effect of producing
more residential instability rather than less.

Similar feasibility questions hold true when considering employment. Substance
abuse treatment requires a proportion of time that can affect employment. Some
employers may be able to work around an individual’s treatment and drug testing
schedule, while others will not be as accommodating (Pager, 2003; Pager, 2007). The
need to continuously take off work to meet treatment and testing schedules can lead to
the loss of employment, especially for individuals who are working low-wage, manual or
service orientated jobs (Pager, 2003; Pager, 2007). Consider the case of substance abuse
treatment programs with randomized testing. Individuals are often required to access a

randomized list hours before a test must be conducted. If an individual is randomly

 

21 If residential substance abuse treatment is required, there are also concerns of coerced mobility (Clear,
2007). If residential treatment is required upon release, an individual will transition out of an institution
directly to a residential center. Upon successful completion or termination of treatment services, the
individual will move to an approved residence. The movement equates to one additional residential move
relative to offenders under traditional community-based supervision. If residential treatment is required
any time during the supervision term, an individual will transition out of their residence, to a residential
facility, and to an approved residence. The movement equates to two additional residential moves relative
to traditional offenders.

96

selected on a given day and needs to work, a choice between remaining compliant with
supervision orders and testing or maintaining employment by going to work must be
made. Repercussions of either choice are likely, with graduated sanctions or violations of
supervision terms for a missed test, the loss of employment for a missed day of work, or a
combination of the two.

These unintended effects can also occur in the opposite direction. The loss of
suitable housing may make it difficult to meet treatment obligations, which can lead to
violations of supervision terms and create ﬁrrther instabilities in obtaining and
maintaining housing. The loss of employment also can lead to violations of supervision
terms due to non-compliance with treatment orders and can lead to further problems in
gaining employment. It is likely that these unintended consequences are further
exacerbated when there is a loss of housing and employment and the prevailing need to
meet the requirements of substance abuse treatment.

Summation and Direction of Research

There is a growing recognition and demand for the development and
implementation of reentry programs and pro gramming. Correctional administrators,
practitioners, program designers, and researchers are seeking to capitalize on the
newfound focus on rehabilitation, with treatment services beginning to take precedence
over crime control efforts (Byme, 2004; Byme, 2008; Byme & Taxman, 2005).
Comprehensive program models are being drafted and attempts have been made to
implement entire program models as well as piecemeal sub-components of program

models.

97

Many of the available reentry program models are structurally similar and also
contain comparable content. Over time, one model of reentry will become paradigmatic
and serve as the foundation for all other models. Currently, however, Useem and Piehl
(2008) considered the reentry movement to be nothing more than trial and error.
Continued experimentation will help to identify effective programs or program
components, but this initial phase does come at a cost. Every trial run results in an
expenditure of taxpayer funds. Without proven promise or sustained effectiveness it is
likely that these funds may become less and less available and truncate the reentry
movement.

Unfortunately, the rush of state correctional systems to integrate reentry structures
and practices is proceeding with a theoretical understanding that reentry programming
should work, but without much empirical knowledge concerning how and why these
programs should work. The available reentry program models appear to be
comprehensive for the sake of being comprehensive (Marlowe, 2006). Available
evidence on reentry dimensions is still largely compartmentalized and not well integrated.
The reentry dimensions of housing, employment, and substance abuse treatment often
include their own body of research, which narrows the focus to one aspect of reentry to
the exclusion of all others. Unfortunately, this narrowed focus does not reﬂect reentry
programs or programming in practice as each of these dimensions overlaps and interacts
with one another (see Palmer 1994; Palmer, 1995). Research has yet to adequately
address this issue and it is still not clear how reentry dimensions interact with one another

and contribute to the observed outcomes of reentry programs.

98

The current study seeks to deconstruct and explore how the process of transition
into the community is inﬂuenced by interrelationships between three speciﬁc sub-
components of reentry -— housing, employment, and substance abuse treatment —- among a
sample of offenders with severe substance abuse and dependency histories. The focus is
placed upon how these three sub-components of reentry inﬂuence program processes that
can indirectly and directly shape program outcomes. It is expected that differential
outcomes in program processes, relapse, and recidivism will be observed due to the
necessary balance that is needed in meeting the challenges of everyday life, remaining
compliant with treatment services, and attempting to reduce the risk of relapse and

recidivism.

99

CHAPTER III: DATA AND METHODS

Data for this study were obtained from a larger project that assessed the impact of
an intensive reentry-based substance abuse treatment program in a medium-sized,
industrialized, Midwestern city. The program was designed in an effort to meet dual
objectives of responding to individualized need and building or enhancing protective
factors that may counter-balance risks associated with relapse and recidivism. Protective
factors include restoring and maintaining bonds with family members, life skills training
to cope with immediate stressors, housing referral and placement, employment training
and placement, health care processing and obtainment, and additional needs as identiﬁed
by individualized assessment. In an effort to accomplish these objectives, the program
was designed to provide a comprehensive, integrated, and intensive reentry-based
substance abuse treatment program combined with frequent random drug testing across
the ﬁrst 12 months post-release.

The program sought to reach a target population of high-risk male offenders with

signiﬁcant substance abuse or dependency histories as identiﬁed through screening

inventory and case ﬁle reviews”. The research evaluation strategy utilized an
experimental design with the random assignment of eligible offenders to the program
participant group (treatment) and traditional community supervision group (control). The
control group was directly released according to traditional pre-release plans and
community based correctional supervision. This group may have participated in some

form of substance abuse treatment in the community and were subject to drug testing and

 

22 Determination of signiﬁcant substance abuse or dependency histories through assessment was
determined by Substance Abuse Subtle Screening Inventory (SASSI), which has an established level of
predictive validity (Lazowski & Boye, 1998). Offenders with high probabilities of substance dependence
(3) or severe dependence (4) formed the eligible pool of participants.

100

violations for non-compliance with testing procedures, but they do not participate in the
services provided to the participant group. These services included transitional housing,
service provider pre-release in-reach and subsequent program services, or receive
intensive case management oversight for the ﬁrst 12 months post-release (for additional
information on the larger project see Appendix A).

The ﬁnal sample consisted of 511 offenders, with 263 of the offenders being
randomly assigned to the participant group and the remaining 248 assigned to the control
group. The random assignment procedures were carefully followed and participant and
control groups had similar background characteristics in terms of demographics and
criminal and substance abuse histories. The comparable criminal and substance abuse
histories of those assigned to one of the two groups also indicated that the assignment
process reached its target population of hi gh-risk offenders in need of intensive services.
The average offender assigned to either group was likely to have at least one prior
juvenile probation term, two or three prior jail sentences, one or two prior prison
sentences, and an assessment score indicating the need for intensive out-patient substance
abuse treatment services.

The two year post-release follow-up of these 511 offenders served as the
foundation of the research23. Data were gleaned from the ofﬁcial management
information systems of the local Department of Corrections and State Police. These data
sources contained a plethora of narrative and quantitative information on individuals over

time. It is important to recognize that ofﬁcial data may have some degree of upward (i.e.,

 

23Informed by the work of Beck and Shipley (1989), a trend in correctional research is the use of 3 year
post-release follow-up periods to evaluate the impact of correctional programming on outcome indicators
of relapse and recidivism. The current study utilized a 2 year post-release follow-up in an effort to focus on

101

inﬂated, embellished) or downward (i.e., non-reported, overlooked) bias (Charnbliss &

Nagasawa, 1969; Gove et al., 1985; Hindelang et al., 1981; McCleary, 1977, McCleary,

197 8). Equally important is the fact that ofﬁcial data is found to be highly correlated

with victimization and self-report data (Gove et al., 1985; Hindelang et al., 1981). Given

this latter ﬁnding, there is merit to utilize ofﬁcial data sources24

Research Questions

The current research explored the interrelationships between reentry dimensions.
Most salient to this research are the dimensions of housing, employment, and substance
abuse treatment. These three dimensions are commonly included in reentry programming
frameworks, but are seldom subject to empirical analyses to determine their inﬂuence on
one another or for the broader process of transition into the community. The overarching
research framework examined the effects and interactions of these dimensions to identify
how these dimensions indirectly or directly affect program outcome indicators of relapse
and recidivism. It was expected that these reentry dimensions would inﬂuence and be
inﬂuenced by one another and these associations shape program outcomes.
The overarching research framework led to the formulation of a series of

research questions. Three research questions form the foundation for a series of analyses
that were used to examine the extent to which speciﬁc reentry dimensions affected and

Were affected by other reentry dimensions. The three research questions (RQ) included:

\

the immediate transitional experience of balancing a 2 year post-release correctional supervision term,
re'ﬁntry dimensions of housing, employment, and treatment, and the risk of relapse and recidivism.

24
The research focus on housing, employment, and treatment dimensions can also increase conﬁdence in

ﬁle use of official sources of data. Tracking of these dimensions is central to the function of correctional
supervisory agents. Housing prospects, employment prospects, and treatment enrollment must be reported
and veriﬁed by an agent. Once veriﬁed, the reported residence, employer, and treatment provider becomes

102

RQI .' Instability in housing and/or employment will negatively affect treatment dosage
and processes.

As detailed in Chapter 2, research suggested that offenders will generally have
less stable housing (Metraux & Culhane, 2006; Mallik-Kane & Visher, 2008; Visher &
Farrell, 2005) and less stable employment (Solomon et al., 2004a; Uggen, 2000) than the
general population, but there are variable levels of stability among offenders. Since the

procurement of housing and/or employment can provide structure and stability to daily
life, it was anticipated that instability in one or both of these dimensions affected the
stability of substance abuse treatment (see Hser et al., 2006; Jacobs & Western, 2007;
Malllik-Kane & Visher, 2008; National Research Council, 2008; Nelson et al.,. 1999).

R Q2: Instability in housing and/or employment will positively aﬂect outcomes of relapse
and recidivism.

This research question was extrapolated from the ﬁrst research question and seeks
to further elaborate the effects of housing and employment on the process of transition. It
iS probable that instabilities in housing and/or employment damage the structure and
Stability of daily life to such an extent as to place an individual at risk for relapse and
I‘eczidivism (see Jacobs & Western, 2007 ; Mallik-Kane & Visher, 2008; Metraux &
Culhane, 2006; National Research Council, 2008; Petersilia, 2003; Sampson & Laub,

1 993; Travis et al., 2001; Visher & Farrell, 2005). Instability in housing and/or
employment was predicted to increase levels of relapse and recidivism. In addition to

direct effects, it was also anticipated that instability in housing and employment would

\
“.16 foundation for subsequent collateral contacts. Any and all changes are required to be reported in a

t1mely fashion.

103

produce an interactive effect that contributed to levels of relapse and recidivism that were
more pronounced than the levels observed for the individual direct effects.

RQ3: Instability in housing and/0r employment will positively affect outcomes of relapse
and recidivism and this relationship will be mediated (or indirectly aﬂected by) treatment
dosage and processes.

The ﬁnal research question represents a culmination of the previous two research
questions. The observed increase in levels of relapse and recidivism that are expected
from instabilities in housing and/or employment (from research question 2) may be
contingent upon the effects of housing and/or employment instability on treatment dosage
and processes (from research question 1). There is an expected covariation between
treatment dosage, processes, and outcomes that is inﬂuenced directly or indirectly by
housing stability. It is likely that the loss of structure and stability associated with the
procurement of housing and/or employment will reduce the rehabilitative value
commonly associated with treatment to affect relapse and recidivism outcomes (see
Jacobs & Western, 2007; Mallik-Kane & Visher, 2008; National Research Council, 2008;
Petersilia, 2003; Travis et al., 2001; Visher & Farrell, 2005).

The research question suggests that levels of housing stability will affect
treatment dosage and processes, which will contribute to observed relapse and recidivism
outcomes. Similarly, levels of employment stability will affect treatment dosage and
processes, which will also affect relapse and recidivism outcomes. Overall it is expected
that instability in housing or employment will reduce levels of treatment dosage and
increase non-compliance with treatment, which will contribute to higher levels of relapse

and recidivism.

104

Once again an interactive effect is anticipated. High levels of housing instability
and employment instability are expected to reduce levels of treatment dosage received
and increase levels of non-compliance with treatment. In turn, the low levels of treatment
dosage received and high levels of treatment non-compliance will lead to higher levels of
relapse and recidivism.

Dependent Variables

A number of dependent variables will be used for this study. All measures have
been gleaned from ofﬁcial management information systems of the local Department of
Corrections and State Police. The unit of analysis is the individual and all variables will
be measured at the individual level across an individual’s community correctional
supervision term.

Treatment dosage represents the number of months an individual has participated
in substance abuse treatment programming. The tenure of participation in substance
abuse treatment is a common measure of treatment dosage in substance abuse treatment
literature (for reviews see Anglin & Hser, 1990; Simpson et al., 1995; Zhang et al.,
2003). This measure was constructed through the examination of case notes maintained
by an individual’s community correctional supervision agent. All case note entries are
date speciﬁc and provide a variable level of detail concerning treatment activities. These
narrative records were reviewed for contextualized content concerning the initial

enrollment in treatment, progress in treatment, and termination of treatment across an

. . . , . . . 25 . . .
mdrvrdual 5 community supervrsron term . The qualitative case note revrews served as

 

25 . . . . . . ,
It 15 important to note that srnce the measurement of treatment dosage was taken across an mdrvrdual s
supervision term, the dosage received may correspond to one or multiple substance abuse treatment

programs.

105

the foundation to produce quantitative measures that are included in the study (see
Tashokkori & Teddlie, 1998).

Measures of treatment processes can be used to provide a complete picture of
treatment participation by identifying underlying activities that contribute to treatment
dosage levels and overall treatment outcomes (Simpson et al., 1995). Unfortunately,
treatment processes are commonly overlooked in the literature (Onifade et al., 2008).
Two measures are used as indictors of treatment processes. Both measures represent
non-compliance with substance abuse treatment. Treatment program rule violations
represent the number of program violations incurred by individuals. The number of
violations is provided as a summary measure from the management information system
of the local Department of Corrections, but is based upon extrapolation algorithms of
community supervision agent case notes. Abscond is a dichotomous measure of whether
an individual absconded from community correctional supervision during their
supervision term. The measure represents the issuance of an abscond warrant by the
supervision agent for non-compliance with supervision reporting term526.

The remaining dependent variables are commonly associated with outcome
indicators used to assess the effectiveness of correctional programming and practice.
Relapse represents the proportion of positive drug tests relative to the total number of
drug tests administered. This measure was constructed through the use of summary
measures of the total number of positive drug tests observed and the total number of drug

tests submitted from the local correctional management information system.

 

26 This measure should not be confused with absence without leave (AWOL) or absconds from treatment.
While these may be associated with the issuance of an ofﬁcial abscond warrant for failure to corrrply with
supervision reporting terms, the two types of non-compliance may be entirely separate.

106

Relapse can be measured in an assortment of ways and a common measurement
approach is lacking in the literature (Wells et al., 1988). The measurement approach used
for this study is based upon a cumulative, longitudinal ratio that is standardized by the
dominator of the total number of tests administered. The standardization of measurement
allows one to differentiate between low levels of relapse (i.e., no positive drug tests or
very few positive tests relative to the number of tests administered) and high levels of

relapse (i.e., frequent positive drug tests relative to the total number of tests

administered)”.

Recidivism is measured in two ways from two different data sources. The
commonality between the measures of recidivism is the focus on the recidivism event (as
opposed to the timing of the event). Re-arrest is a dichotomous measure that identiﬁes
whether individuals were arrested, but not necessarily convicted, for a felony offense.
Arrest information is obtained through the management information system of the local
State Police. Re—incarceration is a dichotomous measure that identiﬁes whether
individuals had their community correctional status terminated and were returned to
prison. Returns to prison can occur in two ways. Technical violations of supervision
occur from a sustained violation of community correctional supervision leading to the
revocation of community status and re-incarceration on an existing charge. A new
commitment occurs if the individual is convicted of a new sentence of incarceration

while under community correctional supervision for an existing sentence. Both of these

 

27 One of the dangers with utilizing a ratio of positive drug tests to total number of drug tests is the fact that
denominators at the extremes may produce similar ratios. For instance a 1:1 positive drug test to total drug
test ratio would be considered to be the same as a 35:35 ratio. While the latter ratio may be a more severe
indicator of relapse difﬁculties than the former, the current study conceptualizes both ratios as being
comparable since they will require similar community correctional supervision responses (sanctioned or
non-sanctioned) and may be shaped by relative degrees of housing and/or employment stability.

107

measures of re—incarceration are obtained through the management information system of
the local Department of Corrections.

This study utilizes multiple measures of recidivism in order to increase
conﬁdence in the validity of measurement. Each ofﬁcial data source of recidivism has
limitations. State police arrest data is contingent upon the reporting of committed crimes
(Maltz, 1984). Correctional data is often shaped by organizational pressures and bounded
by partnered and hierarchical decision-making (Maltz, 1984; McCleary, 1977; McCleary,
1978). Maltz (1984) suggests that arrest data is a better indicator of individual conduct,
even if the arrest information does not include conviction. However, it is advisable to
supplement arrest measures with additional sources such as correctional records (Maltz,
1984)

Independent Variables

Two independent variables will be used for this study. Both of the measures have
been gleaned from ofﬁcial management information systems of the local Department of
Corrections and are measured at the individual level of analysis across one’s community
correctional supervision term. Housing stability represents the number of housing
movements made by individuals in the sample. Housing stability, movement, or mobility
are common terms in criminological literature, but the measurement options underlying
these terms are not entirely clear. The choice to utilize the number of housing moves is
informed by research that suggests that offenders are likely to move often due to reliance
on temporary accommodations (Metraux & Culhane, 2006; Mallik-Kane & Visher, 2008;

Visher & Farrell, 2005).

108

The housing stability measure was constructed through the examination of
community correctional supervision agent case notes (similar to the construction of the
measure of treatment dosage). The case notes were reviewed and analyzed for
contextualized content regarding address changes. Speciﬁc addresses were not provided,
but were rather spoken of in generalities (e. g., “moving to new apartment”) or relative to
new living arrangements (e. g., “requests approval to move in with girlﬁiend”). Any
housing changes must be conditionally approved and subsequently veriﬁed for ﬁnal
approval by an individual’s supervision agent per supervision term orders.

Efforts were made to capture any and all housing movements. A broad approach
captures traditional residential movement, movement to and from inpatient residential
treatment providers, and movement to and from graduated sanction facilities (e. g.,
technical violation centers). Conceptualizing housing movement in this fashion reﬂects
the totality of movement experienced by offenders — movement that is shaped by human
agency as well as correctional coercion (Clear, 2007). The only exclusions to the
measurement and aggregation of housing moves were for housing movements to/from
jails and to/from hospitals since these types of movement generally keep individuals
away from their place of residence for a short amount of time (i.e., less than a month).
Once again, these qualitative reviews led to the production of quantitative measures that
are included in the study.

It is important to note that the measure of housing stability may actually be
thought of as being a measure of housing instability. This is due to the fact the number of

housing moves is the basis for the measure. A high number of housing moves

109

corresponds to unstable housing, while a low number of housing moves reﬂects stable
housing.

Employment stability represents the number of months an individual has been
working. The choice to utilize the number of working months is informed by research
that suggests that tenure in the labor market can provide daily sources of stability and
structure (Listwan et al., 2006; National Research Council, 2008; Simon, 1993). The
construction of the measure was once again developed from community correctional
agent case notes (similar to the measures of treatment dosage and housing stability).

Agent case notes were examined for content concerning employment applications,

interviews, procurement, progress, and termination across an individual’s supervision

term28. Detail on the employer’s name, contact information, and type of employment
were available in some instances. Most often the description of employment would

simply pertain to work status (e. g., “currently working,” “working full-time”) or the type

9’ ‘6 ” ‘6

of employment (e. g., “janitorial services, construction, restaurant”) with additional
narrative detail. All case note entries are data speciﬁc, which allows for the extrapolation
of the length of employment for every employment opportunity. Agents are required to
check the employment statuses of offenders on their caseload per supervision term
orders. Periodic veriﬁcations of employment status with the individual’s employer are

made once employment is procured and the employer serves as an important collateral

contact for the agent.

 

28 . . . . .

Snmlar to the construction of the measure of treatment dosage, employment stabrlrty was measured
across an individual’s supervision term. This suggests that the number of months an individual was
enrployed may pertain to one or multiple employment experiences.

110

The case note reviews produced a quantitative measure that is included in the
study. Efforts were made to only capture those employment experiences that would
provide reasonably consistent and legitimate income. Key to this measurement
conceptualization is the thought of obtaining social capital that may contribute to upward
mobility (Wilson, 1990; Wilson, 1996). This did not necessitate full-time employment or
payroll based employment. However, this conceptualization did exclude sporadic
handyman jobs and chore based work for relatives that tended to represent one-time
employment opportunities with wages that could only provide support for a few days.
Control Variables

A number of variables will be used to control for individual demographics,
criminal and substance abuse history, and correctional supervision activity across the
supervision term. All of the measures are gleaned from the management information
system of the local Department of Corrections. Control measures for individual
demographics include age, race, educational background, and marital status. Age is
represented in years. Write is a dichotomous measure of race (1=white; 0=non-white).
Educational background consists of two separate dummy variables that use less than high
school diploma without a general equivalency degree (GED) as the reference category.
GED Education is a dichotomous indicator of those who received a GED degree
(1=GED; 0=reference category). High school graduate or more is a dichotomous
indicator of those who received a high school diploma and who may have continued their
education with college or professional degree coursework (1=High school graduate or
more; 0=reference category). Multiple measures of educational background were used

since it is expected that higher levels of education may be associated with higher levels of

111

social capital (Coleman, 1988). Marital status consists of two separate dummy variables
that use the status of single as the reference category. Divorced or widowed is a
dichotomous indicator of status (1=divorced/widowed; 0=single). Married is also used
as a dichotomized indicator (1=manied; 0=single). Multiple measures of marital status
are used since they may produce differential social capital effects (Coleman, 1988) and
effects on criminal behavior (Huebner, 2005; Laub & Sampson, 1993; Sampson & Laub,
1993)

Control measures for criminal and substance abuse history include institutional
history, current offense type, and participation in correctional substance abuse treatment
services. Juvenile commitment is a dichotomous measure of whether an individual has a
juvenile commitment history (1=at least one juvenile commitment; 0=no juvenile
comrrritrnents). Prior prison term is a dichotomous indicator of whether an individual
has a history of prison sentences (1=at least one prior prison term; 0=no prior prison
term). Multiple offense types are used in an attempt to differentiate effects that may
pertain to conviction offense categorizations (Rosenfeld, 2008). Drug offense serves as
the reference category. Person is a dichotomized indicator of conviction for a crimes
against person (1=person offense; 0=drug offense). Property is an indicator of conviction
for a property offense (1 =property offense; 0=drug offense). Public safety indicates a
conviction for a public safety offense (1=public safety; 0=drug offense). Prior
correctional substance abuse treatment is a measure of the number of exposures to and
participation in corrections based substance abuse treatment prior to release. These past

correctional services include institutional and corrnnunity based treatments.

112

 

 

Finally, control measures for activity across the supervision term include
supervision intensity, program completion, randomized group assignment,
hospitalization, mental health enrolhnents, and the pursuit of higher or professional
education. Supervision intensity measures the number of in-person contacts made by an
individual’s community correctional supervision agent. A higher level of in-person
contacts is conceptualized to refer to high levels of supervision intensity. Graduate of
treatment is a dichotomized measure of whether an individual successfully completed a
substance abuse treatment program (1=graduate of at least one treatment program; 0=did
not complete treatment program). Treatment group is a dichotomized indicator of
random assignment to the treatment group condition (1=treatment group; 0=control
group). Individuals in the treatment group participated in intensive, reentry-based
substance abuse treatment services while control group members received traditional
supervision services. Measures of supervision intensity, graduate of treatment, and
treatment group were constructed from summary measures within the ofﬁcial data source.

Hospitalization is a dichotomized indicator of hospital admission (1=admitted at
least once; 0=never admitted). Mental health treatment is a dichotomized indicator of
participation in mental health treatment services (1=received mental health services at
least once; 0=did not receive mental health services). This measure represents services
above and beyond substance abuse treatment and is not conceptualized as being
synonymous with substance abuse treatment services. Higher education is an indicator
of the enrollment and pursuit of college or professional degree coursework (1=pursued

higher education; 0=did not pursue higher education). Measures of hospitalization,

113

 

mental health treatment, and higher education were constructed upon review and analysis
of community correctional supervision agent case notes in within the ofﬁcial data source.
Analysis Methods and Modeling Strategy

The ﬁrst stage of the analysis will provide generalized descriptive information on
all of the variables included in this research. The second stage of analysis seeks to
explore the 3 research questions presented earlier in this chapter. Multivariate linear and
logistic regression models will serve as the foundation for the analyses. Linear regression
is a robust statistical technique that is used for continuous dependent variables and rests
upon assumptions of normally distributed variables and residuals, constant and
independent residual error structures across independent variables, and non-collinearity
among independent variables (Allen, 1997; Berry & Feldman, 1985; Berk, 2004; Fox,
1991). Logistic regression is a second robust statistical technique that is used for
dichotomous dependent variables (Agresti, 2002; Allen, 1997; Long, 1997; Pampel,
2000). Logistic regression eases the assumptions held by linear regression models and
also utilizes a method of coefﬁcient estimation (i.e., maximtnn likelihood) that is
different from linear regression (i.e., ordinary least squares) (Agresti, 2002; Allen, 1997;
Long, 1997; Pampel, 2000). Elaborations or modiﬁcations to regression models will be
made according to the results of the foundational linear or logistic models and
appendicized as necessary.

Multivariate modeling is used to assess the impact of the independent variables of
housing and employment stability on each of the dependent variables and answer the ﬁrst
two research questions. Regression analyses will be conducted step-wise. This approach

is used to observe the relative strength of the independent variables and note changes in

114

direction or magnitude that occur with the inclusion of additional variables into the
regression equation. The step-wise approach is not used in a traditional fashion with
forward or backward elimination techniques to develop “best” ﬁtting models (Agresti,
2002; Berk, 2004).

Four step-wise iterations will be examined. First, initial analyses will be
unconditional and ascertain the bivariate relationship between an independent variable
and a dependent variable. Second, multivariate analyses of both of the independent
variables without the control variables will be conducted. Third, a model consisting of
only the control variables will be produced. Finally, conditional models will be analyzed
that contain the independent variables and control variables.

Full conditional models will also make use of an interaction term. This term will
be used to determine if conditional models may be impacted by the combination of high
levels of housing and employment instability. The dichotomized interaction term will be
constructed from the observed median sample distributions of housing stability and
employment stability and entered into regression equations (Hardy, 1993; J accard, 2001;
J accard et al., 1990).

The ﬁnal research question requires the conditional modeling of the independent,
control, treatment dosage and treatment processes variables to ascertain their direct
effects on the overall dependent variables of relapse and recidivism. An interaction term
will be censored and presented for these full conditional models to assess the joint
inﬂuence of housing and employment instability. In addition to direct effects, the ﬁnal
research question requires the examination of indirect effects (or mediation). The method

used for the determination of indirect effects is based upon Baron and Kenny (1986).

115

The Baron and Kenny (1986) method requires a series of three steps. First, a
given independent variable must be associated with the outcome variable of interest

(established in research question 2). Second, the same independent variable must be

associated with a suspected mediator variable (established in research question 1)”.
Finally, the suspected mediator variable must be associated with the outcome variable of
interest after controlling for the effects of the given independent variable (established in
research question 3). All of these associations must be non-zero or statistically
signiﬁcant. If the three steps are observed, then partial indirect effects are likely. If the
ﬁnal step results in a non-signiﬁcant direct effect between the given independent variable
and outcome variable of interest, the results would suggest that the suspected mediator

variable completely mediates the relationship between the two variables.

 

29 The ﬁrst two steps in the Baron and Kenny method are exchanged with one another in order to follow
the research questions proposed in the study.

116

CHAPTER IV: RESULTS
Descriptive and Bivariate Statistics

Table 1 provides the descriptive and bivariate statistics for the control variables
conditioned by randomized group assignment. The average individual included in the
sample is a 35 year old, single, non-white male. Approximately 50% of the sample has
received a generalized equivalency degree (GED), 35% have not graduated from high
school or obtained a GED, and 15% have completed a high school degree and/or
continued forward in their education with college or professional coursework.

Twenty percent of the sample has had at least one prior juvenile commitment,
while 50% of the sample had served at least one prior prison sentence. In terms of the
current offense type, 32% of the sample was convicted on a person offense, 28% on a
property offense, 24% on a drug offense, and the remaining 16% on a public safety
offense. The sample averaged one prior enrollment in substance abuse treatment services
provided (and subsidized by) the Department of Corrections.

In terms of supervision activity across the supervision term, the sample averaged

45 in-person contacts with their supervision agent”. The intensity of supervision varied
widely by individual as indicated by the standard deviation coefﬁcient. Thirty-seven
percent of the sample completed and graduated from a substance abuse treatment
program during their supervision term. Twenty-three percent were hospitalized at least

one time during their parole term and 13% were enrolled in mental health treatment.

 

30 . . . . . . . . . . .

It becomes readily apparent that any drscussron of supervrsron mtensrty wrll refer to relatrve mtensrty.
The sample averaged 45 in-person contacts across a 24 month follow—up period equates to two contacts per
month. This rate of contact is consistent with national averages (Petersilia, 1999).

117

Seventeen percent enrolled in higher education, which includes college or professional
degree coursework.

Table 1: Control Variable Distributions (n=511).

 

 

Mean (SD)
Demographics
Age“ 35 (9.00)
White 33% (.47)
GED Education 50% (.50)
HS Grad or More* 15% (.36)
Less than HS (no GED) 35% (.48)
Divorced/Widowed 16% (.37)
Married“ 9% (.29)
Singlea 75% (~43)
Criminal and Substance Abuse History
At Least 1 Juv Commitment 20% (.40)
At Least 1 Prior Prison Term 50% (.50)
Term for Person Offense 32% (.47)
Term for Property Offense 28% (.45)
Term for Public Safety Offense 16% (.37)
Term for Drug Offensea * 24% (~43)
Num of Prior DOC SATX 1 (1.51)
Supervision Activity Across Supervision Term
Supervision Intensity 45 (25.32)
Grad of Any SATX" 37% (.48)
Hospitalized 23% (.43)
Mental Health Referral/Enroll 13% (.34)
_Enrolled in Higher/Pro Edu** 17% (.38)

 

*p<.05, **p<.01; NOTE: Standard deviations presented in parentheses.
a. Variable serves as the reference category for dummy variable series.

Table 2 presents the overall and group based descriptive statistics for the
independent variables. On average, individuals in the sample moved residences
approximately three times and were employed for approximately seven months during
their parole term. The standard deviation for these two measures is proportionately quite
large and suggests that individuals in the sample had a variable degree of residential
movement and employment experiences.

The overall distribution of the independent variables allows for the construction

of an interaction term that represents high levels of housing and employment stability.

118

Using the conditional distributions of the current sample, the dichotomized interaction
term identiﬁes median splits of the housing stability measure (i.e., individuals with four
or more residential moves) and employment stability measure (i.e., individuals who were
unemployed for more than half of their supervision term). Overall, 29% of the sample
met the criteria of being above the sample median of high residential movement and were
employed for less than half of their supervision term

In terms of substance abuse treatment processes, members of the sample received
a dose of approximately eight months of treatment. There is a substantial degree of
variability with regard to the number of treatment months between individuals. Members
of the sample appeared to be rather compliant with treatment participation, averaging one
treatment program violation across their supervision term. Compliance with treatment
did not translate to compliance with supervision terms. Forty-four percent of the sample

absconded from supervision at least once during their supervision term.

_Table 2: Independent Variable Distributions (n=511).

 

 

_ Mean (SD)
Housing Stability
Total Moves *** 3 (2.35)
Employment Stability
Num of Months Working 7 (6.81)
Interaction Term
Res*Emp Instability *** 29% (.45)
Substance Abuse Treatment Dosage
Estimated Months in TX*** 8 (5.79)
Substance Abuse Treatment Processes
Num of TX Prog Violations*** l (1.22)
\Absconded" 44% (.50)

 

*p<.05, **p<.01, ***p<.001; NOTE: Standard deviation presented in parentheses.

Table 3 provides descriptive information for the dependent variables. The
average proportion of positive drug tests for the entire sample is .23. This suggests that

approximately 23% of the drug tests administered resulted in a positive screen. Twenty-

119

eight percent of the sample were re-arrested and 33% were re-incarcerated during their
supervision term.

Table 3: Dependent Variable Distributions (n=51l).

 

 

Mean (SD)
Proportion Positive Tests* .23 (.26)
Re-Arrested 28% (.45)
Re-Incarcerated 33% (.47)

 

*p<.05, **p<.01, ***p<.001; NOTE: Standard deviation presented in parentheses.
Research Question 1: Unconditional, Partial Conditional, and Conditional Multivariate
Analyses of Treatment Dosage and Processes

The goal of the ﬁrst phase of analysis was to examine the relationships between
housing and employment stability and substance abuse treatment process indicators of
treatment dosage, treatment violations, and absconding. Linear and logistic regressions
were utilized to assess the relationships. Models were presented stepwise, beginning with
unconditional bivariate models and ending with fully conditional models that include the
independent variables of housing and employment stability, all of the control variables,
and an interaction term of housing and employment instability.

Table 4 provides the unconditional and partial conditional linear regression
models. Housing and employment stability were signiﬁcantly and positively related to
the amount of substance abuse treatment dosage received in unconditional bivariate
models. This effect were also present for the partial conditional multivariate model that
controls for the effect of housing stability and employment stability. The standardized
coefﬁcients suggest that housing stability was a moderate predictor of treatment dose,
while employment stability is a weaker predictor. This strength of prediction was also

reﬂected in the explained variance. Housing stability explained approximately 18% of

120

the variation of months in substance abuse treatment, while employment stability only
explains 5% of the variance.

Table 4: Partial Linear Regressions of Substance Abuse Treatment Dosage on
Independent and Control Variables (n=511).

 

 

 

b (SE) StdB b (SE) StdB b (SE) StdB b (SE) StdB
Age .02 (.03) .03
White" -.42 (.41) -.03
GEDa -.15 (.41) -.01
HS Grad Plusa .28 (.57) .02
Divorce/Widowb .86 (.55) .05
Marriedb -.19 (.66) —.01
Prior J Commite -.40 (.47) -.03
Prior Prisone .23 (.42) .02
Person Crimec .08 (.51) .01
Property CrimeC -.37 (.52) -.03
PSafety Crimec -.54 (.59) -.03
Past rx Services -.05 (.12) -.01
Sup Intensity .08 (.01)
.35***
Grad ofrx" 4.47 (.41)
.37***
Tx Groupd 3.94 (.38)
.34***
Hospitalizede 1.20 (.44)
.09**
MH Treatment" -.06 (.57) -.00
H Educ .81 (.50) .05
House Stability 1.05 (.10) 1.07 (.09)
.43*** 44*“
Emp Stability .19 (.04) .20 (.03)
.22*** .24***
TR-Square .18 .05 .24 .52
Model F ll3.87*** 26.19*** 7973*” 2935*”
\df 1 1 2 18

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance.

121

Unexpectedly, as the number of housing movements increased the amount of
substance abuse treatment dose also increased. The logic of the research question
assumed that housing instability would reduce the levels of treatment dosage received,
but this does not appear to be the case among the partial models that do not include
control variables. Employment stability appeared to be consistent with the logic of the
research question. The more stable the employment in terms of the number of months
worked, the greater the dose of substance abuse treatment.

A number of the control variables related to supervision activity across the
supervision term were also signiﬁcantly related to substance abuse treatment dosage. As
the intensity of supervision increased, the dosage of substance abuse treatment also
increased. Graduates of substance abuse treatment programs, members of the treatment
group, and those that were hospitalized at least once during their supervision term
received higher dosage levels of treatment than those who did not graduate from a
treatment program, were members of the control group, or were not hospitalized. These
variables appear to be moderate predictors of treatment dose, with the exception of
hospitalization which is a very weak predictor. It is important to note that the control
variables appear to explain approximately 52% of the variance of substance abuse
treatment dose irrespective of the main independent variables of housing and
employment stability.

Table 5 provides the full conditional linear regression models. In general, most of
the unconditional and partial conditional effects remained the same in the full conditional
models. Housing and employment stability continue to be positively associated with

substance abuse treatment dose after controlling for control variables, one another, and an

122

interaction term“. Similarly, supervision intensity, substance abuse treatment graduates,
and members of the treatment group are positively associated with treatment dose.
Housing stability continued to be a stronger predictor of substance abuse
treatment dose relative to employment stability. However, both of these predictors were
weaker than the control predictors of supervision intensity, substance abuse treatment
program graduates, and assignment to the treatment group. It is also important to note
that the explained variance increased very little with the inclusion of housing and
employment stability measures. The model that only included control variables
explained 52% of the variance and the model that includes all of the variables explains an
additional 4% of the variance. This suggested that housing and employment stability

may not be contributing much to the prediction of treatment dose.

 

31 The results for the primary independent variables of housing and employment stability in the full
conditional models were conﬁrmed through negative binomial regression with robust standard errors. See
Appendix B for further information.

123

Table 5: Full Linear Regressions of Substance Abuse Treatment Dosage on Independent
and Control Variables @=511).

 

 

 

b (SE) StdB b (SE) StdB b (SE) StdB b (SE) StdB
Age .01 (.02) .02 .02 (.02) .04 .02 (.02) .03 .02 (.02) .03
Whitee -.58 (.39) -.05 -.75 (.42) -.06 -.98 (.40) -.98 (.40)
-.08* -.08*
GEDa -.10 (.39) -.01 -.23 (.40) -.02 -.19 (.39) -.02 -.19 (.39) -.02
HS Grad plusa .09 (.55) .01 .19 (.56) .01 -.03 (.54) -.00 —.03 (.54) -.00
Divorce/Widowb .91 (.53) .06 .76 (.55) .05 .80 (.53) .05 .79 (.53) .05
Marriedb -.13 (.64) -.01 -.26 (.65) -.01 -.21 (.63) -.01 -.21 (.63) -.01
Prior 1 Commite -.48 (.45) -.03 -.36 (.47) -.02 -.45 (.45) -.03 -.45 (.45) -.03
Prior Prisone .12 (.41) .01 .07 (.42) .01 -.08 (.40) -.01 -.07 (.41) -.01
Person Crimec -.02 (.49) -.00 -.02 (.50) -.00 -.14 (.48) -.01 -.14 (.48) -.01
Property Crimec -.42 (.50) -.03 -.27 (.51) -.02 -.31 (.49) -.02 -.31 (.50) -.02
PSafety Crimec -.45 (.57) -.03 -.62 (.59) -.04 -.54 (.57) -.03 -.54 (.57) -.03
Past TX Services -.14 (.12) -.04 -.02 (.12) -.Ol -.11 (.12) -.03 -.ll (.12) -.03
Sup Intensity .07 (.01) .08 (.01) .07 (.01) .06 (.01)
,30*** _34*** .29*** .29***
Grad ofTXe 4.21 (.40) 4.11 (.42) 3.76 (.41) 3.76 (.41)
-35*** .34*** .31*** .31***
Tx Groupd 3.45 (.38) 3.99 (.38) 3.48 (.37) 3.48 (.37)
.30*** .34*** .30*** .30***
Hospitalizede .79 (.43) .06 1.20 (.43) .76 (.42) .06 .76 (.42) .06
.09**
MH Treatmente -.17 (.55) -.01 .15 (.57) .01 .08 (.55) .01 .08 (.55) .01
H Educ .95 (.48) .65 (.50) .04 .76 (.48) .05 .76 (.48) .05
.06*
House Stability .49 (.08) .52 (.08) .50 (.11)
.20*** .21*** .20***
Emp Stability .09 (.03) .11 (.03) .11 (.03)
.ll*** .13*** .13***
House*Empf .12 (.60) .01
R-Square .55 .53 .56 .56
Model F 31.39*** 28.76*** 3135*“! 29.80***
(if 19 19 20 21

 

‘p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

124

There are some exceptions to the ﬁndings from previous unconditional models”.
The measure of participation in higher education via college or professional degree
coursework becomes signiﬁcant once housing stability is controlled. Those who are
participating in higher education during their supervision term receive a higher dosage of
substance abuse treatment than those who are not seeking higher education. This effect
becomes insigniﬁcant once employment stability is controlled but does not appear to
change in direction or strength. Race was signiﬁcant in the full models that control for
housing stability and employment stability with or without the interaction term of
housing and employment instability. On average, whites received a lower dose of
substance abuse treatment than non-whites.

The interaction term of housing instability and employment instability did not
provide any additional explanatory power to the variables of housing and employment
stability. Individuals who have experienced substantial housing moves and who lacked
stable employment did not experience an increase or a decrease in substance abuse
treatment relative to those with relatively more stable housing and employment. The
signiﬁcance of the variables of housing stability and employment stability with the
inclusion of the interaction term indicates that each of these variables has a direct effect
on treatment dosage.

Table 6 provides the unconditional and partial conditional linear regression

models of the number of substance abuse treatment program violations on independent

 

32 Once housing stability is controlled for, the measure of hospitalization becomes insigniﬁcant. The
direction and strength of the predictor does not change, but does appear to interact and become subsumed
by housing stability. This ﬁnding differs from previous unconditional models, but was not replicated in full
conditional negative binomial regression models with robust standard errors (see Appendix B for further
information). The negative binomial regression models suggested that individuals who had been

125

and control variables. Housing and employment stability were statistically signiﬁcant,
but their effects are in two different directions. Housing stability was a moderate
predictor of substance abuse treatment program violations as indicated by the
standardized coefﬁcient of .46 and the ability to explain approximately 21% of the
variance in the number of treatment violations. Employment stability was a relatively
weak predictor and only explains approximately 4% of the variance in treatment
violations. In combination, housing and employment stability appeared to provide
slightly more explanatory power (explaining 25% of the variance in treatment program
violations) than the control variables (explaining 22% of the variance in treatment
program violations).

Housing stability was positively associated with the number of substance abuse
treatment violations. Individuals who have experienced more housing movements are
also likely to have experienced more treatment program violations. Employment stability
was negatively associated with the number of treatment violations. Individuals who have
more stable employment are likely to have fewer treatment violations. Both of these

effects are in the direction expected by the logic of the research question.

 

hospitalized during their supervision term were 1.15 times more likely to have higher proportions of
positive drug tests than those who had not been admitted to a hospital.

126

Table 6: Partial Linear Regressions of Treatment Program Violations on Independent
and Control Variables (n=511).

 

 

 

b (SE) StdB b (SE) StdB b (SE) StdB b (SE) StdB
Age .01 (.01) .06
Whitee .23 (.11)
09*
GEDa -.15 (.11) -.06
HS Grad Plusa -.13 (.15) -04
Divorce/Widowb -.02 (.15) -.01
Marriedb -.07 (.18) -.02
Prior J Commite .19 (.13) .06
Prior Prisone -.11 (.l l) -.05
Person Crimec .30 (.13)
.11*
Property (:timec .37 (.14)
.14**
PSafety Crimec .09 (.16) .03
Past TX Services .16 (.03)
.20***
Sup Intensity .01 (.002)
. 154:1“!
Grad of Tx" -.53 (.11)
_21***
Tx Groupd -73 (JO)
.30***
Hospitalizede .18 (.12) .07
MH Treatment" -.22 (.15) -.08
H radiie -.29 (.13) -.07
House Stability .24 (.20) .24 (.02)
.46*** .46***
Emp Stability -04 (.01) -.03 (.01)
ﬂ21*** ‘19***
R-Square .21 .04 .25 .22
Model F 139.67*** 2302*" 85.54*** 7.82***
df 1 1 2 18

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefficient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance.

A number of control variables continue to inﬂuence the relationships with
treatment processes in general and more speciﬁcally the number of treatment program

violations. Race, current conviction offense type, past exposure to previous correctional

127

substance abuse treatment, and supervision activity all appear to inﬂuence treatment
program violations. Whites who were convicted of a person or property crime and were
exposed to a number of prior correctional substance abuse treatments had higher numbers
of treatment program violations, on average, than non-whites, those convicted of a drug
crime, or those with few or no past exposures to correctional substance treatment.
Relative to supervision activity across the supervision term, as the intensity of
supervision increases, the number of observed treatment program violations increases.
Individuals assigned to the treatment group were likely to have more treatment program
violations than members of the control group. Finally, those who successfully completed
a substance abuse treatment program often had fewer treatment program violations than
those who did not complete a treatment program.

Table 7 provides the full conditional linear regression models of the number of
substance abuse treatment program violations on independent and control variables.
Housing stability continued to be signiﬁcantly related to the number of substance abuse
treatment violations in the conditional models. The standardized coefﬁcients suggested
that housing stability has the strongest effect on the prediction of treatment program
violations across all of the conditional models, with more housing movement leading to

more violations. Employment stability was negatively related to the number of substance

abuse treatment violations in each of the full conditional models”.

 

33 The results for the primary independent variables of housing and employment stability in the full
conditional models were conﬁrmed through negative binomial regression with robust standard errors. See
Appendix B for further information.

128

Table 7: Full Linear Regressions of Treatment Program Violations on Independent and

 

 

 

Control Variables
Std B (SE) Std B (SE) Std B (SE) Std B (SE)
Age .01 (.01) .03 .01 (.01) .05 .00 (.01) .03 .00 (.01) .03
Whitee .16 (.10) .06 .35 (.11) .25 (.10) .24 (.10)
.13" .10* .10M
GEDa -.12 (.10) -.05 -.12 (.11) -.05 -.10 (.10) -.04 -.11 (.09) -.03
HS Grad plusa -.22 (.14) -.06 -.10 (.15) -.03 -.19 (.13) -.06 -.20 (.14) -.05
Divorce/Widowb .00 (.13) .00 .01 (.14) .00 .03 (.13) .01 -.01 (.13) -.01
Marriedb -.04 (.16) -.01 -.04 (.17) -.01 -.02 (.16) -.01 .01 (.16) .01
PrjOI-Jcommite .15 (.11).05 .18(.12).06 .15 (.11).05 .16 (.11).05
Prior Prisone -.16 (.10) -.07 -.05 (.11) -.02 -.12 (.10) -.05 -.10 (.10) -.04
Person Crimec .25 (.12) .10* .33 (.13) .28 (.12) .26 (.12) .10
.12* .11*
Property Crimec .34 (.12) .34 (.14) .32 (.12) .32 (.12) .11
.13** ,12* '12:”:
PSafety Crimec .13 (.14) .04 .12 (.16) .04 .16 (.14) .05 .18 (.14) .06
Past TX Services .12 (.03) .15 (.03) .11 (.03) .10 (.03)
-15*** .19*** .14*** .13***
Sup Intensity .00 (.002) .04 .01 (.002) .00 (.002) .05 .00 (.002) .03
_l6***
Grad ofTXe -.66 (.10) -.41 (.11) -.56 (.10) -.57 (.10)
--26*** -.16*** -.22*** -.23***
TX Groupd .50 (.09) .72 (.10) .50 (.09) .51 (.09)
.21*** .29*** 20*“: 2111:“:
Hospitalizede -.01 (.11) -.00 .19 (.11) .07 .00 (.10) .00 -.02 (.10) -.01
MI-I Treatmente -.34 (.14) -.37 (.15) '.40 (.14) -.38 (.13)
-.09** -.10* -.11** -.11**
H Educ -.15 (.12) -.05 -.16 (.13) -.05 -.11 (.12) -.03 -.12 (.11) -.03
House Stability .23 (.21) .22 (.02) .14 (.03)
.44*** _43*** 26mm:
Emp Stability -.03 (.01) -.02 (.01) -.01 (.01) -.06
_.18*** _.13***
House"‘Empf .63 (.14)
.25***
R-Square .37 .25 .39 .43
Model F 15.53*** 8.50*** 15.61*** 17.81***
df 19 19 20 21

 

‘p<.05, **p<.01, **"‘p<.001; NOTE: Unstandardized coefficients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.
a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

129

The signiﬁcance of the interaction term suggested that individuals with 4 or more

residential moves who have been employed for less than half of their parole term have

more treatment program violations on average than all other individuals in the sample”.
The ﬁnding also suggested that the direct, positive effect of housing movement on
treatment program violations was conditioned by employment stability. The
insigniﬁcance of the coefﬁcient for employment stability suggested that the partial effect
of employment stability was not conditioned by housing stability in the full linear
regression model. However, the replication of the ﬁlll linear regression model with a
negative binomial regression model with robust standard errors estimated a signiﬁcant
employment stability coefﬁcient after controlling for the interaction term (see Appendix
B for further information). This suggested that the negative relationship between
employment stability and treatment program violations was conditioned by housing
stability.

The full conditional models explained between 39% and 43% of the variance in
the number of substance abuse program violations, which was a 17% to 21% increase in
explained variance from the conditional model that only included the control variables
and a 14% to 18% increase from the conditional model that only included the variables of
housing and employment stability. In addition to the strong effects for housing stability,

a number of control variables had consistent effects across all of the conditional

models”. Past enrollment in correctional substance abuse treatment and assignment to

 

34 The result for the interaction term in the full conditional model was conﬁrmed through negative
binomial regression with robust standard errors. See Appendix B for further information.

35 . . . . .
Control vanable measures of current convrctron offense type remamed consrstent across the full
conditional models when models were estimated using negative binomial regression with robust standard

130

the treatment group continued to be positively associated with treatment program
violations. Individuals who graduated from substance abuse treatment during their
supervision term had less treatment program violations on average than those who failed
to complete treatment. These effects were also conditioned by the interactive effect of
housing and employment instability.

The full conditional models also modiﬁed some of the previous ﬁnding ﬁ'om the
partial conditional models. Race continued to be positively related to treatment program
violations with whites having more violations on average than non-whites, but was
unrelated to the model which. only controls for housing stability. The measure of housing
stability may have subsumed the effect of race, but it is also likely that the effect of
employment stability is associated with or conditioned by race (see Pager, 2007).
Supervision intensity was found to be insigniﬁcant for most of the conditional models
except for the model which controls for employment stability. It appears that the effect
of supervision intensity was associated with and truncated by the measure of housing
stability. The standardized coefﬁcient for supervision intensity was .15 in the conditional
model that only includes the control variables and .16 in the conditional model that
includes the control variables and employment stability. Conditional models that
included the control variables and housing stability, control variables with housing and
employment stability, and control variables with housing stability, employment stability,
and the interaction term produced standardized coefﬁcients of .04, .05, and .02

respectively.

 

errors (see Appendix B). Individuals convicted of crimes against persons or property crimes had more
treatment program violations, on average, relative to individuals convicted of drug offenses.

131

One unanticipated ﬁnding was also observed with the full conditional models.
Enrolhnent in mental health treatment during the supervision term was negatively
associated to the number of substance abuse treatment program violations once housing
and employment stability were controlled. That is, those with mental health issues had
fewer treatment program violations than those without mental health issues. This effect
was also observed for the ﬁll] conditional model that included an interaction term. It is
possible that the identiﬁcation of mental health needs can lead to provider ﬂexibility in
determinations of treatment program violations and/or compliance. It is equally probable
that this same type of provider ﬂexibility is given to those convicted of drug offenses,
while those convicted of persons or property crimes are given less ﬂexibility.

For the ﬁnal treatment process consideration, Table 8 provides the unconditional
and partial conditional logistic regression models of supervision absconding on
independent and control variables. The partial models suggested that the relative
explanatory power of the models is quite low, with pseudo variance explained
coefficients near 10% to 20%. The decreasing log-likelihood ratio statistic suggested that
the additional variables included to the models are providing better model ﬁt than
simplistic models with few variables. Overall, the partial conditional model that only
includes control variables had more explanatory power and better model ﬁt than models

which include housing and/or employment stability.

132

Table 8: Partial Logistic Regressions of Supervision Absconding on Independent and

 

 

 

Control Variables
b (SE) ExpB b (SE) ExpB b (SE) ExpB b (SE) ExpB
Age -.02 (.01) .98
White“ -.10 (.22) .91
GEDa .06 (.22) 1.06
HS Grad Plusa -.08 (.31) .92
Divorce/Widowb .09 (.30) 1.10
Marriedb -.24 (.37) .78
Prior J Commit“ .51 (.25)
1.66*
Prior Prison“ .43 (.23) 1.54
Person Crimec .28 (.27) 1.33
Pro Crime“ .64 (.28)
pm 190*
PSafety Crime“ .08 (.32) 1.09
Past TX Services .25 (.07)
1.28***
Sup Intensity .00 (.004) 1.00
Grad ofrx“ -1.15 (.23)
.32***
Tx Groupd -77 (21)
2.17***
Hospitalized“ .25 (.23) 1.29
MB Treatment“ .13 (.30) 1.14
H Edu“ -.65 (.28)
52*
House Stability .25 (.04) .26 (.04)
l.28*** 130*“:
Emp Stability -.08 (.02) —.09 (.01)
.92*** .91***
Cox and Snell R2 .07 .07 .14 .17
Nagelkerke R2 .10 .09 .19 .22
~2LL 661.72 663.72 624.12 608.80
Model C1112 39.37*** 37.38*** 76.98*** 92_30*c*
df 1 1 2 18

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.
a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, C. Contrasted
against the reference category of the non-instance.

Housing stability and employment stability were found to be statistically

signiﬁcant in opposing, but expected, directions as speciﬁed in the research question. In

133

terms of housing stability, more housing movement is associated with a greater likelihood
of absconding. With each additional housing movement the odds of absconding increase
by a factor of just over 1. Employment stability was negatively related to absconding.
Individuals who have stable employment were less likely to abscond.

Control variables of criminal and substance abuse histories as well as supervision
activity across the supervision term were also associated with absconding. Individuals
who had previously been adjudicated and committed to a juvenile facility at least one
time were just less than two times more likely to have absconded than individuals who
were not committed as a juvenile. Individuals who were convicted of a property crime as
their instant offense were nearly two times more likely to abscond relative to those who
were convicted on drug offenses. Past enrollment in correctional substance abuse
treatment services was also positively associated with absconding. The odds of
absconding increase 1.28 times for every exposure to correctional substance abuse
treatment services.

Relative to supervision activity, individuals assigned to the treatment group were
just over 2 times more likely than members of the control group to abscond. Individuals
who successfully completed a substance abuse treatment program during their
supervision term were less likely to abscond relative to those who were unable to
complete a treatment program. This negative effect on absconding was also observed for
enrollment in higher education. Individuals who pursued college coursework or
professional degrees were less likely to abscond than those who did not pursue advanced

education.

134

Table 9 presents the full conditional logistic regression models of absconding on
independent and control variables. Housing and employment stability were signiﬁcantly
related to absconding in the expected direction, with increased housing movement

increasing the likelihood of absconding and increased employment stability decreasing

the likelihood of absconding36. The ﬁrll conditional models provided additional
explanatory power compared to the previous unconditional and partial conditional models
as indicated by the conservative Cox and Snell pseudo explained variance. However, the
proportion of explained variance hovered around 20% to 26%. The log-likelihood ratio
statistic for the full conditional models continued to be lower than the unconditional and
partial conditional models, but the rate of reduction was not very large (and in fact
increases for the partial conditional model that only controls for employment stability).
Nevertheless, the addition of housing and employment stability variables contributed to a
better ﬁtting model.

The effects of control variables for prior substance abuse treatment history, and
current treatment activity continued to be signiﬁcantly related to absconding across all of
the ﬁll] conditional models”. Prior treatment history was positively associated with
absconding. Individuals who had experienced more exposure to correctional substance
abuse treatment were more likely to abscond. It is important to note that his effect

persisted even after controlling for prior prison admissions. This suggested that increased

 

36 The results for the primary independent variables of housing and employment stability in the full
conditional models were conﬁrmed through negative binomial regression with robust standard errors. See
Appendix B for further information.

37 Individuals convicted of a property crime appeared to be nearly two times more likely to abscond than
those convicted for a drug offense in the full conditional logistic regression models. The signiﬁcance of the
effect was not observed in full conditional models that utilized negative binomial regression with robust
standard errors (see Appendix B for further information).

135

exposure to correctional substance abuse treatment inﬂuenced absconding regardless of
whether an individual has no history or a lengthy history of prison sentences. The group
effect of assigrnnent to the treatment condition and participation in the treatment program
was observed. Members of the treatment group were two times as likely to abscond
relative to individuals in the control group that experience traditional supervision
services. Successful completion of a substance abuse treatment program continued to be
negatively associated with absconding. This suggested that treatment may serve as a
protective factor and decrease the likelihood of absconding for those who complete
treatment.

Variables in the partial conditional control model were inﬂuenced by the
inclusion of housing and/or employment stability variables in terms of signiﬁcance, but
did not change in direction or relative magnitude. Full conditional models that controlled
for the effect of housing stability reduce and effectively negated the effect of past
juvenile commitments. This suggested that a juvenile criminal history may partially
inﬂuence employment stability, but does not inﬂuence housing stability or contribute to
the effects of employment stability once housing stability is controlled. Enrollment in
college coursework or professional degrees continued to have a negative effect on
absconding, but this effect becomes insigniﬁcant once housing stability and employment

stability were controlled.

136

Table 9: Full Logistic Regressions of Supervision Absconding on Independent and
Control Variables (n=51 1).

 

 

 

b (SE) ExpB b (SE) ExpB b (SE) ExpB b (SE) ExpB
Age -.02 (.01) .98 -.02 (.01) .98 -.03 (.02) .97 —.03 (.02) .97
Whit; -.26 (.23) .77 .16 (.23) 1.18 -.02 (.25) .98 -.02 (.25) .98
GED“ .11 (.23) 1.12 .11(.22)1.12 .16 (.24) 1.17 .16 (.24) 1.17
HS Grad plusa -.18 (.32) .83 -.03 (.31) .97 -.15 (.33) .86 —.15 (.33) .86
Divorce/Widowb .15 (.31) 1.16 .17 (.31) 1.19 .23 (.32) 1.26 .22 (.33) 1.25
Mamedb -.20 (.38) .82 -22 (.37) .80 -.19 (.39) .83 -.18 (.39) .83
prior J Commite .49 (.26) 1.64 .51 (.26) .50 (.27) 1.66 .51 (.27) 1.66
1.67*
prior him“ .39 (.24) 1.48 .59 (.24) .56 (.25) .56 (.25)
l.81** 1.76* 1.76*
person Crimec .24 (.29) 1.28 .36 (.28) 1.44 .33 (.29) 1.39 .32 (.29) 1.38
property Crimec .67 (.29) .57 (.28) .63 (.30) .63 (.30)
196* 1.77* 1.88* 1.88*
Psafety Crimec .14 (.34) 1.15 .14 (.33) 1.15 .20 (.36) 1.23 .21 (.35) 1.23
Past rx Services .20 (.08) .23 (.07) .19 (.08) .19 (.08)
1.22** l.26** 1.21** 1.21*
Sup Intensity -.01 (.01) .00 (.004) 1.00 -.01 (.01) .99 -.01 (.01) .99
.99*
Grad of Txe -1.46 (.25) -.92 (.23) -123 (.26) -1.24 (.26)
23*“I .40*** 29*“: 29*“:
TX Group," .49 (.22) .75 (.21) .48 (.22) .48 (.23)
1.64* 2.11*** 1.61* 1.61*
Hospitalizede -.05 (.25) .95 .27 (.24) 1.31 -.03 (.25) .97 -.04 (.25) .96
MH Treatmente .10 (.32) 1.10 -.05 (.31) .96 -.08 (.33) .92 -.O8 (.33) .93
H M: -.59 (.29) —.57 (.28) -.50 (.30) .61 -.51 (.30) .60
.56* .57*
House Stability .34 (.06) .33 (.05) .32 (.07)
l.4l*** 140*!” 1.38***
Emp Stability -.07 (.02) -.07 (.02) -.07 (.20)
.93*** ,97*** .93***
House*l=.mpf .11 (.36) 1.12
Cox and Snell R2 .24 .20 .26 .26
Nagelkerke R2 .32 .26 .35 .35
-2LL 562.08 589.54 546.39 546.30
Model Chiz 139.02*** lll.56*** 154.70*** 154.80***
df l9 19 20 21

 

‘p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.
a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves (median split of sample) and who were employed for less than half

of their supervision term

137

The inclusion of housing and/or employment stability variables modiﬁed the

signiﬁcance of the coefﬁcient of some of the control variables that were previously

insigniﬁcant in the partial conditional control model”. Prior prison sentences were
positively associated with absconding in conditional models that controlled for
employment stability. Individuals who have served one or more prior prison terms were
nearly two times as likely as those serving their ﬁrst prison term to abscond. Supervision
intensity was negatively associated with absconding in the conditional model that
controlled for the effect of housing stability. This effect is very small since the odds ratio
approximates one, but did suggest that intensive supervision may reduce the likelihood of
absconding.

The interaction term of housing and employment instability did not inﬂuence the
likelihood of absconding. Additionally the interaction term did not inﬂuence the direct
effects of housing stability or employment stability on the likelihood of absconding.
These ﬁndings suggested that while housing stability and employment stability have
signiﬁcant direct effects on absconding, the combination of their effect does not
substantially increase or decrease the likelihood of absconding.

Summation of Results for Research Question 1: Treatment Dosage and Processes

The goal of this ﬁrst phase of analysis was to examine the relationships between
housing and employment stability on substance abuse treatment process indicators of
treatment dosage, treatment violations, and absconding. Overall, housing stability and

employment stability inﬂuenced all of the treatment process indicators to some degree.

 

38Age was statistically signiﬁcant in full conditional negative binomial regression models with robust
standard errors (see Appendix B for further information). In these models, older individuals were less
likely to abscond.

138

As the number of housing movements increase, the level of treatment dosage, the number
of treatment program violations, and the likelihood of absconding all increased. As the
stability of employment increased, the number of months in substance abuse treatment
increased and the number of treatment program violations and the likelihood of
absconding decreased. The joint effect of housing and employment instability is only
observed for the prediction of treatment program violations and suggested that the direct
effect of housing stability is conditioned by the effect of employment stability. This joint
effect was not observed for employment stability in linear regression models, but was
observed in negative binomial regression models with robust standard errors (see
Appendix B). While the measures of housing and employment stability inﬂuenced
treatment processes, the model ﬁt indices and explained variance statistics suggested that
housing and employment stability may have more of an effect on the treatment process
models of non-compliance (i.e., treatment violations and absconding) rather than models
of treatment dosage.

A number of the control variables also provided insights into how treatment
dosage and processes may be inﬂuenced. Not surprisingly, control variables associated
with treatment progress or supervision activity were important predictors of all of the
treatment processes. Individuals who successﬁrlly completed a substance abuse
treatment program received a larger dose of treatment, had less treatment program
violations, and were less likely to abscond relative to those who did not complete a
treatment program. There were consistent group effects across the conditional models.
Individuals assigned to the treatment group and participated in the intensive, reentry-

based substance abuse treatment program spent received a larger dose of treatment, but

139

also had more treatment program violations and an increased likelihood of abscond
relative to the control group who received traditional supervision services.

Some of the control variables were only associated with treatment dosage, while
others were only associated with treatment non-compliance. Supervision intensity was
consistently related to treatment dosage, with increased intensity of supervision leading to
a larger dose of substance abuse treatment. The effect of supervision intensity was not as
consistent with models assessing substance abuse treatment violation or absconding. In
the treatment violation models, supervision intensity appeared to be partially inﬂuenced
by the inclusion of employment stability, but not housing stability. In the absconding
models, supervision intensity was inﬂuenced by controls for housing stability, but not for
controls of employment stability.

Conviction offense type and prior correctional substance abuse treatment service
history were consistently related to treatment non-compliance and unrelated to treatment
dosage. Property offenders were more likely than drug offenders to have a higher
number of substance abuse treatment program violations. Individuals who have
experienced a higher number of past correctional substance abuse services were also
more likely to have a higher number of current substance abuse treatment violations and
were more likely to abscond. In the models of treatment program violations, those who
were convicted of crimes against persons had more violations, on average, than drug
offenders.

A number of the control variables were also inﬂuenced by the stepwise addition
or censoring of housing and employment stability variables across and within treatment

process models. Whites spent less time in treatment than nonwhites after housing and

140

employment stability were controlled, but had more treatment violations than nonwhites
after employment stability was controlled with or without subsequent controls for
housing stability. Enrollment in college or professional degree coursework was
positively associated with treatment dosage in the model that controlled for housing
stability and was negatively associated with absconding in models that only controlled for
housing or employment stability.

The remainder of the additive or censored effects pertained to speciﬁc treatment
process models. Individuals who were hospitalized at some point across their supervision
term received a larger dosage of substance abuse treatment. Individuals who participated
in mental health treatment had less substance abuse treatment program violations than
those who did not participate in mental health treatment once housing or employment
stability variables were included in the models. Individuals with a juvenile commitment
history were more likely to abscond, but only after employment stability was controlled.
Those with prior prison admissions were also more likely to abscond, but only after
employment stability was controlled with or without the inclusion of the measure of
housing stability.

Research Question 2: Unconditional, Partial Conditional, and Conditional Multivariate
Analyses of Relapse and Recidivism Outcome Measures

The goal of the second phase of analysis was to examine the relationships
between housing and employment stability and the outcome indicators of relapse, re-
arrest, and re-incarceration. Linear and logistic regressions were utilized to assess the
relationships. Models are presented stepwise, beginning with unconditional bivariate

models and ending with full conditional models that include the independent variables of

141

housing and employment, all of the control variables, and a check of the interaction term
of housing and employment instability.

Table 10 provides the unconditional and partial conditional linear regression
models of relapse on independent and control variables. Housing and employment
stability were signiﬁcantly related to relapse in the unconditional bivariate models as well
as the partial conditional multivariate model. Both of the effects were in the direction
expected by the logic of the second research question. As the number of housing
movements increased, the proportion of positive drug tests also increased. As
employment stability increased, the proportion of positive drug tests decreased.

The standardized coefﬁcients suggested that housing stability and employment
stability were relatively similar, but weak predictors of relapse. The conditional
multivariate model that includes both measures of stability explained only 8% of the
variance in the proportion of positive drug tests. At the low end, the unconditional
bivariate model of employment stability only explained 3% of the variability in relapse.

The conditional multivariate model of control variables also explained
approximately 8% of the variance of relapse. Similar to the estimated coefﬁcients for
housing and employment stability, the estimated coefﬁcients for the control variables
were weak. A number of the control variables were signiﬁcantly related to relapse.
Whites had less positive tests on average relative to nonwhites. The remaining

relationships were associated with supervision activity across the supervision term.

142

Table 10: Partial Linear Regressions of Substance Abuse Relapse on Independent and
Control Variables (n=51 1).

 

 

 

b (SE) StdB b (SE) StdB b (SE) StdB b (SE) StdB
Age -.00 (.002) .06
White“ -.07 (.03)
-.12**
G131)“l -.01 (.03) -02
HS Grad Plusa -.04 (.04) -.O6
Divorce/Widowedb .00 (.03) .00
Manned" -.02 (.04) -.03
Prior J Commite .00 (.03) .00
Prior Prison“ .05 (.03) .10
Person Crime“ -.03 (.03) -.06
Property Crime“ -.03 (.03) -.04
PSafety Crime“ -.05 (.04) -.07
Past TX Services -.00 (.01) -.00
Sup Intensity .00 (.00) .07
Grad of Tx“ -.06 (.03)
-.11*
Tx Groupd .05 (.02)
.10*
Hospitalized“ .07 (.03)
.12**
MH Treatment“ -.06 (.04) -.03
H Edu“ -.02 (.03)
-.09*
House Stability .02 (.01) .02 (.01)
.21*** _20***
Emp Stability —.01 (.002) -.01 (.002)
218*** ﬂ17***
R-Square .05 .03 .08 .08
Model F 23.88*** 17.32*** 20.61%“0511g 2.48***
df 1 1 2 18

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance.

Individuals who completed a substance abuse treatment program during their
supervision term had lower proportions of positive drug tests, on average, than those who

did not complete a treatment. Individuals assigned to the treatment group and

143

participated in intensive, reentry-based substance abuse treatment services had higher
proportions of positive drug tests than those who were assigned to the control group and
experienced traditional supervision services. Those who were hospitalized at least one
time during their supervision term had a higher proportion of positive drug tests, on
average, than those who were never admitted to the hospital. Finally, individuals who
were enrolled in college or professional degree coursework had lower proportions of
positive drug tests than those who did not seek to further their education.

Table 11 presents the full conditional linear regression models of relapse. In
general, the effects observed in the unconditional and partial models remain the same in
the full conditional models. Housing stability continued to be positively associated with

relapse after control variables, employment stability, and an interaction term of housing

and employment instability were controlled”. As housing movement increased the
proportion of positive drug tests also increased. Employment stability continues to be
negatively associated with relapse. While these direct effects continued to be observed,
there does not appear to be an interactive effect of housing and employment instability
that moderates levels of relapse.

According to the standardized coefﬁcients, the measures of housing and
employment stability were the strongest predictors of relapse. When each measure was
included in the full conditional models, housing stability was slightly stronger of a

predictor than employment stability. The strength of the models must be cautioned,

 

39 The results for the primary independent variables of housing and employment stability in the full
conditional models were conﬁrmed through negative binomial regression with robust standard errors. See
Appendix C for further information.

144

however, since the full conditional models only explained 10% to 13% of the variability
in relapse levels.

Table 11: Full Linear Regressions of Substance Abuse Relapse on Independent and
Control Variables (n=51 1).

 

 

 

Std B (SE) Std B (SE) Std B (SE) Std B (SE)
Age -.00 (.002) -.07 -.00 (.002) -.06 -.00 (.002) -.07 -.00 (.002) -.08
Whitee -.07 (.03) -.05 (.03) -.O8 -.06 (.03) -.06 (.03)
-.13** -.10* -.10*
GEDa -.01 (.03) -.02 -.01 (.02) -.01 -.00 (.03) -.01 -.01 (.03) -.01
HS Grad Plusa -.05 (.04) -.O7 -.04 (.04) -.05 -.04 (.04) -.06 -.05 (.04) -.06
Divorce/Widowedb .00 (.03) .Ol .01 (.03) .01 .01 (.03) .01 .01 (.03) .01
Marriedb -.02 (.04) -.02 -.02 (.04) -.02 -.02 (.04) -.02 -.01 (.04) -.02
Prior J Commite -.00 (.03) -.01 -.00 (.03) -.00 -.01 (.03) -.01 -.01 (.03) -.01
Prior Prisone .04 (.03) .09 .06 (.03) .05 (.03) .05 (.03)
.11* .10* .10*
Person Crimec -.04 (.03) -.O7 -.03 (.03) -.05 -.03 (.03) -.06 -.03 (.03) -.06
Property Crimec -.03 (.03) -.05 -.03 (.03) -.05 -.03 (.03) -.06 -.03 (.03) -.06
psafety Crimec -.05 (.04) -.07 -.05 (.04) -.O7 -.04 (.04) -.06 -.04 (.04) -.06
Past TX Services -.00 (.01) -.03 -.00 (.01) -.01 -.01 (.01) -.03 -.01 (.01) -.04*
Sup Intensity .00 (.001) .02 .00 (.000) .08 .00 (.001) .04 .00 (.001) .03
Grad ofo“ -.07 (.03) -.03 (.03) -.07 -.05 (.03) -.09 -.05 (.03)
-_13** -.09*
Tx Groupd .03 (.02) .06 .05 (.02) .03 (.02) .06 .03 (.02) .06
.10*
Hospitalized“ .06 (.03) .07 (.03) .06 (.03) .06 (.03)
.09* .13** .10* .09*
MH Treatmente -.03 (.04) -.04 -.04 (.03) -.05 -.04 (.04) -.05 -.05 (.04) -.05
H Educ -.05 (.03) -.08 -.05 (.03) -.08 -.05 (.03) -.07 -.04 (.03) -.07
House Stability .02 (.01) .02 (.01) .01 (.01)
' .21*** .19*** .14*
Emp Stability -.01 (.002) -.01 (.002) -.01 (.002)
-.14** -.13** -.10*
House*Empf .05 (.04) .08
R-Square .12 .10 .13 .13
Model F 339*" 2.87*** 3.61*** 352*"
df 19 19 20 21

 

"p<.05, **p<.01, "*p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

145

The control variable of hospitalization had the only consistent effect across the
full conditional models. Similar to the partial model that only included control variables,
individuals who have been hospitalized at least one time had a higher proportion of
positive drug tests than those who were never admitted to a hospital. The remaining
control variable effects observed in the previous partial model are inﬂuenced by the
stepwise inclusion or censoring of the independent variables of housing and employment
stability.

Whites continued to have lower proportions of positive drug tests relative to
nonwhites, but this relationship is not observed in the full conditional model that only
controlled for employment stability. This suggested that the differential effect of race
may be negated and subsumed by the effect of employment stability and ﬁrrther masked
once the partial effect of housing stability is included to the model. The effects of

activity during the supervision term diminished after controlling for housing and

employment stability“). Members of the treatment group continued to have higher
proportions of positive drug tests, but this effect was only observed in the model that only
controlled for employment stability. Once housing stability is controlled for the effect
becomes non-signiﬁcant. Finally, whether an individual was enrolled in college or

professional degree coursework became obsolete once housing or employment stability is

controlled.

 

40 Graduates of substance abuse treatment programs had signiﬁcantly lower proportions of positive tests on
average than non-graduates, but this effect appeared to be inﬂuenced by employment stability in full
conditional linear regression models. The negative effect of the successful completion of a substance abuse
treatment program on relapse did not appear to interact or be conditioned by employment stability in full
conditional negative binomial regression models with robust standard errors (see Appendix C for father
information). In these models, program graduates were observed to have lower levels of positive drug test
proportions than those who did not successfully complete a treatment program net all other variables.

146

Individuals who had previously incurred a prison sentence had higher proportions
of positive drug tests than those on their ﬁrst prison term for full conditional models that
included employment stability. The control variable of past prison sentences was

associated with relapse in the full conditional models, but was not observed in the partial

conditional model that only included control variables“. This suggested that
employment stability may interact and enhance the effects of criminal history record
information.

Table 12 provides the unconditional and partial conditional logistic regression
models of re-arrest on independent and control variables. It is immediately apparent that
housing stability had no effect on re-arrest. Employment stability was signiﬁcant and
negatively related to re-arrest in the unconditional model as well as the partial conditional
model that controlled for housing stability. This suggests that as employment stability
increased, the likelihood of re-arrest decreased. While signiﬁcant, employment stability
appeared to have a weak effect as indicated by the odds ratio approximating 1.00 and the
conservative pseudo variance explained statistic of 1%.

The partial conditional model that only includes the control variables suggests
that the control variables were more closely related to re-arrest relative to the independent
variables of housing and employment stability. The power of the conditional control
model was largely relative. Only 12% of the variability in re-arrest can be explained by

the control variables. At the outset it appears that the partial logistic regression models

 

41 Additionally, the number of prior exposures to correctional substance abuse treatment services was
observed to be negatively related to relapse after controlling for the direct and interactive effects of housing
and employment stability in linear regression models. The effect of the number of past exposures to
correctional substance abuse treatment on relapse was not observed in full conditional negative binomial
regression models with robust standard errors. See Appendix C for further information.

147

provide a cautious representation of the relationships between the independent and
control variables on re-arrest as well as the complexity of modeling re-arrest events.

Table 12: Partial Logistic Regressions of Re-Arrest on Independent and Control
Variables (n=511).

 

 

 

b (SE) ExpB b (SE) ExpB b (SE) ExpB b (SE) ExpB
.97*
White“ -.06 (.23) .94
GED“I .49 (.25)
1.64*
Hs Grad Plus“ .79 (.32)
2.21**
Divorce/Widowedb .31 (.31) 1.36
Married" .09 (.38) 1.10
Prior J Commit“ -. 14 (.27) .87
Prior Prison“ .39 (.24) 1.47
Person Crime“ -.01 (.29) .99
Property Crime“ .26 (.29) 1.30
PSafety Crime“ .34 (.36) .71
Past TX Services -.07 (.07) .94
Sup Intensity -.02 (.01)
.98***
Grad of TXe -.56 (.25)
.57*
Tx Group“ .31 (.22) 1.37
Hospitalized“ .07 (.27) 1.07
MH Treatment“ -.29 (.36) .75
H Edu“ -.31 (.30) .73
House Stability -.02 (.04) .98 -.02 (.04) .98
Emp Stability -.04 (.02) -.04 (.02)
.96** .96**
Cox and Snell R2 .001 .01 .01 .12
Nagelkerke R2 .001 .02 .02 .17
-2LL 609.33 603.15 602.83 546.49
Model ChiZ .26 6,44'” 676* 63.10am”:
df 1 l 2 18

 

*p<.05, **p<.01, ***p<.OOl; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance.

148

Control variables related to the likelihood of re-arrest included demographic
variables and variables associated with supervision activities across the supervision term.
Individual age was negatively related to the likelihood of re-arrest, with older individuals
being less likely to be re-arrested. The effect of individual age was signiﬁcant, but
relatively weak with an odds ratio approaching one. The strongest effects for the control
variables in the conditional control model were for educational status. The effects were
in an unexpected direction with stronger educational backgrounds leading to higher
probabilities of re-arrest. Individuals who completed a GED were just under two times
more likely to be re-arrested than those who did not complete high school or obtain a
GED. Those who have received a high school diploma (not GED based) and who may
have continued their education through college or professional degree coursework were
just over two times more likely to be re-arrested than those who did not complete high
school or obtain a GED.

Relative to supervision activities across supervision terms, the intensity of
supervision an individual received was negatively associated with re-arrest. Individuals
who were intensely supervised were less likely to be re-arrested. This effect is
signiﬁcant, but relatively weak with an odds ratio of .98. The successful completion of a
substance abuse treatment program was also associated with re-arrest. Program graduates
were less likely to be re-arrested relative to those who did not complete a treatment
program.

Table 13 provides the full conditional logistic regression models of re-arrest. In
large part, the full conditional models closely resemble the previous unconditional and

partial conditional models. Approximately 12% of the variability in the likelihood of re-

149

arrest can be explained by the full conditional models. The effects of the control
variables of age, educational status, and supervision intensity are consistent across the
full conditional models in terms of direction and strength. Once again, older offenders,
those who have not completed high school or GED equivalency, and those who were
subjected to relatively intense supervision were less likely to be re-arrested after
controlling for all of the independent and control variables.

The most notable difference between the full conditional and unconditional or
partial conditional models is the effect of employment stability“. After controlling for

all of the control variables, the effect of employment stability was parceled out and no

longer associated with re-arrest43. The relative effect of employment stability maintains

direction and magnitude, but was no longer signiﬁcant.

 

42 The other notable change between partial logistic regression models and the full conditional logistic
regression models was for graduates of substance abuse treatment program. Those who successfully
completed a treatment program were less likely to be re-arrested than those who do not complete a
treatment program, but this effect was not observed for the conditional model that only controls for
employment stability. The insigniﬁcance of the effect remains despite similarity in magnitude and
direction of the partial effect. Caution is needed with interpreting the effect of graduates of substance
abuse treatment programs on re-arrest. The signiﬁcance of the effect was only observed in full conditional
logistic regression models. The signiﬁcance of the coefﬁcient was not replicated in full conditional
negative binomial regression models with robust standard errors (see Appendix C).

43 The results for the primary independent variables of housing and employment stability in the full
conditional models were conﬁrmed through negative binomial regression with robust standard errors. See
Appendix C for further information.

150

Table 13: Full Logistic Regressions of Re-Arrest on Independent and Control Variables

 

 

 

(n=511).
b (SE) ExpB b (SE) ExpB b (SE) ExpB b (SE) ExpB

Age -.03 (.02) -.03 (.02) -.03 (.02) -.03 (.02)

.97* . .97* .97* .97*
Whitee -.08 (.23) .92 .02 (.24) 1.02 -.01 (.24) .99 -.01 (.24) .99
(3131)“1 .50 (.25) .51 (.25) .52 (.25) .52 (.25)

1.66* 1.66* 1.68* 1.68*
HS Grad Plusa .77 (.32) .81 (.32) .79 (.32) .79 (.32)

2.17* 2.26" 2.20" 2.20“
Divorce/Widowedb .32 (.31) 1.38 .33 (.31) 1.39 .34 (.31) 1.41 .34 (.31) 1.40
Marriedb .10 (.38) 1.11 .10(.38)1.11 .11(.38)1.12 .11(.38)1.12
Prior Juv Commite -.15 (.27) .86 -.14 (.28) .87 —.15 (.28) .86 -.15 (.28) .86
Prior Prison Admite .37 (.25) 1.45 .43 (.25) 1.54 .41 (.25) 1.51 .41 (.25) 1.51
Person Crimec -.02 (.29) .98 .01 (.29) 1.01 -.00 (.30) 1.00 -.00 (.30) 1.00
Property Crimec .25 (.29) 1.29 .24 (.29) 1.28 .24 (.29) 1.27 .24 (.29) 1.27
PSafety Crimec -.32 (.36) .73 -.32 (.36) .73 -.30 (.36) .74 -.30 (.36) .74
Past TX Services -.08 (.07) .92 -.O7 (.07) .93 -.O9 (.07) .92 -.O9 (.07) .92
Sup Intensity -.02 (.01) -.02 (.01) -.02 (.01) -.02 (.01)

-98*** .98*** .98*** .98***
Grad ofo‘“ -.61 (.25) -.45 (.26) .64 -.51 (.26) -.51 (.26)

.54* .60* .60*

Tx Groupd .23 (.23) 1.26 .30 (.22) 1.35 .23 (.23) 1.26 .23 (.23) 1.26
Hospitalized“ -.01 (.27) .99 .06 (.27) 1.06 -.01 (.27) .99 -.0'1 (.27) .99
MH Treatment“ -.31 (.36) .74 -.34 (.36) .71 -.35 (.36) .71 -.35 (.36) .71
H Educ -.29 (.31) .75 -.28 (.31) .76 -.26 (.31) .77 -.26 (.31) .77
House Stability .09 (.05) 1.09 .08 (.05) 1.09 .08 (.07) 1.08
Emp Stabili -.02 (.02) .98 -.02 (.02) .98 -.02 (.02) .98
House*Emp .03 (.37) 1.03
Cox and SnellR2 .12 .12 .12 .12
Nagelkerke R2 .17 .17 .18 .18
-2LL 543.53 544.71 542.19 542.19
Model ChiZ 66.06*** 64.88*** 67.40*** 67.40***
df 19 19 20 21

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.
a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

Table 14 presents the unconditional and partial conditional logistic regression

models of re-incarceration on independent and control variables. Housing and

employment stability were signiﬁcantly and negatively related to re-incarceration across

151

the models. As housing movement increased (or becomes relatively more unstable), the
likelihood of re-incarceration decreased. This effect is unexpected. The second research
question presumed that unstable housing would increase the likelihood of deleterious
outcome effects such as re-incarceration. The effect of employment stability is consistent
with the logic of the second research question. As employment stability increases, the
likelihood of re-incarceration also decreases.

Overall, the measure of employment stability appeared to be a stronger predictor
of re-incarceration than the measure of housing stability. Employment stability explained
11% of the variability of the likelihood of re-incarceration, while housing stability only
explained 4% of the variability of re-incarceration likelihood. Employment stability also
provides a larger reduction to the log-likelihood ratio, which corresponded to an
improved model ﬁt. The partial effects of the independent variables were largely
overshadowed by the effects of the control variables. As indicated by the pseudo
explained variance statistics and log-likelihood ratio statistic, the control variables
provide a better model from which to assess relationships with re-incarceration.

A number of the control variables were associated with re-incarceration. Past
educational status was one of the strongest predictors of re-incarceration. Individuals
who obtained a GED were just over two times more likely to be re-incarcerated relative
to those who did not obtain a high school degree or a GED. Those who obtained a high
school degree through traditional means and who may have enrolled in college or
professional degree coursework were just over three times more likely to be re-

incarcerated compared to those who did not obtain a high school degree or GED.

152

Table 14: Partial Logistic Regressions of Re-Incarceration on Independent and Control
Variables (n=511).

 

b (SE) ExpB b (SE) ExpB b (SE) ExpB b (SE) ExpB

 

 

Age -.03 (.02) .97
White“ -.41 (.27) .67
GED“ .82 (.28)
2.28**
HS Grad Plus“ 1.12 (.36)
3.06**
Divorce/Widowedb .21 (.35) 1.23
Marriedb .19 (.43) 1.21
Prior J Commit“ .58 (.31) 1.78
Prior Prisone .22 (.27) 1.24
Person Crime“ .57 (.33) 1.76
Property Crime“ .18 (.34) 1.20
PSafety Crime“ -.07 (.39) .93
Past TX Services -.01 (.08) .99
Sup Intensity -.05 (.01)
.951"!!!
Grad orTx“ -1.91 (.31)
.15***
Tx Groupd .60 (.25)
181*
Hospitalized“ -.37 (.31) .69
MH Treatment“ .58 (.39) 1.79
H Edu“ -.37 (.33) .69
House Stability -20 (.05) -.22 (.05)
.82*** .81***
Emp Stability -.13 (.02) -.13 (.02)
.88*** .88***
Cox and Snell R2 .04 .11 .15 .33
Nagelkerke R2 .06 .16 .21 .46
-2LL 629.49 588.37 566.17 446.56
Model Chi2 20.57*** 61 .69** 83.89*** 203.50***
df 1 l 2 l8

 

‘p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance.

The remaining control variables associated with re-incarceration consisted of
measures of supervision activity across the supervision term. On average, those who

were subject to more intense supervision were less likely to be re-incarcerated. Similarly,

153

individuals who successfully completed a substance abuse treatment program were less
likely to be re-incarcerated relative to those who did not complete a treatment program.
Finally, treatment group assignment was positively related to re-incarceration. Those
who were assigned to the treatment condition and participated in intensive, reentry-based
substance abuse treatment were nearly two times as likely to be re-incarcerated relative to
the control group who received traditional supervision services.

Table 15 provides the full conditional logistic regression models of re-
incarceration. The full conditional models provide little additional explanatory power
from the partial conditional model that included all the control variables. The pseudo
variance explained statistic only increased by 4% in the full conditional model that
includes all independent and control variables. The perpetual reduction in the log-
likelihood ratio statistic suggests that the inclusion of additional variables provided a
stronger model for the assessment of re-incarceration relationships.

Housing stability is no longer signiﬁcantly related to re-incarceration aﬁer the
control variables are entered into the conditional models. Employment stability, on the
other hand, continued to be negatively related to re-incarceration after the control
variables were entered into the model and the interaction term is controlled. The effect of

employment stability was also a direct effect and was not conditioned by the interaction

of elevated levels of housing and employment instability“.

 

44 The results for the primary independent variables of housing and employment stability in the full
conditional models were conﬁrmed through negative binomial regression with robust standard errors. See
Appendix C for further information.

154

Table 15: Full Logistic Regressions of Re-Incarceration on Independent and Control

Variables (n=511).

 

 

 

b (SE) ExpB b (SE) ExpB b (SE) ExpB b (SE) ExpB
Age -03 (.02) .97 -.03 (.02) -.03 (.02) -.03 (.02)
97* 97* 97*
Whit; -39 (.27) .68 -.04 (.29) .96 .02 (.29) 1.02 .02 (.29) 1.03
. may .81 (.28) .90 (.29) .89 (.29) .88 (.29)
2.25** 2.46** 2.43** 2.41**
Hs Grad plusa 1.14 (.36) 1.23 (.38) 1.29 (.38) 1.28 (.39)
3.13** 3.42*** 3.65*** 3.58**
Divorce/Widowedb .19 (.35) 1.21 .30 (.37) 1.35 .28 (.37) 1.32 .30 (.37) 1.35
Mamedb .18 (.43) 1.19 .30 (.45) 1.35 .30 (.45) 1.35 .26 (.45) 1.30
prior Juv Commie .60 (.32) 1.83 .63 (.32) .68 (.33) .67 (.33)
188* 198* 195*
prior prison Admtte .22 (.27) 1.25 .46 (.29) 1.58 .49 (.29) 1.63 .47 (.29) 1.61
198* 2.05* 2.13*
propeity Crimec .20 (.34) 1.22 .10 (.35) 1.11 .12 (.35) 1.13 .14 (.35) 1.15
Psafety Crimec —.09 (.39) .92 .03 (.41) 1.03 .01 (.41) 1.01 -.01 (.41) .99
Past Tx Services .00 (.08) 1.00 -.04 (.08) .96 -.02 (.08) .98 -.01 (.08) .99
Sup Intensity -.04 (.01) -.04 (.01) -.04 (.01) -.04 (.01)
-96*** .96*** .96*** .96***
Grad ofo6 -1.88 (.31) -1.48 (.32) -1.40 (.32) -1.39 (.32)
.15*** .23*** .25*** .25***
TX Groupd .67 (.26) .53 (.26) .62 (.27) .61 (.27)
195** 1.69* 187* 184*
Hosptmiizede -.32 (.32) .73 -.42 (.32) .66 -.33 (.33) .72 -32 (.33) .72
MH Treatmente .59 (.39) 1.80 .37 (.40) 1.45 .38 (.40) 1.47 .36 (.40) 1.43
H Educ -.40 (.33) .67 -22 (.35) .80 -.25 (.35) .78 -.24 (.36) .78
House Stability -.08 (.06) .93 -.12 (.07) .89 -.03 (.09) .97
Emp Stability -.11 (.02) —.12 (.02) -.13 (.03)
.89*** .89*** .88***
House*Empf -.58 (.44) .56
Cox and Snell R2 .33 .37 .37 .37
Nagelkerke R2 .46 .51 .5 1 .52
-2LL 444.99 417.95 414.37 412.62
Model Chiz 205.07*** 232.11*** 235.69*** 237.43***
df 19 19 20 21

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.
a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non—instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term

155

All of the control variables associated with re-incarceration in the partial
conditional model that only included control variables continued to be related to re-
incarceration after controls for housing and employment stability with or without the
interaction term. All of the effects continued to be in the same direction and relatively
the same magnitude. Educational status affects re-incarceration, with those with more of
an educational background being more likely to be re-incarcerated than those with less of
an educational background. Negative associations with re-incarceration continued to be
observed for those who have received relatively intense supervision and who successfully
completed a substance abuse treatment program during their supervision term. Once
again, those who were assigned to the treatment group and who received intensive
reentry-based substance abuse treatment programming were more likely to be re-
incarcerated relative to individuals who experienced traditional supervision services.

A number of control variables were associated with re-incarceration after the

inclusion of the measure of employment stability and persisted once the partial effect of

housing stability was controlled“. Older individuals were less likely to be re-
incarcerated. Those who were convicted of a crime against persons were approximately
two times more likely to be re-incarcerated relative to those who were convicted on drug
charges. In all, the partial effects of age and conviction offense type seem to be

conditioned and enhanced by the partial effect of employment stability.

 

45 Individuals who had previously been committed to a juvenile institution were nearly two times more
likely to be re-incarcerated relative to those who never experienced a juvenile commitment in the full
conditional logistic regression analyses. This ﬁnding was not observed in replicated models that used
negative binomial regression with robust standard errors (see Appendix C). Instead, the negative binomial
model suggests that individuals who have previously been incarcerated were 1.27 times more likely to be
incarcerated during their current supervision term relative to those who just completed their ﬁrst
institutional sentence. This effect was only apparent after controlling for the interaction of housing and
employment instability.

156

Summation of Results for Research Question 2: Relapse and Recidivism Outcome
Measures

The goal of the second phase of analysis was to examine the relationships
between housing and employment stability and the outcome indicators of relapse, re-
arrest events, and re-incarceration instances. Housing and employment stability had
mixed effects on the outcome indicators. Housing stability was only related to relapse
and suggested that as housing movement increased, the ratio of positive drug tests also
increased. As the stability of employment increased, the ratio of positive drug tests and
the likelihood of re-incarceration decreased. Employment stability was unrelated to re-
arrest after all of the control variables were entered into the full conditional models. The
joint effect of housing and employment instability was not observed for any of the full
conditional models.

Given these mixed effects, it is not surprising to observe that the model ﬁt indices
and explained variance statistics suggest that housing and employment stability may have
more of an effect on the outcome models of relapse rather than models of re-arrest and re-
incarceration. For models assessing relationships with relapse, housing and employment
stability appear to contribute to similar levels of explained variance as the control
variables. This contribution needs to be prefaced. The estimated partial coefﬁcients in
relapse models were all relatively small. Only 10% of the variability in relapse can be
explained by a conditional model that includes housing and employment stability, all of
the control variables, and an interaction term.

In terms of the recidivism measures, only 10% of the variability in the likelihood

of re-arrest can be explained in a full conditional model, with much of the explanatory

157

power being generated from the control variables. Similarly, much of the modest
explained variance (approximately 30% to 40%) associated with the likelihood of re-
incarceration is due, in large part, to the control variables with partial inﬂuence from the
negative effect of employment stability.

The control variables shaped some of the outcome variables, but these effects
were not consistently found across all three outcome variable indicators. Instead, a few
consistent results were observed across the indicators of recidivism (i.e., re-arrest or re-
incarceration) and the remaining results were observed within speciﬁc outcome variable
indicators. Educational background status was positively associated with re-arrest and re-
incarceration, with those possessing a GED or traditional high school diploma (with or
without enrollment in college or professional degree coursework) being more likely to
recidivate relative to those lacking a GED or high school diploma. Supervision intensity
was negatively associated with recidivism. Individuals who were exposed to relatively
higher levels of supervision intensity were less likely to be re-arrested or re-incarcerated.

Within the outcome indicator of relapse, direct effects for race and activities
during the supervision term were observed. On average, whites had lower proportions of
positive drug tests than non-whites. Individuals who were admitted to the hospital were
observed to have higher proportions of positive drug tests relative to those who were
never admitted to the hospital during their supervision term.

Direct effects on recidivism measures were contingent upon the type of
recidivism measure. The direct effect of age was only observed for the re-arrest outcome
indicator and suggested that older offenders were less likely to be re-arrested during their

supervision term. The negative effects of age on re-incarceration likelihood were

158

observed, but only after controls for employment stability were introduced into the
regression models. The control variable effects on re-incarceration were based upon
substance abuse treatment activity across the supervision term. Individuals who

successfully completed a treatment program were less likely to be re-incarcerated relative

to those who did not complete treatment“. A group effect on re-incarceration was
observed. Those who were assigned to the treatment group and participated in intensive,
reentry-based substance abuse treatment services were more likely to be re-incarceration
than those under traditional supervision services.

Control variables were also indirectly inﬂuenced by the stepwise addition or
censoring of housing and employment stability variables within outcome indicator
models. Regarding relapse, those with prior prison admissions had higher proportions of
positive drug tests than those on their ﬁrst prison term in models that included a measure
of employment stability. The positive effect of treatment group assignment and
associated intensive treatment programming was negated once housing stability was
controlled. Finally, the negative relationship between enrollment in college or
professional degree coursework and positive drug test proportions was controlled once
housing or employment stability measures are included into the conditional models.

The censoring or additive effect of independent and control variables was also
observed in models assessing re-incarceration likelihood. Age and offense type were
associated with re-incarceration once a measure of employment stability was entered into
a conditional model. Older individuals were less likely to be re-incarcerated after

employment stability was controlled. Individuals who were serving a supervision term

 

4 . . . . . .
6 Graduates were also observed to have lower levels of relapse 1n full conditional negative bmorrual

159

for a crime against persons were more likely to be re-incarcerated after controls for
employment stability.

Research Question 3: Conditional Multivariate Analyses of Relapse and Recidivism
Outcome Measures Controlling for Treatment Dosage and Processes

The goal of the ﬁnal phase of analysis was to examine the relationship between
housing and employment stability on the outcome measures of relapse, re-arrest
incidents, and re-incarceration events after controlling for substance abuse treatment
indicators of treatment dosage and processes. Multivariate modeling in this manner will
assist in the determination of the relative contributions of housing and employment
stability after subjecting the measures to the conditional effects of treatment processes.
Linear and logistic regressions will be used to assess the relationships. Models are
presented stepwise. An initial full conditional model that includes measures of housing
and employment stability, treatment process indicators, and all the control variables is
followed by a secondary model that adds an interaction term that represents elevated
housing and employment instability.

Table 16 presents the full conditional linear regression models of relapse on
independent, treatment processes, and control variables. Overall the model explains 18%
of the variability in the levels of relapse. The standardized coefﬁcients suggest that the
strongest effects consist of treatment process measures of dosage and programmatic
violations. Of the two independent variables, only employment stability was

signiﬁcantly related to the proportion of positive tests after controlling for treatment

 

regression models with robust standard errors (see Appendix C for further information).

160

processes“. Once again, the estimated coefﬁcients in the relapse models are observed to
be relatively weak, which was consistent with previous models that did not control for
treatment dosage or processes. The insigniﬁcance of the interaction term suggests that
the direct effect of employment stability was not conditioned by the effects of housing
stability.

Two of the three treatment process variables were related to relapse after
controlling for the effect of housing and employment stability“. As the dose of
substance abuse treatment increased, the proportion of positive drug tests increased. The
number of treatment program violations was also positively related to relapse and
suggests that as the number of violations increased, the proportion of positive drug tests
also increased. These direct effects also provide additional insights to determinations of
the direct and indirect effects of housing and employment stability.

The direct effect of housing stability was unrelated to positive drug test
proportions after treatment process measures are controlled. Models that did not control
for the effect of treatment process measures suggested that the direct effect of housing
stability was positively associated with proportions of positive drug tests. Moreover,
previous models suggested that housing stability was positively associated with each of

the individual indicators of treatment processes.

 

4 . . . . .

7 The results for the primary independent variables of housrng and employment stability 111 the full
conditional models were conﬁrmed through negative binomial regression with robust standard errors. See
Appendix D for further information.

4 . . . .
8 The results for the treatment process varlables 1n the full condlnonal models were conﬁrmed through
negative binomial regression with robust standard errors. See Appendix D for ﬁrrther information.

161

Table 16: Full Linear Regressions of Substance Abuse Relapse on Independent,

Treatment Processes, and Control Variables (n=511).

 

 

 

b (SE) Std B b (SE) Std B
Age -.00 (.001) -.08 -.00 (.001) -.08
White“ —.05 (.03) -.10* -.05 (.03) -.10*
(35153 .00 (.02) .00 .00 (.02) .00
HS Grad plusa -.04 (.03) -05 -.04 (.03) -.05
Divorce/Widowedb '-00 (-O3) '-00 -.00 (.03) -.00
Manied“ —.01 (.04) -.01 -.01 (.04) -.01
Prior J Commit“ -.01 (.03) -.01 -.01 (.03) -.01
Prior Prisone -06 (07—) -11* .06 (.02) .11*
Person Crimec '-04 (-O3) '08 -.04 (.03) -.08
Property CI‘ImCC -.05 (.03) -.08 -.O4 (.03) -.O8
PSafety Crime“ -.04 (.04) -.06 -.04 (.04) -.06
Past TX Services -.01 (.01) -.O6 -.01 (.Ol) -.06
Sup Intensity -.00 (.001) -.05 —.00 (.001) -.05
Grad ofoe -.O6 (.03) -.12* -.07 (.03) -.12*
Tx Group“ -.03 (.03) -.07 -03 (.03) -.06
Hospitalized“ .05 (.03) .08 .05 (.03) .08
MH Treatment“ -.02 (.03) -.03 -.02 (.03) -.03
H Educ -.05 (.03) -.07 -.05 (.03) -.07
House Stability .00 (.01) .04 .00 (.01) .02
Emp Stabili -.01 (.002) -.12** -.01 (.002) -.12*
House*Emp .01 (.04) .03
Dose of SATX .01 (.003) .26*** .01 (.003) .26***
SATX Program Violations .04 (.01) .20*** .04 (.01) 20*”
Absconde .02 (.03) .04 .02 (.03) .04
R-Square .18 .18
Model F 474*“: 4544"”
df 23 24

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in

parentheses and followed by a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

In all, these ﬁndings suggest that the positive effects of housing stability on the

proportion of positive drug tests are completely mediated by the effects of treatment

162

process variables of treatment dosage and treatment program violations”. As the number
of housing movements increased, the amount of substance abuse treatment dosage also
increased, which contributes to relapse with an increase in the proportionality of positive
drug tests. Additionally, as the number of housing movements increased, the number of
treatment program violations also increased. In turn, the number of program violations
inﬂuenced relapse with more violations leading to higher drug test proportions.

The direct effect of employment stability was negatively associated with relapse
after controlling for treatment process variables. The effect was observed for both
models and suggested that the direct effect was not conditioned by the interactive effects
of housing and employment instability. As the stability of employment increased, the
proportionality of positive drug tests to total drug tests decreased. The direct negative
effect of employment stability on relapse was also observed in past conditional models
that did not control for treatment dosage or processes.

In previous models, employment stability was positively associated with
substance abuse treatment dosage levels and negatively associated with the number of
treatment program violations and the likelihood of absconding. Given the signiﬁcance of
the relationships between employment stability and the indicators of treatment processes
and the relapse outcome measure, it appears that employment stability indirectly affected
relapse through treatment dosage and treatment program violations. Higher levels of
employment stability were associated with higher treatment dosage levels. In turn, higher

dosage levels were associated with relapse and higher proportions of positive drug tests.

 

49Methods used for the determination of mediation or indirect effects were based upon Baron and Kenny
(1986)

163

Stable employment led to fewer number of treatment program violations, which was
associated with relapse and lower proportions of positive drug tests.

Control variables continue to inﬂuence observed relationships with relapse after
controlling for treatment processes. On average, whites had lower proportions of positive
drug tests relative to non-whites. Individuals who have previously served a prison
sentence had a higher proportion of positive tests compared to those who are under
supervision for their ﬁrst prison sentence. The completion of a substance abuse treatment
program during the supervision term continued to affect relapse outcomes. Individuals
who have successfully completed treatment had lower proportions of positive drug tests
than those who did not complete a treatment program. The remainder of the control
variables that inﬂuenced relapse in past models — history of prison admission, past
obtainment of correctional substance abuse treatment services, treatment or control group

assignment, and the pursuit of higher education during the supervision term — were no

longer related to relapse after controlling for treatment processes50

Table 17 provides the conditional logistic regression models of re-arrest on
independent, treatment processes, and control variables. The model explained 18% of the
variability in the likelihood of re-arrest. As suggested by the minute reduction in log-
likelihood ratios, there does not appear to be much of a nominal difference between the
two models. Once again, the strongest direct effects appeared to be driven by treatment

dosage and process variables.

 

50 Hospital admission during the current supervision term was observed to be unrelated to relapse in the
full conditional linear regression models controlling for the partial effects of treatment processes. This lack
of association was not observed in replicated models that utilized negative binomial regression with robust
standard errors (see Appendix D). Instead, individuals who were hospitalized during their supervision term
were observed to have a rate of relapse that was 1.21 times higher than those who were not admitted to a
hospital.

164

Only housing stability appeared to be related to the likelihood of re-arrest after

controlling for treatment dosage and process variables“. However, the effect of housing
stability was reduced to insigniﬁcance once the joint contribution of housing and
employment instability was controlled in the full logistic regression models. When
negative binomial regression models with robust standard errors were used the effect of
housing stability on re—arrest likelihood was observed with or without controls for the
interaction between housing and employment stability (sees Appendix D). The
interaction term did not have a direct association with the likelihood of re-ar'rest in
logistic or negative binomial models, but did appear to inﬂuence housing stability in
logistic models.

All of the treatment process variables were directly related to re-arrest likelihood

after controls for the partial effects of housing and employment stability as well as the

joint contribution of the two measures of stabilitysz. Substance abuse treatment dose was
negatively related to re-arrest. On average, individuals who received a larger dose of
treatment were less likely to be re-arrested. Unexpectedly, the number of treatment
program violations was negatively related to re-arrest. Individuals with more treatment
program violations were less likely to be re-arrested. Finally, individuals who absconded
at least one time during their supervision term were over two times more likely to be re-

arrested than those who remained active throughout their supervision term.

 

51 The results for the primary independent variables of housing and employment stability in the full
conditional models were conﬁrmed through negative binomial regression with robust standard errors. See
Appendix D for further information.

52 The results for the treatment process variables in the full conditional models were conﬁrmed through
negative binomial regression with robust standard errors. See Appendix D for further information.

165

Table 17: Full Logistic Regressions of Re-Arrest on Independent, Treatment Processes,
and Control Variables (n=511).

 

 

 

b (SE) ExpB b (SE) ExpB
Age -.03 (.02) .97 -.03 (.02) .97
White} -03 (.26) .97 -.02 (.26) .98
GED“ .47 (.26) 1.60 .47 (.26) 1.60
HS Grad plusa .78 (.33) 2.19* .78 (.33) 2.19*
Divorce/Widowedb .42 (.32) 1.52 .41 (.32) 1.50
Mat-tied" .13 (.40) 1.14 .14 (.40) 1.15
Prior J Commie -.21 (.28) .81 -.20 (.29) .82
Prior Prison“ .33 (.26) 1.39 .33 (.26) 1.40
Person Crimec -04 (31) 1-04 .04 (.31) 1.04
property Crime“ .25 (.31) 1.28 .25 (.31) 1.29
PSafety Crime“ -.37 (.37) .69 -.35 (.37) .70
Past TX Services -.07 (.08) .94 -.O7 (.08) .93
Sup Intensity -.02 (.01) 98** —.02 (.01) .98**
Grad ofo6 -.18 (.31) .84 -.20 (.31) .82
TX Group“ .66 (.27) 193* ..68 (.27) 1.97**
Hospitalizede .01 (.29) 1.01 .01(.29)1.01
MH Treatment“ --54 (.38) .58 -.53 (.38) .59
H Educ -21 (.32) .81 -.21 (.32) .81
House Stability .17 (.07) 1.19** .25 (.08) 1.16
Emp Stabili -.01 (.02) .99 -.01 (.02) .99
House*Emp .25 (.39) 1.29
Dose of SATX -.09 (.03) .92** —.09 (.03) .92**
SATX Program Violations -.47 (.13) .63*** -.48 (.13) .62***
Absconde .88 (.26) 2.41*** .88 (.26) 2.40“**
Cox and Snell R2 .18 .18
Nagelkerke R2 .26 .26
-2LL 508.68 508.26
Model ChiZ 100.9l*** 10133:”:-
df 23 24

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

The direct effect of housing stability was positively associated with the likelihood
of re-arrest after controlling for the effects of treatment process variables. It appeared

that as the number of housing movements increased, the likelihood of re-arrest also

166

increased. However, caution is needed with interpreting the effects of housing stability.
Once the interaction term was introduced, the effect of housing stability maintains a
similar direction and magnitude of effect but was no longer signiﬁcantly related to re-
arrest.

Previous models used to assess the relationship between housing stability and re-
arrest did not control for the partial effects of treatment processes. These models
indicated that housing stability was not directly related to re-arrest likelihood. Once
treatment processes were controlled, however, the direct effect of housing stability on re-
arrest was observed. These ﬁndings suggest that the direct effect of housing stability is
enhanced by the inclusion of the partial effects of treatment process variables. However,
there was no evidence to suggest that the relationship between housing stability and re-
arrest likelihood was mediated or indirectly affected by the treatment process variables.

Employment stability was unrelated to re-arrest likelihood after controlling for
treatment processes. Previous models that did not control for the effects of treatment
processes also suggested that employment stability did not have a direct effect on re-
arrest. Employment stability was found to be positively associated with substance abuse
treatment dosage and negatively associated with the number of treatment violations, and
the likelihood of absconding in previous models. The current and previous models
suggest that employment stability only provided direct effects on treatment processes and
did not indirectly affect re-arrest likelihood through treatment processes.

After controlling for the partial effects of the independent and treatment processes
variables, a number of control variables were related to re—arrest likelihood. The effects

of two control variables were consistent with previous models that assessed the

167

relationships between housing stability, employment stability, and re-arrest. Educational
background status was positively associated with re-arrest. Individuals who received a
traditional high school diploma and who may have pursued advanced degrees were just
over two times more likely to be re-arrested relative to those who did not obtain a high
school diploma or GED. The differential effects between those who obtained a GED and
those who did not obtain a diploma or GED were not observed once treatment processes
were controlled. Supervision intensity continued to be a relevant factor in re-arrest
likelihood. Individuals who received relatively intense supervision were less likely to be
re-arrested, while those who received low levels of supervision were more likely to be re-
arrested.

A few notable control variable differences are observed between previous models
that assessed the relationships between housing stability, employment stability, and re-
arrest and the current models of re-arrest likelihood that control for the effects of housing
stability, employment stability, and treatment process measures. Age and whether an
individual had successﬁrlly completed a substance abuse treatment program were no
longer associated with re-arrest once the partial effects of treatment processes were
considered. Group assignment was observed to differentiate the likelihood of re-arrest
once treatment processes were controlled. This effect was not observed in previous
models. Net housing stability, employment stability, and treatment process
considerations, this group effect suggests that individuals assigned to the treatment
condition and who received intensive, reentry-based substance abuse treatment were
nearly two times as likely to be re-arrested relative to those who received traditional

supervision services.

168

Table 18 presents the conditional logistic regression models of re-incarceration on
independent, treatment processes, and control variables. Overall the models explained
41% of the variability in the likelihood of re-incarceration. The log-likelihood ratio
levels appear to be similar, which suggests that one model did not provide a better ﬁt
over the other. The standardized coefﬁcients suggest that strong effects were observed
for treatment processes (speciﬁcally absconding) and the control variables that represent
educational background and activities across the supervision term.

Employment stability appeared to be the only primary independent variable

related to the likelihood of re-incarceration after controls for the treatment process

variables53. Housing stability was related to re-incarceration likelihood in the full
conditional logistic regression model that did not control for the interaction of housing
and employment instability. Once the interaction was controlled, the association between
housing stability and re-incarceration was reduced to insigniﬁcance. Models using
negative binomial regression with robust standard errors did not replicate the ﬁndings for
housing stability (see Appendix D). These models suggest that housing stability was not

related to re-incarceration.

 

53The results for the primary independent variable of employment stability in the ﬁrll conditional models
were conﬁrmed through negative binomial regression with robust standard errors. See Appendix D for
further information.

169

Table 18: Full Logistic Regressions of Re-Incarceration on Independent, Treatment
Processes, and Control Variables (n=511).

 

 

 

b (SE) ExpB b (SE) ExpB
Age -.03 (.02) .97 -.03 (.02) .97
Whitee -.02 (.31) .98 -.02 (.31) .98
GEDa .94 (.31) 2.56" .94 (.31) 2.55”
HS Grad Plusa 1.41 (.41) 4.09*** 1.40 (.41) 406*"
Divorce/Widowedb .28(.39)1.33 .31 (.39) 1.36
Mam-ed" .50 (.48) 1.65 .46 (.48) 1.58
Prior J Commite .55 (.34) 1.74 .53 (.34) 1.70
Prior Prisone .36 (.31) 1.43 .35 (.31) 1.41
Person Crimec -62 (37) 1-36 .64 (.37) 1.90
PSafety Crimec "-15 C44) -86 -.17 (.44) .84
Past TX Services -.06 (.09) .94 -.05 (.09) .95
Sup Intensity -.O4 (.01) .96*** -.04 (.01) .96***
Grad ofo“ -.83 (.37) 44* -.78 (.38) .45...
TX Group“ .83 (.32) 2.30** .80 (.32) 2.22**
Hospitalized“ -.42 (.34) .66 -.41 (.35) .66
MH Treatmente .36 (.41) 1.43 .35 (.41) 1.42
H Edue -.07 (.37) .93 -.O6 (.37) .94
House Stability -.17 (.08) .84* -.08 (.10) .92
Emp Stabili --10 (.03) .90*** -.12 (.03) .89***
House*Emp -.63 (.47) .53
Dose of SATX -.08 (.04) .92* -.08 (.04) .92*
SATX Program Violations -.O9 (.13) .91 -.O6 (.14) .94
Abscond" 1.47 (.31) 4.36*** 1.48 (.31) 4.41***
Cox and Snell R2 .41 .41
Nagelkerke R2 .57 .57
-2LL 381.05 379.26
Model ChiZ 269.00*** 270.792”:
(if 23 24

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with standard errors in
parentheses and followed by a standardized coefficient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

The treatment process variables of treatment dosage and absconding were directly

related to re—incarceration after controlling for the partial effects of housing and

170

employment stability“. Substance abuse treatment dose was negatively related to re-
arrest and suggests that higher dosages of treatment received are associated with reduced
likelihood of re-incarceration. Absconding was positively related to the likelihood of re-
incarceration. Individuals who absconded at least one time were approximately four
times more likely to be re-incarcerated relative to those who remained active throughout
their supervision term.

Previous models used to assess the direct effect of housing stability on re-
incarceration did not control for the partial effects of treatment processes. These models
indicated that housing stability was not directly related to re-incarceration likelihood.
While the partial effects of treatment processes may inﬂuence the effect of housing
stability on re-incarceration in logistic regression models, this effect was not veriﬁed with
negative binomial models. Given the totality of these observations, there was no
evidence to suggest that the relationship between housing stability and re-incarceration
was mediated by treatment processes.

The direct effect of employment stability was negatively associated with the
likelihood of re-incarceration after controlling for treatment processes. The effect was
observed after the partial effect of the interaction between housing and employment
instability was controlled. As the stability of employment increased, the likelihood of re-
incarceration is decreased. The negative association between employment stability and
re-incarceration was also observed in previous models that did not control for treatment

processes.

 

54 The results for the treatment process variables in the full conditional models were conﬁrmed through
negative binomial regression with robust standard errors. See Appendix D for further information.

171

Employment stability was found to be positively associated with treatment dosage
levels and negatively associated with the likelihood for absconding in previous models.
That is, individuals with relatively stable employment received higher dosages of
treatment and were less likely to abscond. In all, the signiﬁcance of the relationships
between housing stability, treatment dosage, absconding likelihood, and the overarching
outcome measure of re-incarceration likelihood indicate that employment stability
indirectly affected the likelihood of re-incarceration through treatment dosage levels and
absconding likelihood. Higher levels of employment stability were associated with
higher levels of treatment dose, which reduces the likelihood for re-incarceration.
Additionally, higher levels of employment stability were associated with a reduced
likelihood in absconding. In turn, those who do not abscond were also less likely to be
re-incarcerated.

A number of control variables were related to re-incarceration after controlling for
the partial effects of the independent and treatment process variables. All of the effects
were consistent with previous models that assessed the relationships between housing
stability, employment stability, and re-incarceration. Educational background status was
positively associated with re-incarceration likelihood. Individuals who completed a GED
were just over 2.5 times more likely to be re-incarcerated compared to those who did not
obtain a high school diploma or GED. Those who obtained a high school diploma and
who may have enrolled in college or professional degree coursework were 4 times more
likely to be re-incarcerated relative to those who did not obtain a high school diploma or

GED.

172

Activities during the supervision term were again found to inﬂuence re-
incarceration likelihood once treatment processes were controlled. Individuals who were
exposed to relatively intense supervision were less likely to be re-incarcerated. Those
who successfully completed a substance abuse treatment program during their
supervision term were also less likely to be re-incarcerated relative to those who were
unable to complete any substance abuse treatment services. Once again a direct group
effect was observed. Individuals who were assigned to the treatment group and
subsequently participated in intensive, reentry-based substance abuse treatment were just
over two times more likely to be re-incarcerated relative to those who experienced
traditional supervision services.

Once treatment processes were controlled, the control variable for prior juvenile

commitment was no longer associated with the likelihood for re-incarceration55
Individuals with past juvenile commitments were no more or less likely to be re-
incarcerated after controlling for treatment processes. It appears that this control variable
partially interacts and become subsumed into the direct effects of the treatment process
variables.

Summation of Results for Research Question 3: Relapse and Recidivism Outcome
Measures Controlling for Treatment Dosage and Processes

 

55 Two additional variables did not appear to be related to re-incarceration after controlling for the partial
effects of treatment processes in full conditional logistic regression models. Age and those convicted for
crimes against persons were signiﬁcant in previous logistic regression models. Replication of the full
conditional models controlling for the effects of treatment processes with negative binomial regression and
robust standard errors suggests that the variables of age and conviction offense type were associated with
re-incarceration likelihood (see Appendix D for further information). Older individuals were less likely to
be re-incarcerated, while those convicted of crimes against persons were 1.25 times more likely to be re-
incarcerated relative to drug offenders. Another indicator of criminal history — prior prison sentence — was
signiﬁcantly related to re-incarceration likelihood in a negative binomial model that controlled for the
interactive effect of housing and employment instability. Those who had previously served a prison term
were 1.14 times more likely to be re-incarcerated relative to those who are serving their ﬁrst prison term

173

The goal of the ﬁnal phase of analysis was to examine the relationships between
housing and employment stability on the outcome measures of relapse, re-arrest
incidents, and re-incarceration events after controlling for substance abuse treatment
process indicators of treatment dosage, treatment program violations, and absconding.
All of the models appeared to beneﬁt from the inclusion of treatment process measures
relative to previous models that did not control for the measures. The model ﬁt indices
remained proportionately similar for relapse models, while the models of re-arrest and re-
incarceration experienced substantial reductions in log-likelihood ratios.

The variance explained statistics suggest that the inclusion of treatment process
measures provided higher percentages of explained variability within each outcome
measure (in upwards of 5% to 6% depending on outcome measure). Relapse and re-
arrest outcome models had relatively similar levels of explained variance. Re-
incarceration outcome models provided just over double the amount of explained
variance as the relapse or re-arrest outcome models.

Overall, housing and employment stability had mixed effects on the outcome
indicators. Housing stability was observed to be related to re-arrest, with relatively
higher levels of housing movement being associated with a higher likelihood of re-arrest.
As employment stability increased, relapse and likelihood of re-incarceration decreased.
The direct effect of employment stability on the outcome indicators remained after
controlling for the joint contributions of housing and employment instability. The
interaction of housing and employment stability was not directly related to any of the

outcome indicators.

174

Treatment process measures were observed to have relatively stronger
associations with the outcome variables compared to the independent variables of
housing and employment stability. Treatment dosage had consistent direct effects across
the outcome indicators after controlling for housing stability, employment stability, and
control variables. Relatively high dosages of substance abuse treatment were associated
with higher relapse proportions of positive drug tests and reduced likelihood of re-arrest
or re-incarceration.

The remaining indicators of treatment processes had mixed effects. The number
of substance abuse treatment program violations were positively associated with relapse
proportions of positive drug tests and negatively associated with the likelihood of re-
arrest. Individuals who absconded at least one time during their supervision term were
two times more likely to be re-arrested and just over four times more likely to be re-
incarcerated relative to those who remained active on their supervision term. It is worthy
to note that despite the mixed effects across outcome indicators all three measures of
treatment processes were signiﬁcantly related to re-arrest likelihood.

The control variables inﬂuenced some of the outcome variables, but these effects
were not consistently found across all three outcome measures. Race and prior criminal
history of prison admission differentiated the effects of relapse. The successful
completion of a substance abuse treatment program was negatively associated with
positive drug test proportions and lowered likelihood of re-incarceration, but was not
associated with re-arrest likelihood. Educational background status and activities across
the supervision term were associated with recidivism likelihood, but not with relapse.

Individuals who obtained a high school diploma and who may have pursued higher

175

education were more likely to be re-arrested or re-incarcerated relative to those without a
high school diploma or GED. Additionally, individuals who obtained a GED were more
likely to be re-arrested relative to those who did not obtain a high school diploma or
GED. The intensity of supervision received across the supervision term was negatively
associated with re-arrest or re-incarceration, with more intense supervision reducing the
likelihood of recidivism. Finally, a group effect differentiated the likelihood of
recidivism. Individuals assigned to the treatment group received intensive reentry-based
substance abuse treatment, but were more likely to be re-arrested or re-incarcerated
relative to those who experienced traditional supervision services.

The ﬁnal series of analyses also provide insights into whether the direct effects of
housing and employment stability on the outcome measures may be indirectly mediated
by the effects of the treatment process measures. Table 19 provides a summary table of
results for the supplemental Sobel’s test (1986) of indirect effects. The Sobel test takes
the product of the coefﬁcient assessing the relationship between the independent variable
and the suspected mediator variable and the coefﬁcient assessing the relationship
between the suspected mediator variable and the outcome variable controlling for the
effects of the independent variable. The coefﬁcients and their respective standard errors
are corrected for binary mediators or outcomes (MacKinnon & Dwyer, 1993). This
adjustment requires that each coefﬁcient to be multiplied by the standard deviation of the
independent variable and divided by the standard deviation of the outcome variable in
order to achieve measurement standardization across estimated linear or logistic
coefﬁcients. Once corrected, the coefﬁcients are used to test the null hypothesis that the

indirect effect of the independent variable to the outcome variable through the suspected

176

mediator variable is zero. Signiﬁcant ﬁndings lend support to the observation of a

mediator or indirect effect“.

Table 19: Summary of Aﬂ'usted Sobel’s Tests of Indirect Effects Q1=511).

 

 

Sobel Test Statistic

Housing Stability 9 Dose 9 Relapse 2.97***
Housing Stability 9 Violations 9 Relapse 376*"
Housing Stability 9 Abscond 9 Relapse NS
Housing Stability 9 Dose 9 Re-Arrest NS
Housing Stability 9 Violations 9 Re-Arrest NS
Housing Stability 9 Abscond 9 Re-Arrest NS
Housing Stability 9 Dose 9 Re-Incarceration NS
Housing Stability 9 Violations 9 Re-Incarceration NS
Housing Stability 9 Abscond 9 Re-lncarceration NS
Employment Stability 9 Dose 9 Relapse 2.47***
Employment Stability 9 Violations 9 Relapse -1.79***
Employment Stability 9 Abscond 9 Relapse NS
Employment Stability 9 Dose 9 Re-Arrest NS
Employment Stability 9 Violations 9 Re-Arrest NS
Employment Stability 9 Abscond 9 Re-Arrest NS
Employment Stability 9 Dose 9 Re-Incarceration -1.75*
Employment Stability 9 Violations 9 Re-Incarceration NS
Employment Stability 9 Abscond 9 Re-Incarceration -2.82**

 

*p<.05, **p<.01, ***p<.001; NOTE: Sobel Tests have been adjusted for binary mediators or outcomes
(MacKinnon & Dwyer, 1993). NS refers to non-signiﬁcant indirect effect path as speciﬁed by the Baron
and Kenny (1986) method for determining indirect effects.

The results suggested that indirect effects were present for two of the three
outcome variables. Re-arrest likelihood was the only outcome variable that did not
appear to be indirectly affected by treatment dosage and processes. Based on information
from previous models, housing and employment stability did not have direct effects on
re-arrest likelihood when treatment process measures are left uncontrolled. Since an
initial direct effect between housing stability, employment stability, and re-arrest was not

observed, the indirect effects of treatment processes are not present. Instead, housing and

 

56 Table 19 provides supplemental Sobel’s test statistics after the presence of a mediator or indirect effect
has been made with the Baron and Kenny (1986) method. Appendix E provides alternative Sobel test
statistics for each indirect path effect irrespective of an established direct effect.

177

employment stability directly affected treatment processes and the treatment processes
directly affected re-arrest likelihood.

Substance abuse treatment dose and the number of treatment program violations
appeared to completely mediate the relationship of housing stability and the proportion of
positive drug tests. Complete mediation is observed since the initial direct effect of
housing stability on relapse became insigniﬁcant once the treatment process variables
were entered into the conditional model (see Baron & Kenny, 1986; Kline, 2005).
Individuals with a relatively high number of housing movements appeared to receive
higher dosages of substance abuse treatment and these higher dosages were related to
higher proportions of positive drug tests. Moreover, the high number of housing
movements tended to lead to a higher number of treatment program violations, which
also corresponded to a higher proportion of positive drug tests. At the opposite end of the
continuum, these ﬁndings suggested that those with relative stable housing received
smaller dosages of treatment, which contributed to proportionately fewer positive drug
tests.

The relationship between employment stability and relapse also appeared to be
indirectly affected by substance abuse treatment dose. The indirect effect appeared to be
a partial mediator since the direct effect of employment stability on the outcome variable
modiﬁes the magnitude of the coefﬁcient but the coefﬁcient remained signiﬁcant after
controls for treatment processes. Individuals with relatively stable employment appeared
to receive a higher dosage of treatment across their supervision term, which contributed

to higher proportions of positive drug tests. By implication, unstable employment

178

appeared to be associated with lower dosages of treatment and low proportionality of
positive drug tests.

The number of substance abuse treatment program violations also appeared to
partially mediate the relationship between employment stability and relapse. Stable
employment appeared to reduce the number of treatment program violations. This
reduction in violations was associated with lowered proportion of positive drug tests.

Substance abuse treatment dose and absconding also appeared to mediate the
relationship between employment stability and re-incarceration. Individuals with
relatively stable employment tended to receive a higher dose of substance abuse
treatment and these levels of treatment dose were associated with a reduced likelihood of
re-incarceration. The inverse of the effect suggested that unstable employment
negatively affected the levels of dosage received and increased the likelihood for re-
incarceration. Stable employment also reduced the likelihood of absconding from
supervision, which reduced the likelihood for re-incarceration.

In all, these results suggested that substance abuse treatment dose may be an
important mediator of housing and employment stability on relapse and an important
mediator of employment stability on re-incarceration likelihood. Treatment non-
compliance also appeared to be an important mediator. The number of treatment
program violations mediated the relationship between housing stability and relapse and
employment stability and relapse. Absconding mediated the relationship between

employment stability and re-incarceration likelihood.

179

CHAPTER V: DISCUSSION AND CONCLUSION

State correctional systems are working to develop prisoner reentry infrastructures
that may ease the process of transition into the community and improve relevant program
outcomes. An assortment of reentry-based programs have been implemented and will
continue to be reﬁned and expanded with the passage of time. The programs often utilize
a number of different service delivery structures, strategies, and content to affect the
transition process. As such, there is a large degree of reentry programming variability
across and within state correctional systems. Unfortunately, very few reentry programs
have been subject to formal evaluation. The theoretical promise of reentry programming
has yet to be reliability translated into practice. By implication, the knowledge base used
to inform the development and implementation of reentry-based programming is
currently lacking and needs further development.

The primary purpose of the current study was to explore how speciﬁc reentry
dimensions interact and affect the transition into the community. The research attempted
to inform future pro grams by un-packaging the black box that surrounds reentry-based
programming and assessed the relative contributions of reentry and treatment
components. Central to the study were the interrelationships between the reentry sub-
components of housing, employment, and substance abuse treatment. It was expected
that these three sub-components inﬂuence treatment processes that may directly or
indirectly affect program outcomes of relapse and recidivism. More speciﬁcally, the
overarching research questions suggested that housing and employment stability would
contribute to beneﬁcial treatment processes, reduced levels of relapse, and reduced

likelihood of recidivism. Additionally, housing and employment stability would also

180

indirectly lead to reduced levels of relapse and recidivism likelihood through associations
with treatment processes that are beneﬁcial to treatment participants.

Overall, the results suggest that the stability of housing and employment can
directly inﬂuence treatment processes and relevant correctional outcome indicators.
Housing stability was associated with substance abuse treatment dosage, treatment
program violations, likelihood of absconding, and levels of relapse. Individuals who had
relatively stable housing generally received a low dose of treatment, had few treatment
program violations, were less likely to abscond, and had low proportions of positive drug
tests. Employment stability was found to be associated with treatment dosage, program
violations, likelihood of absconding, relapse, and re-incarceration. Individuals who had
relatively stable employment generally received a high dose of treatment, had few
treatment program violations, were less likely to abscond, had low proportions of positive
drug tests, and were less likely to be re-incarcerated.

After controlling for the partial effects of treatment processes, the direct effect of
housing stability on levels of relapse was not replicated. Housing stability was found to
be associated with re-arrest likelihood. Individuals with relatively stable housing were
less likely to be re-arrested. The ﬁndings for the direct effects of employment stability
were replicated after controlling for the partial effects of treatment processes. Individuals
with relatively stable employment had low proportions of positive tests and were less
likely to be re-incarcerated. Once again, employment stability was observed to be
unrelated to re-arrest likelihood.

While these direct effects were apparent, it is important to note that the effects of

housing and employment were relatively weak. Control variables used throughout the

181

models generally provided a substantial amount of explained variance across speciﬁed
treatment process or programmatic outcomes. Exceptions were observed in models that
assessed treatment program violations and relapse without controls for treatment
processes. In terms of the former, housing and employment stability measures provided
slightly more explanatory power than the control variables. Within these models,
housing stability was found to be the strongest predictor of treatment program violations.
In terms of the latter, housing and employment stability provided comparable explanatory
power to the control variables. Within these models, housing stability was found to be
the strongest predictor of relapse and is subsequently followed by the measure of
employment stability.

Treatment processes were also observed to have direct effects on outcomes after
controlling for the partial effects of the independent variables. Those who received a
higher dosage of substance abuse treatment had relatively higher proportion of positive
drug tests, but were less likely to be re-arrested or re-incarcerated. Treatment program
violations were associated with relapse and re-arrest. Higher levels of treatment program
violations led to higher proportions of positive drug tests, but reduced the likelihood of
re-arrest. Absconding was also observed to be related to re-arrest and re-incarceration.
Those who absconded were over 2 times more likely to be re-arrested and over 4 times
more likely to be re-incarcerated relative to those who did not abscond (i.e., remained
active) throughout their supervision term.

These ﬁndings suggest that participation in treatment may serve as a protective
factor from recidivism, but not relapse. Treatment program violations may be indicative

of non-compliance with treatment, but this does not appear to translate to recidivism risk.

182

It may be the case that program sanctions are enough to maintain or modify behavior
within the treatment program and prevent non-compliant behavior from escalating to the
point of affecting one’s supervision status. For instance, absconding appears to inﬂuence
treatment processes, but also inﬂuences one’s supervision status. In turn, these dual
effects contribute to higher risks for recidivism. Additional research is necessary to
question these interpretations, but there does appear to be some evidence to suggest that
the treatment non-compliance measures may produce differential outcome effects.

In addition to the assessment of direct effects, the ﬁnal set of conditional models
controlled for the partial effects of treatment processes on outcome indicators and
provided some indication of indirect or mediation effects. Six indirect or mediation
effects were observed across models estimating direct effects on treatment processes,
direct effects on outcome indicators, and direct effects on outcome indictors controlling
for treatment processes. Two of the six indirect effects were associated with housing
stability. The positive association between housing stability and relapse appeared to be
completely mediated by treatment dosage and treatment program violations. As the
number of housing movements increased, the levels of treatment dosage increased, which
contributed to higher proportions of positive drug tests. Increased housing movement
also increased the number of treatment program violations, which also increased the
proportion of positive drug tests. Alternatively, relatively stable housing lowered the
levels of treatment dosage received and also reduced the levels of treatment program
violations incurred, which also decreased positive drug test proportions.

The remaining indirect effects were associated with employment stability and

were only partial indirect effects. Once again indirect effects on relapse were observed.

183

Stable employment contributed to higher levels of treatment dosage, which led to higher
proportions of positive drug tests. Stable employment also led to fewer treatment
program violations and lower proportionality of positive drug tests. The alternative
interpretation suggests that unstable employment was associated with low treatment
dosages, which led to low positive drug test proportions, but increased treatment program
violations, which leads to elevated positive drug test proportions.

Treatment processes also indirectly shape the relationship between employment
stability and re-incarceration. Stable employment increased the dosage of treatment
received, which reduced the likelihood of re-incarceration. Stable employment also
decreased the likelihood of absconding, which reduced the likelihood of re-incarceration.
On the other hand, unstable employment reduced the treatment dosage received and
increased the likelihood of absconding; both of which increased the likelihood of re-
incarceration.

There were a few notable surprises from the primary independent variables of
interest. First was the lack of consistent association with recidivism indicators. The
research question presupposed that housing and employment stability would affect re-
arrest and re-incarceration with stability in both measures being associated with a reduced
likelihood of recidivism. A direct effect of housing stability on re-arrest and re-
incarceration likelihood was not observed in the initial conditional models that did not
control for the effects of treatment processes. However, once treatment processes were
controlled in the ﬁnal models, housing stability was found to inﬂuence the likelihood of

re-arrest.

184

A review of the previous ﬁndings provides some possible explanations.
Treatment processes appear to be relevant, and highly signiﬁcant, variables that were
censored from previous models. The inclusion of treatment process variables modiﬁes
the effects of housing stability. It appears that housing stability interacts with the
treatment processes, which modiﬁes the estimated coefﬁcient and inﬂates the statistical
signiﬁcance of housing in re-arrest outcome models (i.e., estimated coefﬁcient changes
direction and magnitude, but maintains similar standard error).

Employment stability was observed to have a direct effect on re-arrest in
unconditional and partial conditional models, but this effect was negated after the
inclusion of control variables. This ﬁnding was observed for models that censored and
included controls for the partial effects of treatment process measures. Employment
stability was consistently found to affect re-incarceration likelihood across all of the
models used for analyses. As such, it appears that employment stability may be a more
reliable and valid indicator of recidivism when the measure of recidivism is
conceptualized through correctional based deﬁnitions.

Second was the direction of the effect for housing stability on substance abuse
treatment dosage. The research question assumed that housing stability would be
associated with higher dosages of treatment, while housing instability would be
associated with a low dose of received treatment. The ﬁndings do not support this
assumption. lrrstead, the opposite is observed to be true; housing stability was found to
be associated with low dosages of treatment, while housing instability appeared to

correspond to higher dosages of treatment.

185

Two mechanisms may be at play that contributes to this effect. On one hand,
individuals with stable housing were found to have few treatment program violations and
were less likely to abscond. These individuals are likely to be compliant and making

gains in other pro-social areas. By making such gains, the perceived need for treatment

may become secondary to maintaining social and economic capita157. On the other hand,
individuals with unstable housing were found to have an elevated number of treatment
program violations and were more likely to abscond. These violations may affect
treatment progress by extending time in treatment through continuation of current
services or discharge and referral to subsequent treatment services that may more
adequately address needs. Either option is likely to inﬂate the level of treatment dosage
received. Additionally, housing instability may be perceived as being at risk for drug use

according to their supervisory agent and be referred to additional substance abuse
treatment services, which elevates the levels of treatment dosage receivedss. Treatment

services, in this regard, can effectively become a source of stability when the stability of

housing is lost or challenged.

 

57 A simple bivariate test confums this assumption. Using median splits of the distribution of housing
movement into high (i.e., 4 or more housing moves) or low (i.e., 3 or fewer housing moves) designations, it
appears that individuals with few housing moves were signiﬁcantly more likely to have been employed for
over half of their supervision term (40%) than those who have many housing moves (27%). There were no
differences between the two groups with regard to marital status, educational background, or the pursuit of
higher education. Interestingly, while remaining compliant with substance abuse treatment services, those
with few housing moves were signiﬁcantly less likely to successﬁrlly complete a treatment program (30%)
than those with many housing moves (47%). This relationship contributes to the ﬁnding of low treatment
dose among those with stable housing. Dosage was measured by estimated months in treatment and the
successful completion of a program generally required numerous months in treatment. The relationship
also adds support to the notion that the fulﬁllment of monetary needs trump treatment needs.

58 A simple bivariate test conﬁrms this assumption. Using median splits of the distribution of housing
movement into high (i.e., 4 or more housing moves) or low (i.e., 3 or fewer housing moves) designations, it
appears that individuals with many housing moves received more intense supervision (by an average of
approximately 14 additional in-person contacts) than those with few housing moves. Moreover, those with

186

Finally, a dichotomous interaction term was used across models that controlled
for the joint contribution of housing and employment instability. Using the overall
distributions of the sample, the term referred to those individuals who had many housing
moves and who were employed for less than half of their supervision term as speciﬁed by
conditional median distributions. Instability was purposely selected as the focus since the
reentry movement focus on pre-release and post-release planning attempts to ease the
transition process and, by default, attempts to increase stability in dimensions of reentry.

Surprisingly, the interaction term did not appear to be associated with treatment
process or outcome indicator models used in the analyses. The only exception is for the
model that assessed the treatment process indicator of substance abuse treatment program
violations. Within the model, the observed effects suggest that the effect of additional
housing movements leading to a high average of treatment program violations is
conditioned by employment stability. Individuals with 4 or more housing moves who
have been employed for less than half of their supervision term tend to have more
treatment program violations, on average, relative to the remainder of the sample.

Part of the dearth of ﬁndings regarding the interactive effects of housing and
employment stability may be due to the solitary consideration of elevated housing and
employment instability. Additional combinations of the interaction between housing and
employment stability were constructed and explored to provide additional context of the
complexities that surround considerations of housing and employment (see Appendix F).
The results conﬁrm the interactive effect of the primary independent variables on the

treatment process indicator of treatment program violations and suggest that stable

 

many housing moves are also more likely to have completed treatment than those with few housing moves,
which also increases observed treatment dosage levels.

187

housing may reduce program violations regardless of stable or unstable employment.
The results also indicate that the solitary consideration of housing and employment
instability masked other important housing and employment combinations that affect re-
incarceration likelihood. Stable or unstable housing appear to interact with stable
employment and increase the likelihood of re-incarceration.

A number of the control variables utilized throughout the analyses provided
valuable insights into the potential for differential effects and how such effects inﬂuence
treatment processes and/or relevant correctional outcomes. The previous chapter
provided individualized discussion of all of the effects for speciﬁc control variables.
There are a few important issues and trends that have been observed with regard to the
control variables and their inﬂuence on treatment processes and program outcomes.

First is the issue of self selection. One of the most consistent results observed
across the analyses are the beneﬁcial outcomes that result from the successful completion
of a substance abuse treatment program during the supervision term. Individuals who
completed a treatment program received a higher dosage of treatment, incurred less
treatment program violations, were less likely to abscond, and were less likely to be re-
incarcerated relative to those who did not complete any sort of substance abuse treatment
program. It also appears that individuals who completed substance abuse treatment
programs may have reduced levels of relapse in replicated negative binomial models.

Self selection is a common problem for correctional evaluators and evaluations in
general (Campbell & Stanley, 1966). A certain proportion of subjects will separate
themselves from the sample population regardless of evaluative design and will directly

beneﬁt from treatment services. This is not detrimental to future program development

188

since a proﬁle of successful participants can be developed. The problem is the difﬁculty
of identifying successful participants prior to admission to a program and the initiation of
services.

Assessment techniques continue to be developed and implemented that will assist
in this process. For instance, motivational interviewing is becoming an increasingly
popular supplement to static and dynamic risk assessments (Checinski & Ghodse, 2004;
Longshore et al., 2004; Miller & Tonigan, 1996; Taxman, 2008). The technique is used
to gauge perceptions of the potential of future self change and self development of
clients. In line with theories of cognitive change, this approach attempts to provide the
tools to foster cognitive transformation by ensuring that an individual is motivated and
ready for change. The difﬁculty is that within this very same theoretical perspective is
the notion that cognitive change cannot occur until an individual is ready to make the
transformation, which may render services to induce motivation and subsequent
treatment services as futile.

Second, there is the question of the dynamics surrounding supervision intensity.
There appears to be two different trends associated with intensity that are associated with
two different sources of supervision. The traditional focus of supervision intensity
generally refers to supervision from an individual’s supervision agent. The available
literature has been critical of the effects of intensive Supervision with intensity being
associated with increased likelihood of re-incarceration due to inﬂated technical
violations (Petersilia & Turner, 1991; Turner et al., 1992). However, there is some
evidence to suggest that the pairing of intensive supervision with treatment services can

reduce the effect of supervision on re-incarceration and beneﬁt treatment participants

189

(National Research Council, 2008; Petersilia & Turner, 1991; Taxman, 2008; Useem &
Piehl, 2008). The second avenue of supervision that is often overlooked in the literature
is derived from participation in community-based treatment services. Treatment service
providers can also act as agents of supervision and inﬂuence program processes and
outcomes.

To some extent, the current ﬁndings appear to support the notion that intensive
supervision paired with treatment services can induce treatment compliance and beneﬁt
individuals enrolled in treatment services. Individuals exposed to relatively intense
supervision received a greater dose of treatment and were less likely to recidivate. This
ﬁnding needs to be prefaced by the type of treatment services received. Individuals who
participated in the intensive, reentry-based substance abuse treatment also received a
greater dose of treatment. However, these individuals had more treatment program
violations, were more likely to abscond, and were also more likely to be re-incarcerated
relative to those who participated in traditional treatment services net all of the control
variables (which includes supervision intensity). When treatment processes were
controlled in the ﬁnal models, individuals who participated in the intensive treatment
were also more likely to be re-arrested relative to those exposed to traditional treatment
services.

There appears to be a differential supervision effect that may be due to

participation in the intensive, reentry-based substance abuse treatment pro gram59

Individuals within the program were assigned to a number of program staff within two

 

59 In a simple bivariate test there was no statistical difference between the correctional supervision
intensity levels of the group assigned to the treatment condition (mean of 44 and standard deviation of
25.06) and the control condition (mean of 46 and standard deviation of 46.55).

190

different community service providers who were in continual contact with an individual’s
supervision agent thorough informal and formal communication channels. It is likely that
the continual oversight created a “ﬁshbowl effect,” where the knowledge of an
individual’s behavior is monitored more closely by a greater number of people with
enforcement or regulatory authority relative to those individuals who were only subjected
to traditional community-based services. Non-compliant behavior is observed
immediately, by a large number of people, and this may partially explain the ﬁnding of
elevated levels of treatment program violations and absconding among the participants of
the intensive reentry—based treatment program.

Future research needs to explore the supervision effects of community-based
treatment services, how they are related to the supervision effects of correctional
supervision, and how they shape program processes and relevant program outcomes.

One of the most promising avenues for research would be the elaboration of McCleary’s
(1975; 1977; 1978) latent power dynamics that exist within correctional ﬁeld ofﬁces.
This elaboration could explore how the demands of community service providers
inﬂuence perceptions of individuals managed under a supervision agent’s caseload. It is
possible that service provider pressure on agents to intervene for non-compliant
participants. In order to be perceived as “doing something,” maintain a generalized
reputation, and regain some degree of autonomy, agents may be more prone to utilize
severe sanctions (e. g., technical returns to prison) over graduated sanctions in the
community.

Third is the potential for effects speciﬁc to those who have been previously

processed through the correctional treatment system. Veterans of the system appear to be

191

more likely to be non-compliant with substance abuse treatment services independent of
the effects of age. The ﬁndings suggested that individuals who had a longer history of
participation in correctional substance abuse treatment had more treatment program
violations and were more likely to abscond. Individuals who have served a prior prison
sentence were more likely to abscond and had a higher proportion of positive drug tests
after controlling for the effects of treatment processes. In negative binomial models,
those who have served a previous prison sentence had a higher proportion of positive
drug tests across conditional models and were also more likely to be re-incarcerated after
controlling for the interaction of housing and employment instability.

It is possible that the continual exposure to similar forms of correctional substance
abuse treatment services (i.e., content and structure) leads to ﬁ'ustration and non-
compliance during the current transition into the community (Marlowe, 2006). It may
also be possible that program non-compliance occurs due to knowledge of how to “play
the game.” These individuals may know how to push the rule boundaries and are aware

that the punishments for non-compliance with substance abuse treatment services are

likely to result in graduated sanctions instead of returns to prison60. There may also be
some degree of ﬂexibility on the part of service providers. Program violations may be
relayed to Supervision agents, but the individual may be allowed to continue participation
in treatment.

The potential ﬂexibility afforded to veterans of correctional treatment by service

providers and/or their supervisory agent only appears to apply to within treatment

 

6O . . . . . . . .

There were no bivariate relatronshrps between past levels of part1c1patron in substance abuse treatment
and recidivism indicators of re-arrest and re-incarceration. Similarly, there were no bivariate relationships
between past prison admission and the recidivism indicators.

192

program behavior. Once again, absconding affords little to no ﬂexibility. Past
participation in correctional substance abuse treatment is not associated with recidivism
indicators, but absconding was signiﬁcantly increased the likelihood of re-arrest and/or
re-incarceration.

Finally there are a few concerns regarding individual effects that may affect
treatment processes, which may contribute to program outcomes. Controlling for all
other partial effects, individuals with higher background educational statuses were more
likely than those lacking a GED or high school diploma to be re-arrested or re-
incarcerated. These higher educational statuses include those who possess a GED as well
as those who possess a high school diploma and who may have received college or
professional degree coursework.

This ﬁnding was unexpected. One potential explanation relates to the content of
services offered to the sample. Through assessment and criminal history information,
members of the sample were determined to represent a high risk population in need of
outpatient or inpatient substance abuse treatment services. It is likely that the provided
services were provided in a utilitarian fashion with the content of intervention being
designed for those lacking any sort of educational background. As such, the needs of
those who have some form of educational background may have been overlooked. The
applicability of services may have contributed to the increased likelihood of recidivism.
Future research may beneﬁt into further exploration into the needs of individuals who
possess a traditional education foundation and whether these needs are met during the

transition process.

193

Another interesting ﬁnding concerned differential relapse. The ﬁndings suggest
that individuals who have been hospitalized at least one time have signiﬁcantly higher
proportions of positive drug tests than those who were never admitted. The difﬁculty
with interpreting this ﬁnding is the fact that two different scenarios may have taken place.
On one hand, a hospital admission may be due to an existing medical problem or incurred
injury that leads to admission. Existing medical conditions may or may not be related to
a history of illegal drug use. In this case, prescribed medication likely follows release
from the hospital. On the other hand, a hospital admission may be the result of self-
medication. In this case, prescribed medication may or may not be provided. Either
scenario appears to affect substance abuse testing and contributes to elevated levels of
positive drug tests.

It is important to attempt to separate these effects in order to provide an
appropriate programmatic response. Program ﬂexibility is also needed to ensure that
relapse and positive drug test results are not artiﬁcially inﬂated by prescribed medication.
Moreover, drug testing schedules and responses must be ﬂexible enough to ensure that
sanctions are not enforced for positive drug tests associated with medication.

Study Limitations

A number of limitations to the study need to be addressed. Most of the limitations
pertain to measurement issues. Nearly all of the measures utilized for this study were
available or constructed from local Department of Corrections or State Police
management information systems. These ofﬁcial sources of data often contain a mixture
of offender behavior in addition to administrative or operational behavior of

organizational representatives (McCleary, 1977; 1978). For instance, McCleary (1977)

194

has argued that the records maintained by parole agents are more instrumental to the
management of agent image to supervisory staff than reﬂective of interactions with
offenders on an agent’s caseload. As such, it is important to recognize that some degree
of bias is inherent to the data sources and measurement used in this study.

The problems associated with ofﬁcial data sources are particularly damaging for
the main independent variables used in the analyses. Ideally, measures of residences,
employers, and participation in substance abuse treatment would be readily available for
extract and could be assessed for reliability. Unfortunately, this was not the case. The
measures of housing stability, employment stability, and substance abuse treatment
dosage were constructed from reviews of narrative agent case notes. The control variable
measures of hospitalization, mental health treatment referral or enrollment, and
enrollment in higher education were also constructed in a similar fashion.

The case notes provide free form text notations, contain an assortment of
summary codes that can be queried within and across records, and are date speciﬁc. The
quality of content entered into the narrative information is highly variable with some
entries including rich detail and others providing nothing more than a few code words.
The information in the case notes also is contingent upon the process of information
exchange and translation. Correctional supervision agents rely upon their interactions
with offenders to obtain direct or indirect information about their experiences while under
supervision. Once this information is received, the message must be correctly
documented and properly entered into the management information system. At any point
in this exchange process information may be lost, may not be translated into the

management information system, or mis-information may be provided. With these issues

195

in mind, the measures of housing stability, employment stability, substance abuse
treatment dosage, hospitalization, mental health treatment referral or enrollment, and
enrollment in higher education may serve as crude proxy measures since their true value
may have been slightly under or over-reported.

The remaining measures used in the study are readily available and easily
extracted from management information systems. These measures are commonly utilized
by the research and evaluation divisions of the home organizations to conduct in-house
research and audits of information quality. This does not mean that the remaining
measures are without controversy. For instance, ofﬁcial measures of recidivism have
been subject to debate due to many of the same biases that have been previously
mentioned (Maltz, 1984; McCleary, 1977; McCleary, 1978). Re-arrest and re-
incarceration measures were used in an effort to minimize potential bias through multiple
indicators of recidivism.

The analytical strategy utilized for the current study provided some valuable
insight into the process of reentry. However, the statistical methods may not provide a
complete description of the reentry process. Regression models included static measures
of individual background and also included a number of measures that represented
dynamic factors that may have inﬂuenced the transition process across an individual’s
supervision term. The dynamic factors were simpliﬁed and subjected to cross-sectional

analysis techniques to explore the magnitude and direction of partial effects that can

196

inﬂuence the reentry experience. Subsequent research may beneﬁt from longitudinal

models that assess how the partial effects of dynamic variables change over time“

The models used in the study may not have adequately explained variation in the
outcome variables. The model ﬁt statistics are low to moderate across models. At the
low end, the models account for 12% of the variability in re-arrest likelihood. At the high
end, models account for 56% of the variability in substance abuse treatment dosage. The
main independent variables of housing and employment stability appear to provide the
strongest effects in relatively weak models. The proportion of variance explained by the
contributions of only the two independent variables is 8% in relapse outcome models,
which increases to 13% with inclusion of all of the control variables. Relative to the
outcomes for substance abuse treatment program violations, the two independent

variables account for 25% of the variability in violations, which increases to 43% with

inclusion of all of the control variableséz.

 

61 It would be particularly useful for future research (and future replications of the current study) to utilize
survival analyses. It is likely that the measures of housing, employment, and treatment were conditioned
by time. The potential for confounding with time introduces bias into the measures and subsequent
ﬁndings.

62 Multicollinearity diagnostics were conducted. Correlation matrices and variance inﬂation factors (VIF)
were produced for the primary independent variables of housing stability and employment stability, the
treatment process variables, and control variables. Out of 484 possible correlation combinations between
the primary independent variables and control variables, 156 (32%) signiﬁcant relationships were observed
ranging in absolute value from r = .09 to r = .52. The average VIF statistic is 1.43 and ranges in value from
1.10 to 2.37. The VIF statistics suggested that the standard error may have been artiﬁcially inﬂated by the
variables in the model, but the level of inﬂation is not extreme enough (i.e., VIF approximating 4.0) to
disregard the regression models (Fox, 1991). Overall, the correlation coefﬁcients and VIP suggested that
multicollinearity may have inﬂuenced the multivariate models and partially biased the estimated standard
errors. Negative binomial regression models with robust standard errors were used to replicate the results
of the linear and logistic regression models and increase the precision of the estimated coefﬁcients (see
Appendices B, C, and D).

A related concern with the multivariate models used in this study is the level of power used to estimate
coefficients and determine statistical signiﬁcance. Power calculations are contingent upon samples and the
statistical technique to be utilized (Cohen, 1988). Ad-hoc checks of the power of a full conditional linear
regression model with 45 covariates, an anticipated medium effect size (.15), alpha level of .05, and desired

197

Future analyses may beneﬁt from more advanced statistical methods. The method
for determining indirect or mediation effects in the current research utilized the Baron
and Kenny (1986) approach that relies upon the initial establishment of direct effects to
infer indirect effects. A number of direct and indirect effects were observed in the
current study, which establishes a framework that future research can examine. There is
a need to further explore the variance-covariance relationships between the treatment
processes and outcome variables. Methods such as structural equation modeling could be

used to determine the interrelationships between the treatment process variables and how

these associations are inﬂuenced by housing and employment stability63. This analytical
strategy would allow one to explore how the independent variables affect substance abuse
treatment dosage, which may then affect treatment program violations, and contribute to
overall levels of absconding. Once established, the interrelationships between the
outcome variables of relapse, re-arrest, and re-incarceration should be examined and

conditioned by housing stability, employment stability, and the treatment processes

variables“. This strategy would allow one to observe how the independent and treatment

 

statistical power of .80 requires a minimum sample size of 226 without consideration of measurement error
in the covariates. In order to identify small effect sizes (.02) with the same speciﬁcations, a minimum
sample of 1,475 is required. Elaborations of linear regression with generalized linear models generally
require larger sample sizes (Long, 1997). Assessments of interactive effects also require larger sample
sizes (Jaccard, 2001; Jaccard et al., 1990). In all, the consideration of power suggests that the multivariate
models used in the analyses were underpowered and may only be able to identify medium or large effect
srzes.

63 . . .
The correlatrons between substance abuse treatment dosage, treatment program Violations, and

absconding were all statistically signiﬁcant and range in absolute value of .11 to .35, which suggests that
covariance structures between the variables should be explored.

The correlation between re-arrest and re-incarceration was statistically signiﬁcant and moderate (.41).

Surprisingly the correlations between the proportion of positive drug tests and the measures of recidivism
were not statistically signiﬁcant. An alternative measure of relapse was the raw number of positive tests.
When this measure was substituted for the measure of relapse used in the study the correlations between

198

process variables inﬂuence relapse, which shapes recidivism outcomes as well as how re-
arrest likelihood contributes to re-incarceration likelihood.

In order to further explore the process of reentry, future research may also beneﬁt
from dynamic life history event analyses through hierarchical grth curve modeling
(see Homey et al., 1995). This analytical strategy requires time speciﬁed data that is
restructured into a time series format to form an event calendar to explore within-
individual change over time. For instance, the 2 year post-release follow-up data used in
this study could be restructured into 104 weeks. Measures of housing stability,
employment stability, treatment processes, and relapse and recidivism outcomes could
then be re—examined and re-structured to their dichotomous weekly occurrence. Growth
curves of treatment processes could then be examined to determine how gains or losses in
housing and employment affect substance abuse treatment dosage and compliance levels
before and after a gain or a loss. Similar interpretations could be made for growth curves
of relapse and recidivism outcomes.

Finally, the results are also limited in their generalizability. The sample consists
of a high risk population of individuals with severe substance abuse and dependency
histories who were released to a conditional correctional supervision term in a medium
sized Midwestern city. The services these individuals were exposed to or participated in
were also relatively geographically limited to agencies within the city or the surrounding
region. As such, the sample should not be considered representative of all individuals
under correctional supervision in the community or those individuals with substance

abuse or dependency histories.

 

relapse, re-arrest, and re-incarceration were all statistically signiﬁcant and range in absolute value of .16 to
.30.

199

Research Implications

A number of implications can be generated from the current study. Attempts to
offer theoretical implications border on the post hoc criticisms that plague the
development of theoretical knowledge on prisoner reentry (see Chapter 2). A direct test
of any one theory or combination of theories was not attempted with this study. The
results suggest that there is good reason for the lack of theoretical knowledge on the
process of transition from the institution into the community — the reentry process is
complex. Lynch’s (2006) identiﬁcation of theoretical reentry emphases of social control,
social development, methods of supervision, and the immediate transitional experience
from prison to community can all be extrapolated to inﬂuence the process of reentry to
some extent based upon the current study. While a direct test of one theoretical
perspective may provide the best insights to ﬁrture theory development, the
interrelationships between the theories would make it extremely difﬁcult to conceptualize
an approach that is not confounded by the remaining perspectives. It may be more
beneﬁcial to begin to focus on broader theoretical themes that can lead to theory
integration.

Future theoretical developments on prisoner reentry must take into account
broader themes of stability and structure. Each of the theoretical perspectives suggested
by Lynch (2006) directly or indirectly speaks to the importance of these themes.

Theories on the immediate transitional experience provide a useful starting point with the
suggestion that the initial challenges faced upon release must be managed so as to reduce
stress or strain that may lead to reversions to past criminogenic behavior. Elaborations of

the theory and its potential integration with the remaining theoretical perspectives needs

200

to further explore how the stability of reentry dimensions provides daily, weekly, or
monthly structure for life in the community.

Housing and employment stability appear to provide some degree of stabilization
to treatment processes by affecting dosage and non-compliance levels. Stable treatment
may then serve an important function of social control and an avenue for self-
development. That is to say that treatment may only have inherent theoretical value if it
is relatively stable and its stability is contingent upon the stability of broader socio-
economic circumstances. It may be the case that instability across these reentry
dimensions damages the potential for obtaining daily, weekly, or monthly structure in the
community, which may lead to recidivism. Additional research is needed to test these
theoretical claims.

The ﬁndings also have direct theoretical implications for the growing body of
literature on “what works” in corrections. “What works” has evolved to become a theory
of programmatic intervention design that is based upon inductive inferences from reviews
of empirical studies. The perspective generally begins with a speciﬁc treatment program
and considers how program structure and content affect relevant outcome indicators.
Since programs are the focus, the perspective overlooks the fact that treatment processes
may be affected by broader social circumstances related to housing and employment
stability. This suggests that the body of literature on what works may be omitting
relevant variables that contribute to the relationship between programming and program
outcomes. Moreover, this omission conﬁrms the fact that ‘that works” literature may

not be directly applicable to the study of prisoner reentry.

201

Some methodological implications can also be developed from the current
research that is applicable to evaluation research. Evaluations of reentry processes must
make sure that a focus is placed upon the effects of reentry dimensions and develop the
ability to reliably capture measures of relevant reentry dimensions. The focus of the
current research was placed upon the dimensions of housing, employment, and treatment
following release from an institution. A focus on these 3 dimensions is somewhat
limited. Additional focus should be placed on a mixture of pre-prison, within-prison,
release, and post-release reintegration experiences that vary according to individual,
social group, community, and state policy dynamics (V isher & Travis, 2003).

The current research also conﬁrms the need to move beyond the simplistic
presentation of recidivism outcomes in an evaluation research design and for reentry
research in general. It is not enough to simply suggest that sample populations are
making gains in procuring housing, employment, or treatment and then present overall
outcomes pertaining to relapse and/or recidivism irrespective of the reentry dimensions.
The current results indicate that the reentry dimension of housing and employment
stability directly affect treatment processes and relapse outcomes. Moreover,
employment stability directly affects re-incarceration outcomes. A research design focus
speciﬁcally on relapse or recidivism outcomes would overlook these associations. The
same oversight would occur if the associations between treatment dosage and compliance
levels are taken into account with relapse and recidivism outcome analyses.

Evaluations of reentry programming may also beneﬁt ﬁ'om an attempt to separate
the effects of supervision that are inﬂuenced by correctional supervision policy from

those that are inﬂuenced by participation in treatment programming. Correctional

202

supervision intensity paired with treatment can serve as a protective factor that may
reduce the potential for supervision intensity to inﬂate levels of re-incarceration (National
Research Council, 2008; Petersilia & Turner, 1991; Taxman, 2008; Useem & Piehl,
2008). However, this research suggests that it may be the type of treatment that is
associated with recidivism risk. Supervision paired with traditional, community-based
substance abuse treatment services appears to have a lower degree of overall supervision
intensity relative to supervision paired with intensive, community-based substance abuse
treatment that involves multiple program providers and personnel. Intensive treatment
may contribute to intensive community supervision, which appears to signiﬁcantly
increase the risk of recidivism.

Policy implications from the current study conﬁrm that reentry dimensions are
important to the study of release and reintegration into the community. Housing,
employment, and treatment appear to be interrelated and can directly inﬂuence overall
correctional outcomes. These outcomes can also be shaped by treatment processes,
which are themselves conditioned by housing and employment stability.

Irnportantly, the reentry dimensions of housing, employment, and treatment are
inﬂuential to the process of transition into the community regardless of whether an
individual has received reentry-based correctional services. Evidence of differential
effects between those who received intensive, reentry-based substance abuse treatment
and those who received traditional correctional services were observed, but these effects
did not appear to constrain the direct effects of housing stability, employment stability, or
treatment dosage and processes. This suggests that existing policies must attempt to do

better to plan, manage, and provide adequate services for those who are able to

203

participate in reentry-based services. Additionally, new policies must be formulated
drased on the lessons learned from existing policies) to provide some level of reentry-
based services to those who experience traditional correctional services.

Treatment is an important reentry dimension and is an issue that has been
subjected to a litany of correctional research. Referral and participation in treatment may
not be the most adequate response to the challenges of community transition, especially if
the intensity of treatment provides levels of supervision that meet or exceed the levels of
supervision provided by community correctional agents. However, treatment may serve
an instrumental purpose above and beyond its rehabilitative value by enhancing social
capital and self-development. The current results suggest that substance abuse treatment
appears to be most effective in reducing levels of recidivism when dosage levels are high.
By implication, treatment appears to be most effective in reducing recidivism when it
stabilizes housing movement (i.e., those with unstable housing receive higher dosages of
treatment) and reinforces the existing stability of employment (i.e., those with stable
employment receive higher dosages of treatment).

Efforts to enhance the efﬁcacy of substance abuse treatment must also recognize
the challenges to increasing dosage levels and attempting to stabilize housing movement.
Higher levels of dosage were found to be associated with higher levels of relapse.
Unstable housing was found to be associated with higher levels of treatment program
violations and an increased likelihood of absconding. This suggests that treatment
service administrators and staff as well as community correctional supervision agents
must recognize that relapse and non-compliance with treatment programming is a part of

the broader transition and adjustment process. Flexibility is needed in providing sanction

204

responses to relapse and program violations. These violations may be indicative of
behavior associated with criminality, but they may also be indicative of amenability to
treatment or broader socio-economic difﬁculties associated with housing or employment.

The formulation of policy for prisoner reentry is a daunting task. It is very simple
to say that reentry dimensions must be considered and efforts to plan for the initial
obtainment housing, employment, and treatment must be made and associated with
contingency plans of how to maintain housing, employment, and treatment. The existing
reentry program model structures being designed and implemented across the country
have called for similar planning efforts. The difﬁculty in considering these dimensions
and including them into reentry strategies is the fact that housing and employment
statuses can change at any time for any reason. Sometimes theses changes are driven by
individual agency or choice, while other times the changes are made for an individual
without their control. In essence, housing and employment stability may be available for
some individuals but the stability of housing and employment for others may be a matter
of chance. This makes prospects for planning — a central dimension of the reentry
movement — difﬁcult.

Even if plans are solidiﬁed to some extent, the quality and reach of services that
can be provided complicates matters. Housing assistance and programming may be
available in some communities, but there is a dearth of research available to determine
what type of assistance or programming is better than others in terms of obtaining and
maintaining housing. A base rate of expected effect sizes for housing programs has yet to
be established, but there is promising research available on halfway houses and

transitional housing (Bloom et al., 2007; Jacobs and Western, 2007; Lowenkamp &

205

Latessa, 2002; Lowenkamp & Latessa, 2005). Employment assistance and programming
has some empirical merit, but offers very small effect sizes (Aos et al., 2006; MacKenzie,
2006). DevelOped reentry program models all make mention of housing and employment
considerations, but few specify the services that will be offered. It is likely that the
services are referral based and include minimal provisions such as application and
paperwork processing and information on available opportunities (see Justice Research
Associates, 2005). Unfortunately these services may not be adequate. Research is
needed in these areas to inform future reentry programs. Researchers may beneﬁt from
exploring ﬁrnding options associated with partnerships with local housing development
and/or employment development agencies to assess services that could contribute to
reentry program model development.

In a purely humanisitic sense, housing and employment should be provided for
individuals being released into the community simply due to the fact that upon release,
many individuals lack housing placements and employment opportunities. Unfortunately
this recommendation is not economically or politically feasible. A few promising
avenues should be explored.

Housing needs could be met by community corrections in concert with local
housing development providers and landlords interested in receiving tax break incentives
for assisting with housing placements for offenders. Determinations of housing
placement can be made during at pre-release and a dialogue between the landlord,
homeowner, or roommate and the individual to be placed can be established to form
rules, responsibilities, and expectations. Such a process would assist with an initial

housing placement, but does not guarantee that the placement will be maintained

206

throughout the supervision term. Aftercare placement options for the loss of housing
should be considered during the initial placement process. New options for housing
movement should be discussed after progression is made in attaining some degree of
social capital (e. g., obtaining a job, completing treatment), which may contribute to self-
development and reinforce one’s current trajectory of adj ustnrent into the community.
The key is to minimize, not exacerbate, housing movement.

A parallel process could take place for employrrrent. Employers may receive tax
break incentives for placing offenders with employment opportunities. Knowledge of
employers who hire ex-offenders is often shared upon release among the informal
network of other offenders who are experiencing the same transition process. Individuals
are also likely to be referred to local employment development agencies that provide
contact information for employers seeking to hire ex-offenders. Instead of relying upon
informal networks or referrals, it may be a better policy for employers to provide
orientations regarding available positions with their agency during pre-release. This
would open a dialogue between the employer and employee on expectations, begin the
administrative paperwork process, and effectively cut out the “middle man” process of
waiting until release and relying upon referrals. The key is to maximize the amount of
time working.

It is also important to consider the marriage between housing and employment.
Research should also continue to explore transitional housing programs that include
immediate employment placement and the continuation or initiation of treatment services
in an effort to inform policy. The work of Bloom and colleagues (2007) and Jacobs and

Western (2007) suggests that such services can promote gains in social capital and

207

produce reductions in relapse and recidivism. It is likely that the beneﬁcial effects
associated with participation in such services is inﬂuenced by the structure and stability
that is associated with minimized housing movement and maximized employment tenure
as suggested by the current research.

Further elaborations of transitional housing programs are needed. These
programs appear to be reserved for those individuals with treatment needs who lack the
social capital or social networks that can assist with housing and/or employment leads. It
is not clear if individuals who have met their housing needs and/or employment needs
upon release would beneﬁt from such services or if the services would produce
unintended consequences. The current research suggests that those with more social
capital (as indicated by educational background status), may place themselves at greater
risk for recidivism when programming is not applicable or when needs are unmet.

Additional concerns are related to service delivery. Transitional housing and
halfway house programs for offenders are directly or indirectly implemented under the
auspices of the local Department of Corrections. By being associated with the
correctional system, transitional housing may be perceived as being a form of extended
punishment (see Justice Research Associates, 2005). Continued replication of
transitional housing programs should utilize a mixture of facilities that are not associated
with the correctional system to determine if the location of service delivery enhances
program effects.

Conclusion
Overall the results indicate that reentry dimensions are inﬂuential to the process

of transition into the community. The reentry dimensions of housing, employment, and

208

treatment are interrelated and affect outcomes related to relapse and recidivism.
Moreover, housing and employment appeared to Shape the treatment processes of dosage
and violations and these processes, in turn, affect relapse and re-incarceration. While the
reentry dimensions may have direct effects on correctional outcomes, the direct effects of
housing and employment on relapse and re-incarceration are indirectly affected by the
treatment processes of treatment dose and violations.

The reentry movement is still in its infancy and state correctional systems
continue to develop, implement, and reﬁne reentry programs. Research and evaluation
on reentry is still few and far between despite the grth of interest in the topic. It is
concerning that contemporary reentry programs continue to be implemented with little to
no empirical foundation. Proceeding without informed decisions may lead to null or
negative (i.e., those that harm participants) correctional outcomes that affect the
expansion of reentry programs and the funding available to implement such programs.

Moreover, public opinion may impede any progress the reentry movement hopes
to make. According to Petersilia (2003), state funding has shifted away from higher
education -- speciﬁcally higher education — and social services to fund prisons and
correctional programs. The question that policy makers, criminal justice administrators,
and academics will have to answer is to what extent society should assist those
individuals who are considered to be “the lowest of the low” due to their criminal
background. In the name of public safety, should offenders receive as much services and
treatment as deemed necessary in hopes of inducing a behavioral change or should the
use of incapacitation be furthered in order to keep offenders out of communities and

society at large. Responses to such questions by the public and the criminal justice

209

system will determine the extent to which the prisoner reentry movement ﬂomishes or

ﬂounders.

210

APPENDIX A

CONTEXTUAL INFORMATION ON BROADER PROGRAM EVALUATION

High-risk male offenders soon to be released into local jurisdictional limits,
placed under 24 month community correctional supervision, and who had signiﬁcant
substance abuse or dependency histories served as the target population for the program
evaluation. All offenders scheduled to be released into the local jurisdiction were
screened for program eligibility. Eligibility for placement in the program was determined
through the pre-parole process and individuals to be placed in the program were notiﬁed
of placement approximately one month and one week prior to release on average. In
addition to the mentioned pre-requisites, offenders were assessed for eligibility according
to offense and risk classiﬁcations. Ineligible participants included those with high
assaultive risk classiﬁcations, sex offense histories, arson histories, life sentences,
pending felony charges or immigration detainer, physical or mental conditions that may
prohibit participation, paroles from other states under interstate compacts, and those
assigned to minimum community correctional supervision status.

The program structure included graduated transitions between two programmatic
phases. The ﬁrst phase took place during the ﬁrst 30 to 45 days post-release and required
residence in a secure transitional facility. Efforts in this phase were focused on the
establishment of housing placements, the obtainment of employment, enrolhnent in
employability and life skills training, and the initiation of substance abuse treatment
services. Contact with family members was also attempted during this phase in order to
assess the need for family therapy and initiate family therapy sessions as necessary.

Participants were assigned to a primary case worker, primary treatment therapist, and

211

 

case coordinator upon intake who respectively assessed participant need (i.e.,
educational, employment, housing, personal identiﬁcation, mental/physical health, health
care, transportation, and clothing), assessed substance abuse or dependency histories and
developed treatment and relapse prevention plans, and developed local referral networks
in the community in addition to serving as an intermediary when problems arise.
Program staffs worked collaboratively to identify goals and objectives for participants
and monitor participant progress on a weekly basis (at minimum). Moreover, a mid-
treatrnent review was held after 2 weeks in the program with these program staff, the
participant, and the participant’s community correctional supervisory agent to determine
if initial goals and objectives are being met and whether to extend the ﬁrst phase or allow
transition into the second phase of the program.

Substance abuse treatment during the ﬁrst phase consisted of up to 10 hours of
direct treatment services, which included 1 individual counseling session, 3
group/didactic sessions, 1 family group therapy session, and 3 Alcoholics Anonymous or
Narcotics Anonymous sessions per week. Participants were subjected to drug tests at
least 2 times per week in addition to any returns to the transitional housing facility upon
approved day pass leaves for housing or employment prospects and other assorted
community-based responsibilities or referrals. Non-compliance with treatment
participation and positive drug tests resulted in programmatic sanctions. These sanctions
may include building restrictions (loss of day passes), delays in progression to the second
phase of the program, re-starts of the entire program, program terminations, and violation
of community supervision status proceedings that may potentially lead to re-

incarceration.

212

The participant transitioned to the second phase of the program after making
sufﬁcient gains in the obtainment of housing placements and employment or employment
prospects in addition to remaining compliant with substance abuse treatment and testing,
as determined by the participant’s primary case worker, primary therapist, case
coordinator, and community supervisory agent. The second phase allows the participant
to move to an approved home residence, continue substance abuse treatment services, and
continue family therapy sessions as needed. The primary therapist and case coordinator
continued to monitor and assist in meeting program goals and objectives as well as
immediate needs. Bi-monthly and monthly meetings were held among program staff to
review progress, assess goals and objectives, and determine need for modiﬁcation to
goals and objectives. Monthly meeting were held with these program staff, the
participant, and the participant’s community supervisory agent to provide feedback on
progress or discuss issues of non-compliance.

Substance abuse treatment during the second phase consisted of up to 9.5 hours of
direct treatment services. These services consisted of 3 group/didactic sessions, 1 family
group session, and 3 Alcoholics Anonymous or Narcotics Anonymous sessions per week
and 2 individual counseling sessions per month. Drug testing in the second phase was
conducted randomly through the use of a toll-free call-in phone number. Participants in
the second phase were required to call the phone number each business day to check the
listing of testing numbers for that day. If the participant’s Department of Corrections’
number ends in one of the two digits listed on that day, the participant must report to the
transitional housing facility for testing. Positive drug tests result in immediate 3 day

sanctions to be served at the transitional housing facility. Continued non-compliance

213

with treatment or testing may lead to additional 3 day sanctions, placement back in the
ﬁrst phase of the program, program re-starts, program terminations, and violation of
community supervision status proceedings that may potentially lead to re-incarceration.

An experimental design with the random assignment was used to determine the
impact of the program on relapse and recidivism. Eligible offenders were randomly
assigned by Department of Corrections identiﬁcation number. Those with odd ending
identiﬁcation numbers were assigned to the participant group, while those with even
ending identiﬁcation numbers were assigned to the control group.

Insights from Process Evaluation Findings

Overall, the reentry-based substance abuse treatment program was found to be
implemented consistent with the program model. The intensity of the substance abuse
treatment services offered met or exceeded the original program model plan for nearly all
components of the program. Substantial participation variability in treatment services
within and across service components of the program were observed.

There were a few exceptions with regard to implemented program intensity.
Family therapy sessions were attended far less often than originally planned. Much of
the problems with these sessions were due to the skepticism of family members in
participating in treatment services within the transitional housing facility. Efforts to
move the treatment services to local community centers induced more participation, but
the intensity of participation remained low throughout program years of operation.

Employment program services were also different than originally planned. At the
outset, the program was designed to provide placement services. In operation, these

placement services were more akin to referral services. Participants attending local

214

 

{
I.
{3.
“+3.
'5'
~ . X
«s:
.1."
1‘31
4
<":'*',
'3
xii
151;!
$1
{2
’
.I':
3:.
,i'
3 .

employability services and incurred the responsibility of conducting much of the
“legwork” required for obtaining employment on their own.

There were a few issues that accompanied the suitable implementation of the
program. First, the screening process may have missed a number of eligible participants
during the ﬁrst 6 months of program operation. This may have biased the eligibility pool

that may have been assigned to participant or control groups, but any bias would have

been marginalGS. There were difﬁculties in reaching the target sample number (500
eligible offenders) that may have been inﬂuenced by missing eligibility-screens. The
anticipated annual number of eligible offenders provided by the local community
correctional ﬁeld ofﬁce far exceeded the actual number that could have been assigned to
the participant or control conditions. Due to this discrepancy, the random assignment
period was extended one year longer than anticipated.

Second, there was a noticeable degree of hostility from participants who entered
the program during the ﬁrst year of operation. The problem was largely due to the fact
that participants were expecting to be released directly into the community, but were
instead placed into a secure transitional housing facility for the ﬁrst phase of the program.
The levels of hostility were apparent despite the pre-release in-reach portion of the
program in which service providers informed future participants of their placement.

Even with these issues in mind, the process ﬁndings provided some valuable

insights. For participants, the ﬁndings indicate that the referral and admission process

 

65 The issue of eligibility screening led to conscientious debates among program stakeholders.
Supplemental data requests on paroles to the local county and reviews of community correctional ﬁeld
ofﬁce ﬁles suggested that the main issue was not systematically related to missing eligible participants.
Instead, the issue was the failure to complete eligibility fomrs. Offenders released to the local county and
missing eligibility screens were likely to have been excluded due to minimum community correctional
supervision status or substance abuse screening assessments not indicating the need for treatment.

215

 

were completed in a timely manner. Pre-parole eligibility determinations were made
approximately 5 weeks prior to release. Pre-parole in-reach contact by service providers
were conducted shortly after the eligibility determinations were made. Participants
entered the program within 2 days on average after release from the institution. The
ﬁndings also suggest that participants were meeting program expectations. The
completion rate for the ﬁrst phase of the program was high (upwards of 90%).
Participants also seemed to be exposed to employment opportunities in the local
community and successful in gaining employment since a majority of participants were
employed at least once.
Insights from Outcome Evaluation Findings

Outcome evaluation ﬁndings examined the efﬁcacy of meeting the programmatic
objectives of reducing the risk for relapse and recidivism (i.e., re-arrest and re-
incarceration) for participants. The overall ﬁndings suggested null to marginal program
impact. In terms of relapse ﬁndings, participants were drug tested at a signiﬁcantly
higher rate than the control group, had more positive drug tests on average, and incurred
their ﬁrst positive test approximately one month and one week faster than the control
group. However, these apparent differences between the two groups were minimized
after controlling for the frequency of drug testing. Both groups had similar rates of
positive drug tests and a similar proportion of group members with at least one positive
drug test. Concerning recidivism, there were no differences between the proportion and
timing of re—arrest or re-incarceration among the two groups.

Among participants, preliminary explorations found some indication of

differential outcome effects. Compliant participants with few programmatic violations

216

and those with longer program lengths of stay had lower frequencies of positive drug
tests and less likelihood for recidivism. These ﬁndings suggest that program processes,
which are beyond considerations of dosage, can contribute to differential effects and
shape outcomes. Preliminary explorations into the associations between program
participation and outcomes failed to produce clear ﬁndings. This is not to say that
differential effects do not exist. Rather, this ﬁnding suggests that further analyses are
necessary in order to determine the extent to which differential treatment dosages shape

program outcomes.

217

APPENDIX B
VERIFICATION OF MULTIVARIATE REGRESSION ANALYSES OF
TREATMENT DOSAGE AND PROCESSES (RESEARCH QUESTION 1)

Negative binomial regression models were used as diagnostic veriﬁcations of the

results of the linear and logistic regression models used in this study66. The technique
was used for count dependent variables whose non-normal distribution produces a
conditional variance that is larger than its mean (Agresti, 2002; Cameron & Trivedi,
1998; Hilbe, 2007; Long, 1997). As a generalized linear method, negative binomial
regression eases the assumptions of linear regression models. The error structure in
negative binomial regression does not require normally distributed residuals or constant
and independent residual error structures (Cameron & Trivedi, 1998; Hilbe, 2007; Long,
1997). The technique also differs from traditional linear based regression modeling in its
underlying distribution and method of coefﬁcient estimation (i.e., pseudo maximum
likelihood) (Agresti, 2002; Cameron & Trivedi, 1998; Hilbe, 2007; Long, 1997).

Negative binomial regression models were estimated with robust standard errors.
Robust standard errors were used to protect against biased estimates. This bias tends to
occur when linear modeling techniques are applied to non-normally distributed data
(Cameron & Trivedi, 1996; Hilbe, 2007; Long, 1997) and when estimates are generated
from residuals that are inﬂuenced by outliers and heteroskedasticity (Hardin & Hilbe,
2001; Long & Ervin, 2000; Wooldridge, 2009). Negative binomial regression corrects
for biased estimates of standard errors by easing the assumptions of linear regression

(Cameron & Trivedi, 1996; Hilbe, 2007; Long, 1997). The addition of a robust standard

218

 

error function to negative binomial regression models produces a more stringent
estimation of coefﬁcients and determination of statistical signiﬁcance. This is due to the
fact that robust standard errors provide more precision in estimation relative to non-
robust standard error estimates that contain larger standard errors (Hardin & Hilbe, 2001;
Long & Ervin, 2000; Wooldridge, 2009).

Tables 20 through 22 provide the replicated treatment dosage and process models
with negative binomial regression with robust standard errors“. Bolded coefﬁcients and
standard errors represent differences between the linear or logistic regression model
presented in the text and the negative binomial regression model presented below. The
negative binomial model was a replication of the full conditional model that includes the
independent variables and control variables with and without controls for an interaction
term.

Table 20 presents the full conditional negative binomial model of substance abuse
treatment dosage on independent and control variables. Four coefﬁcient estimation
differences were observed between the negative binomial models and the linear
regression models in the text. The differences were associated with two control
variables. Whether or not an individual has been hospitalized during their supervision
term is associated with treatment dosage in the negative binomial models. Those who
have been hospitalized received a higher dose of treatment relative to those who were

never admitted to the hospital during their supervision term. The only other difference

 

6 Caution is needed with the interpretation of these regression models since the models are likely to be
underpowered and may only be able to identify medium or large effect sizes.

67 . . . . . .
Negatrve bmomral regressron models wrth robust standard errors were conducted wrth the use of Stata
IOSe software. Stata utilizes a Huber-White adjustment that produces robust standard errors with

219

between model estimations was between educational background statuses (i.e., high
school graduate with or without college or professional degree coursework relative to no
high school diploma or GED). Importantly, the direction of the effect changed in the
negative binomial regression models, but the coefﬁcient remains insigniﬁcant as

suggested in the linear models.

 

corrections for correlated error terms (Hardin & Hilbe, 2001; see also Eicker, 1967; Huber, 1967; White,
1980 for a technical estimation foundation).

220

Table 20: Veriﬁcation of Full Regressions of Substance Abuse Treatment Dosage on
Independent and Control Variables (n=511).

 

 

 

b (SE) IRR b (SE) IRR
Age .00 (.00) 1.00 .00 (.00) 1.00
Whit; -.17 (.07)* .84 -.17 (.07)* .84
GED“ -.00 (.07) .99 -.01 (.06) .99
HS Grad plus“ .04 (.09) 1.04 .03 (.09) 1.03
Divorce/Widowb .12 (.09) 1.13 .1 1 (.09) 1.12
Mam-edb -.01 (.09) .99 -.01 (.09) .99
prior J Commie -.07 (.07) .93 -.07 (.07) .93
prior prism;3 -.10 (.06) .90 -.10 (.06) .90
Person Crimec '08 (~07) 97— "-08 (~07) -92
property Crimec -.10 (.08) .90 -.10 (.08) .90
PSafety Crimec -.11 (.10) .89 -.11 (. 10) .90
Past TX Services —.01 (.02) .98 -.02 (.02) .98
Sup Intensity .01 (.00)*** 1.01 .01 (.00)*** 1.01
Grad ofoe .50 (.06)*** 1.66 .50 (.06)*** 1.65
TX Groupd .54 (.06)*** 1.73 .55 (.06)*** 1.73
Hospitalized“ -14 (.06)* 1.16 .14 (.06)* 1.15
MB Treatmente .04 (.09) 1.04 .04 (.09) 1.04
H Educ .06 (.07) 1.06 .06 (.07) 1.06
House Stability .07 (.01)*** 1.08 .06 (.02)*** 1.07
Emp Stabili .02 (.00)*** 1.02 .02 (.00)*** 1.02
House*Emp .09 (.08) 1.09
-2LL 1420.49 1420.02
Model Chi2 395.22*** 403 .40***
df 20 21

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.
a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, 0. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were enrployed for less than half of their supervision term.

Table 21 presents the full conditional negative binomial model of treatment

program violations on independent and control variables. Eight coefﬁcient estimation

differences were observed between the negative binomial models and the linear

regression models in the text. The eight differences were reﬂective of one independent

variable and ﬁve control variables.

221

Table 21: Veriﬁcation of Full Regressions of Treatment Program Violations on

Independent and Control Variables (n=511).

 

 

 

b (SE) IRR b (SE) IRR
Age .00 (.01) 1.00 .00 (.01) 1.00
White“ .30 (.13)* 1.34 .29 (.13)* 1.34
may -04 (.13) .96 -.04 (.13) .95
HS Grad plusa -.21 (.19) .81 -21 (.18) .81
Divorce/Widowb .02 (.19) 1.02 .01 (.18) 1.01
Maniedb -.00 (.21) 1.00 .05 (.21) 1.05
priorjcommite .16(.15) 1.17 .16(.15) 1.18
prior Prison“ .01 (.13) 1.01 .02 (.13) 1.01
Person Crimec -36 (-17)* 1-43 .34 (.17)* 1.41
property Crimec .40 (.18)* 1.50 .38 (.17)* 1.46
PSafety Crimec .13 (.22) 1.14 .16 (.22) 1.17
Past Tx Services .08 (.03)** 1.09 .07 (.03)** 1.07
Sup Intensity .OO (.00) 1.00 .00 (.00) 1.00
Grad ofo“ -.84 (.16)*** .43 -.81 (.15)*** .44
TX Groupd .76 (.14)*** 2.14 .75 (.14)*** 2.11
Hospitalizede .06 (.13) 1.06 .05 (.13) 1.05
Mg Treatmente -.57 (.18)** .56 -55 (.18)** .57
H Educ -.10 (.19) .91 -.15 (.18) .86
House Stability .26 (.03)*** 1.29 .18 (.03)*** 1.20
Emp Stabili -.06 (.01)*** .94 -.04 (.01)** .96
House*Emp .58 (20)" 1.78
-2LL 483.70 478.98
Model ChiZ 316.21*** 35971:”:
df 20 21

 

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, C. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

Relative to the independent variable, the effect of employment stability persisted
after controlling for the partial effect of the interaction between housing and employment
stability in the negative binomial models. The effects of the control variables for offense
conviction type (i.e., person and property offenses relative to drug offenses) followed a

similar format and were observed to be signiﬁcant in the negative binomial model that

222

controls for the effect of housing and employment instability. The remaining differences
were for the control variables of marital status (i.e., divorced/widowed relative to single
status), criminal history (i.e., prior prison admissions relative to ﬁrst prison admission on
current supervision term), and hospitalization while under supervision. The direction of
the effect of these control variables changes in the negative binomial models, but the
coefﬁcients remain insigniﬁcant as suggested in the linear models.

Table 22 presents the full conditional negative binomial model of abSconding on
independent and control variables. Seven coefﬁcient estimation differences were
observed between the negative binomial models and the linear regression models in the
text. The seven differences were reﬂective of one independent variable and three control
variables.

The only independent variable difference reﬂects a change in direction between
negative binomial and logistic regression models for the interaction term of housing and
employment instability. Importantly, the insigniﬁcance of the estimated coefﬁcient in
both of the models was the same. Of the control variables, the age of the individual was
associated with absconding in the negative binomial models. This suggested that older
individuals were less likely to abscond. Offense conviction type (i.e., property offenses
relative to drug offenses) was not associated with relapse in the negative binomial
models, despite its signiﬁcant association with relapse in logistic regression models. The
remaining control variable that differs between negative binomial and logistic regression
models represents hospitalization during the supervision term. Importantly, the direction
of the effect changes with negative binomial models, but the coefﬁcients remain

insigniﬁcant consistent with logistic regression estimates.

223

Table 22: Veriﬁcation of Full Regressions of Supervision Absconding on Independent

and Control Variables (n=51 l).

 

 

 

b (SE) IRR b (SE) IRR
Age -.01 (.01)* .99 -.01 (.01)* .99
Whitee -.02 (.10) .98 -.02 (.10) .98
(35133 .06 (. 10) 1.06 .06 (.10) 1.06
HS Grad Plusa -.06 (.15) .94 -.06 (.15) .94
Divorce/Widowb .08 (.13) 1.08 .08 (.14) 1.08
Mam-edb -.08 (.17) .92 -.08 (.17) .92
Pﬁorjcommite .15(.10) 1.16 .15(.10) 1.16
prior tan-gone .27 (.10)* 1.31 .27 (.10)* 1.31
Person Crimec -13(-13) 1-14 -13(-13) 1-14
property cit-mg .23 (.13) 1.26 .23 (.13) 1.26
PSafety Cit-mg .05 (. 16) 1.05 .05 (.16) 1.05
Past TX Services .06 (.02)** 1.06 .06 (.02)** 1.06
Sup Intensity -00 (.00) 1.00 -00 (.00) 1.00
Grad ofoe -.6O (.12)*** .55 -.60 (.13)*** .55
TX Groupd .21 (.09)* 1.24 .21 (.09)* 1.24
Hospitalizede .02 (.10) 1.02 .02 (.10) 1.02
MH Treatmente '-01 (~15) -99 '-01 (J4) -99
H Educ -.22 (.16) .80 -.22 (.16) .80
House Stability .13 (.02)*** 1.14 .13 (.02)*** 1.14
Emp Stabili -.04 (.01)*** .96 -.04 (.01)** .96
House*Emp -.04 (.14) .96
-2LL 371.08 371.06
Model ChiZ 213.98*** 213.24***
df 20 21

 

*p<.05, **p<.01, "*p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, C. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves and who were employed for less than half of their supervision term.

224

APPENDIX C
VERIFICATION OF MULTIVARIATE REGRESSION ANALYSES OF RELAPSE
AND RECIDIVISM OUTCOME MEASURES (RESEARCH QUESTION 2)
Negative binomial regression models were used as diagnostic veriﬁcations of the
results for linear and logistic regressions used in this study (see Appendix B for

additional context). Tables 23 through 25 present the replicated relapse and recidivism

models with negative binomial regression with robust standard errors“. Bolded
coefﬁcients and standard errors represent differences between the linear or logistic
regression model presented in the text and the negative binomial regression model
provided below. The negative binomial model was a replication of the full conditional
model that includes the independent variables and control variables with and without
controls for an interaction term.

Table 23 provides the hill conditional negative binomial model of relapse on
independent and control variables. Three estimated coefﬁcients were observed to differ
between the negative binomial models and the linear regression models in the text. The
differences were associated with two control variables. The number of past correctional
substance abuse treatment services an individual had participated in was not associated
with relapse after controls for the interaction of housing and employment instability in
negative binomial models. Whether an individual had completed a substance abuse
treatment program during their supervision term was associated with relapse with or

without a control for housing and employment instability. Graduates of treatment

 

68 Caution is needed with the interpretation of these regression models since the models are likely to be
underpowered and may only be able to identify medium or large effect sizes.

225

 

programs had a lower proportion of positive drug tests relative to those who were unable
to complete a treatment program.

Table 23: Veriﬁcation of Full Regressions of Substance Abuse Relapse on Independent

and Control Variables (n=511).

 

 

 

b (SE) IRR b (SE) IRR
Age -.00 (.01) .99 -.01 (.01) .99
White“ -.26 (.12)* .77 -.26 (.12)* .77
may -.02 (.10) .97 -.03 (.10) .97
HS Grad plusa -.19 (.16) .82 -.19 (.16) .82
Divomi/WHOWb .03 (.15) 1.03 .03 (.15) 1.03
Mamedb -.09 (.18) .91 -.08 (.18) .92
Prior J Commits -.01 (.12) .99 -.01 (.12) .99
prior prisone .25 (.10)* 1.29 .26 (.11)* 1.29
Person Crimec ‘-12(-13) ~88 '-13 (J3) -88
propem, Crimec -.13 (.13) .88 -.13 (.13) .88
PSafety Crime -.17 (.16) .84 -.17 (.16) .85
Past T'X Services -.02 (.03) .98 -.02 (.03) .98
Sup Intensity .00 (.00) 1.00 .00 (.00) 1.00
Grad ofo“ -.22 (.10)* .80 -.22 (.10)* .80
TX Groupd .16(.10) 1.17 .16(.10) 1.17
Hospitalizede .22 (.09)* 1.24 .21 (.09)* 1.24
MH Treatments -.14 (.13) .87 -.14 (.13) .87
H Edi: -.21 (.13) .81 -.21 (.13) .80
House Stability .08 (.02)*** 1.08 .06 (.02)** 1.06
Emp Stabili -.03 (.01)*** .97 —.02 (.01)** .98
House*Emp .15 (.14) 1.16
-2LL 253.68 253.56
Model C1112 112.10*** “4.85“”
df 20 21

 

*p<.05, **p<.01, ***p<.OOl; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves (median split of sample) and who were enrployed for less than half
of their supervision term

Table 24 presents the full conditional negative binomial model of re-arrest on

independent and control variables. Five estimated coefﬁcients differed between the

226

 

negative binomial models and the logistic regression models in the text. The ﬁve
differences were reﬂective of one independent variable and two control variables.

Table 24: Veriﬁcation of Full Regressions of Re-Arrest on Independent and Control
Variables (n=511).

 

 

 

b (SE) IRR b (SE) IRR
Age -.02 (.01)* .98 -.02 (.01)* .98
White“ —.02 (.14) .98 -.02 (.14) .98
613133 .35 (.17)* 1.41 .35 (.17)* 1.41
HS Grad Plusa .51 (.20)** 1.66 .51 (.20)** 1.66
Divorce/Widowb .22 (.18) 1.25 .22 (.18) 1.25
Mamedb .07 (.22) 1.07 .07 (.22) 1.07
prior J Commite -.11 (.16) .89 -.11 (.16) .89
prior prisone .25 (.16) 1.28 .25 (.16) 1.28
Person Crimec '~01 (-18) ~99 "-01 (~18) -99
property Crimec .13(.18) 1.14 .13(.18) 1.14
PSafety Cit-mg -.21 (.23) .81 -21 (.23) .81
Past Tx Services -.05 (.04) .95 -.05 (.04) .95
Sup Intensity -.02 (.00)*** .98 -.02 (.00)*** .98
Grad of Tx“ -36 (-19) .70 -.36 (.19) .70
TX Gimp“ .l4(.14) 1.15 .l4(.14) 1.15
Hospitalizede .00 (.18) 1.00 .00 (.18) 1.00
MH Treatments -.23 (.26) .79 —.23 (.26) .79
H Educ -.16 (.22) .85 -.17 (.22) .85
House Stability .06 (.03) 1.06 .06 (.04) 1.06
Emp Stabili -.01 (.01) .99 -.01 (.01) .99
House*Emp -.02 (.23) 1.02
-2LL 304.33 304.32
Model Ch12 61 .95*** 62.27***
df 20 21

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, d. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves (median split of sample) and who were employed for less than half
of their supervision term

For the independent variable, the direction of the effect for the interaction term of
housing and employment instability changes in the negative binomial model. However,

the coefﬁcient remains insigniﬁcant, which was consistent with the estimates derived

227

from the logistic regression model. For the control variables, the negative association
between graduates of treatment and re-arrest likelihood in the logistic regression models
was not replicated with the negative binomial models. Additionally, the direction of the
effect of hospitalization during the supervision term on re-arrest likelihood changed in the
negative binomial models, but remained insigniﬁcant consistent with the logistic
regression model results.

Table 25 provides the full conditional negative binomial model of re-incarceration
on independent and control variables. Ten estimated coefﬁcients differed between the
negative binomial models and the logistic regression models in the text. The ten
differences were all reﬂective of control variables.

Two of the control variables differed in terms of statistical signiﬁcance between
negative binomial and logistic regression models. Criminal history background of prior
prison admission was observed to be associated with re-incarceration likelihood after
controlling for the effects of housing and employment instability in negative binomial
models. Individuals who had previously served a prison sentence were 1.27 times more
likely to be re-incarcerated relative to those who are serving their ﬁrst prison term. A
history of prior juvenile commitment was not observed to be related to re-incarceration
likelihood in the negative binomial models.

The remaining differential effects were observed for four control variables. The
variables of race (i.e.., white relative to non-white), conviction offense type (i.e., property
and public safety offenses relative to drug offenses), and past participation in correctional

substance abuse treatment switched direction of estimated effect in the negative binomial

228

models. Importantly, all of these control variables remained insigniﬁcant consistent with

the logistic regression models.

Table 25: Veriﬁcation of Full Regressions of Re-Incarceration on Independent and
Control Variables (n=51 l).

 

 

 

b (SE) IRR b (SE) IRR
Age -.02 (.01)* .98 -.02 (.01)* .98
White“ --02 (-12) .98 -.03 (.12) .97
61393 .42 (.14)** 1.53 .42 (.14)** 1.53
Hg Grad plus? .57 (.16)*** 1.77 .57 (.16)*** 1.76
Divorce/Widowb .20 (.14) 1.22 .21 (.14) 1.24
Mamedb .16(.17) 1.17 .16(.17) 1.17
Prior J Commie .18 (.12) 1.19 .17 (.12) 1.18
Prior Prison“ .23 (.12) 1.26 .24 (.12)* 1.27
Person Crimec 29 (45)“: 1-34 ~30 (-15)* 1-35
property Crimec -.00 (.16) 1.00 -.01 (.16) 1.00
PSafety Crimec -.04 (.19) .95 -.06 (.19) .94
Past TX Services .00 (.03) 1.00 .01 (.03) 1.01
Sup Intensity -.02 (.00)*** .98 -.02 (.00)*** .98
Grad ofoe -.95 (.24)*** .39 -95 (.24)*** .39
TX Groupd .26 (.11)* 1.30 .26 (.11)* 1.29
Hospitalizede -.21 (.18) .81 -.21 (.18) .81
MH Treatments .20 (.18) 1.22 .18 (.18) 1.20
H Educ -.O8 (.17) .92 -.08 (.17) .93
House Stability -.04 (.03) .96 -.00 (.05) 1.00
Emp Stabili -.06 (.01)*** .94 -.06 (.01)*** .94
House*Emp -.23 (.21) .80
-2LL 288.31 288.05
Model ChiZ 265.09*** 265.70***
df 20 21

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves (median split of sample) and who were employed for less than half
of their supervision term

229

APPENDIX D
VERIFICATION OF MULTIVARIATE ANALYSES OF RELAPSE AND
RECIDIVISM OUTCOME MEASURES CONTROLLING FOR TREATMENT
DOSAGE AND PROCESSES (RESEARCH QUESTION 3)
Negative binomial regression models were used as diagnostic veriﬁcations of the

results for linear and logistic regressions used in this study (see Appendix B for

additional context). Tables 26 through 28 provide the replicated relapse and recidivism

models with negative binomial regression with robust standard errors69. Bolded
coefﬁcients and standard errors represent differences between the linear or logistic
regression model presented in the text and the negative binomial regression model
presented below. The negative binomial model was a replication of the hill conditional
model that includes the independent variables, treatment process variables, and control
variables with and without an estimated coefﬁcient for an interaction term.

Table 26 provides the full conditional negative binomial model of relapse on
independent, treatment processes, and control variables. Two estimated coefﬁcients
differed between the negative binomial models and the linear regression models. These
two differences pertained to one control variable. Whether one was hospitalized during
their supervision term inﬂuenced relapse in negative binomial models. Individuals who
were admitted to the hospital at least one time had a higher proportion of positive tests
relative to those who were never admitted to the hospital. This association was not

observed in the logistic regression models.

 

69 Caution is needed with the interpretation of these regression models since the models are likely to be
underpowered and may only be able to identify medium or large effect sizes.

230

Table 26: Veriﬁcation of Full Regressions of Substance Abuse Relapse on Independent,
Treatment Processes, and Control Variables (n=511).

 

 

 

 

b (SE) IRR b (SE) IRR
Age —.01 (.01) .99 -.01 (.01) .99
Whit; -.27 (.12)* .77 -.26 (.12)* .77
GED“ -.00 (.10) 1.00 -.00 (.10) 1.00
HS Grad Plusa -.13 (.16) .87 —.13 (.16) .87
Divorce/Widowb -.03 (.15) .96 -.04 (.15) .96
Marriedb -.10 (.18) .91 -.09 (.18) .91
prior J Commite -.01 (.12) .97 -.02 (.12) .98
pm,r prison“ .28 (.10)** 1.32 .28 (.10)** 1.32
Person Crimec ‘-19 (~13) -83 '-19 (~13) ~83
property Crimec -.18 (.13) .83 -.18 (.13) .83
PSafety Crimec -.17 (.16) .84 -.17 (.16) .84
Past Tx Services -.04 (.03) .96 -.04 (.03) .96
Sup Intensity -.00 (.00) 1.00 -.00 (.00) 1.00
Grad ofoe -.26 (.11)* .77 -.27 (.11)* .77
TX Groupd -.09 (.1 1) .91 -.09 (.11) .91
Hospitalized“ ~19 (~09)* 1.21 .19 (.09)* 1.21
MH Treatmente -.09 (.13) .92 -.09 (.13) .92
H Educ -.24 (.13) .78 -24 (.13) .78
House Stability .01 (.02) 1.01 .01 (.02) 1.00
Emp Stabili -.03 (.01)*** .97 -.03 (.01)*** .97
House*Emp .03 (.14) 1.03
Months in SATX .05 (.01)*** 1.05 .05 (.01)*** 1.05
SATX Violations .15 (.04)*** 1.16 .15 (.04)*** 1.16
Absconde .12(.12) 1.13 .12(.12) 1.13
-2LL 249.77 249.77
Model Chi2 137.19*** 137.78.."
df 23 24

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves (median split of sample) and who were employed for less than half
of their supervision term

 

231

Table 27: Veriﬁcation of Full Regressions of Re-Arrest on Independent, Treatment

Processes, and Control Variables (n=511).

 

 

 

b (SE) IRR b (SE) IRR
Age -.02 (.01) .98 -.02 (.01) .98
Whit; -.03 (.14) .96 -.03 (.14) .97
may .31 (.17) 1.36 .31(.17) 1.36
Hs Grad plus“ .47 (.20)* 1.61 .47 (.20)* 1.61
Divorce/Widowb .27 (.17) 1.31 .27 (.17) 1.31
Maeiedb .07 (.23) 1.08 .08 (.23) 1.08
prior J Commit? -.13 (.16) .88 -.13 (.16) .88
Prim-prison“ .17(.16) 1.19 .17(.16) 1.19
Person Crimec -01 018) 1-01 ~01 {-13) 1-01
propeﬂyctimec .13(.18) 1.13 .13(.18) 1.14
PSafety Crimec -24 (.22) .78 -.24 (.22) .79
Past TX Services —.03 (.05) .96 -.04 (.05) .96
Sup Intensity -.01 (.00)** .99 -.01 (.00)** .99
Grad of Txe -.15 (.20) .86 -.16 (.20) .85
TX Groupd .42 (.16)** 1.53 .43 (.16)** 1.53
Hospitalizedc .03 (.18) 1.03 .03 (.18) 1.03
MH Treatments -.31 (.25) .73 -.31 (.25) .74
H Educ -.12 (.21) .89 -.12 (.21) .88
House Stability .11 (.04)** 1.12 .10 (.05)* 1.10
Emp Stabili -.01 (.01) .99 -.00 (.01) 1.00
House*Emp .11 (.23) 1.11
Months in SATX -.06 (.02)** .94 -.06 (.02)** .94
SATx Violations -.28 (.09)*** .75 -29 (.09)*** .75
Absconde .50 (.17)** 1.65 .50 (.17)** 1.65
-2LL 293.93 293.86
Model Chi2 109.77*** 109.96"...
df 23 24

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves (median split of sample) and who were employed for less than half
of their supervision term

Table 27 presents the conditional negative binomial model of re-arrest on
independent, treatment processes, and control variables. One estimated coefﬁcient

differs between the negative binomial model and the logistic regression models in the

232

text. The difference was observed for one independent variable. Housing stability was
observed to remain signiﬁcantly associated with re-arrest likelihood after controlling for
the partial effect of housing and employment instability. High housing movement
increased the risk of re-arrest, with each additional housing movement increasing the risk
for re-arrest by 1.10.

Table 28 provides the conditional negative binomial models of re-incarceration on
independent, treatment processes, and control variables. Eight estimated coefﬁcients
differ between negative binomial models and linear regression models in the text. The
eight differences were reﬂective of one independent variable and four control variables.

Relative to the independent variables, housing stability was not associated with
re-incarceration likelihood in the negative binomial models. A statistically signiﬁcant
relationship was observed in the logistic regression models that did not control for the
partial effect of housing and employment instability. Three control variables were found
to be signiﬁcantly related to re-incarceration likelihood in the negative binomial models.
Older individuals were less likely to be re-incarcerated and those convicted for a crime
against persons were 1.25 times more likely to be re-incarcerated relative to those
convicted on drug offenses. Additionally, those who had previously served a prior prison
term were 1.14 times more likely to be re-incarcerated relative to those serving their ﬁrst
prison sentence. The relationship between criminal history (i.e., prior prison admissions)
and re-incarceration likelihood was only apparent in negative binomial models that
controlled for the interactive effect of housing and employment instability.

The remaining control variable of participation in higher education during the

supervision term switched direction when estimated with negative binomial models. The

233

variable was not related to re-incarceration likelihood. The insigniﬁcance of the
coefﬁcient is consistent with the estimates derived from logistic regression models.

Table 28: Veriﬁcation of Full Regressions of Re-Incarceration on Independent,
Treatment Processes, and Control Variables (n=511).

 

 

 

b (SE) IRR b (SE) IRR
Age -.02 (.01)* .98 -.02 (.01)* .98
Whit; -.02 (.12) .98 -.03 (.12) .97
65138 .40 (.13)** 1.50 .40 (.13)** 1.50
HS Grad Plus“ .57 (.16)*** 1.78 .57 (.16)*** 1.78
Divorce/Widowb .22 (.14) 1.24 .23 (.14) 1.25
Mamedb .23 (.18) 1.26 .22 (.18) 1.25
Priorjcommite .15(.12) 1.16 .14 (.12) 1.15
Prior Prison“ .12 (.12) 1.13 .13 (.12)* 1.14
Person Crimec ~22 (~15)* 1-25 .23 (.15)* 1.25
property Crimec -.09 (.16) .91 -.09 (.16) .92
PSafety Crimec -.11 (.18) .90 -.12 (.18) .88
Past TX Services -.01 (.03) .99 -.01 (.03) .99
Sup Intensity -.01 (.00)*** .98 -.01 (.00)*** .98
Grad ofoe -.65 (.25)** .52 -.64 (.25)** .52
TX Groupd .38 (.13)** 1.46 .36 (.13)** 1.44
Hospitaiizede -22 (.17) .80 —.22 (.17) .80
MH Treatments .25 (.17) 1.28 .24 (.17) 1.27
H Edue .03 (.17) 1.03 .03 (.17) 1.02
House Stability -.04 (.03) .96 -.01 (.05) .99
Emp Stabili -.05 (.01)*** .95 -.05 (.01)*** .95
House*Emp -.20 (.21) .82
Months in SATX -04 (.02)* .96 -.04 (.02)* .96
SATX Violations -.03 (.06) .96 -.03 (.06) .97
Absconde .56 (.12)*** 1.75 .56 (.12)*** 1.75
-2LL 281.10 280.89
Model Chi2 319.92*** 320.88***
df 23 24

 

 

 

‘p<.05, ”p<.01, ***p<.OOl; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses; IRR refers to incident rate ratio, which is a standardized coefﬁcient.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Interaction variable refers to individuals who
experienced 4 or more housing moves (median split of sample) and who were employed for less than half
of their supervision term.

234

 

APPENDIX E
SOBEL TEST (1986) OF INDIRECT EFFECTS WITHOUT THE INITIAL PRESENCE
OF A NON-ZERO DIRECT EFFECT BETWEEN INDEPENDENT AND OUTCOME
VARIABLE FROM BARON AND KENNY (1986)

The method for determining the presence of indirect effects in this study was
based upon Baron and Kenny (1986). The method followed a three step process that was
largely shaped by the initial step. This initial step requires that a signiﬁcant (non-zero)
association must be observed between a given independent variable and an outcome
variable of interest. Without the establishment of this initial effect, Baron and Kenny
(1986) suggested that a suspected mediator variable will not be able to affect the
relationship between two variables.

Many have been critical of the initial step of the Baron and Kenny method and
have suggested that the Sobel test (1986) should serve as the initial test for the presence
of an indirect effect instead of being used as a supplement to Baron and Kenny
(Greenwald & Draine, 1998; MacKinnon et al.., 2000; Shrout & Bolger, 2002). The
rationale for this argument is based on the notion that indirect effects may be overlooked
by placing too much emphasis on the initial presence of a statistically signiﬁcant direct
effect of an independent variable on a dependent variable. By implication, this argument
suggests that indirect effects can be observed without the presence of a direct effect.

Table 29 provides a summary table of alternative results for the Sobel test (1986)
of indirect effects. These alternative results were based upon the use of the test at the
outset and irrespective of the establishment of a direct effect between an independent

variable and a dependent variable. The estimated coefﬁcients and their respective

235

standard errors were corrected for comparability between continuous or binary mediators
and outcomes (MacKinnon & Dwyer, 1993).

Table 29: Summary of Adjusted Sobel’s Test of Indirect Effects without Baron and
Kenny (1986) Clause (n=511).

 

 

 

Sobel Test Statistic

Housing Stability 9 Dose 9 Relapse 297*“
Housing Stability 9 Violations 9 Relapse 376*“
Housing Stability 9 Abscond 9 Relapse .66***
Housing Stability 9 Dose 9 Re-Arrest -2.72
Housing Stability 9 Violations 9 Re-Arrest -3.43*
Housing Stability 9 Abscond 9 Re-Arrest 290*"
Housing Stability 9 Dose 9 Re-Incarceration -1.91
Housing Stability 9 Violations 9 Re-Incarceration -.69*
Housing Stability 9 Abscond 9 Re-Incarceration 364*”
Employment Stability 9 Dose 9 Relapse 2.47***
Employment Stability 9 Violations 9 Relapse -l.79***
Employment Stability 9 Abscond 9 Relapse -.65***
Employment Stability 9 Dose 9 Re-Arrest -2.32*
Employment Stability 9 Violations 9 Re-Arrest 1.75"
Employment Stability 9 Abscond 9 Re-Arrest -2.43**
Employment Stability 9 Dose 9 Re-Incarceration -l.75*
Employment Stability 9 Violations 9 Re-Incarceration .65"
Employment Stability 9 Abscond 9 Re-Incarceration -2.82**

 

*p<.05, **p<.01, ***p<.001; NOTE: Sobel Tests have been adjusted for binary mediators or outconres
(MacKinnon & Dwyer, 1993).

The alternative results suggest that treatment processes indirectly affected nearly
all of the relationships between housing stability, employment stability, and correctional
outcome indicators of relapse and recidivism. The only exceptions were observed for
housing stability and recidivism. The level of substance abuse treatment dose received
did not appear to inﬂuence the relationship between housing stability and re-arrest
likelihood or the association between housing stability and re-incarceration likelihood.

The alternative results provide a liberal interpretation of indirect effects and
conﬁrmed the overall results of numerous interrelationships between the reentry
dimensions of housing, employment, and treatment that can shape correctional outcome

indicators of relapse and recidivism. These alternative results differ from the

236

 

conservative estimate of six direct and indirect effects in the text. Given the exploratory
nature of this study and the ongoing debate over the need to establish an initial direct
effect between variables that may be mediated, it was more appropriate to cautiously
interpret the presence of indirect effects. This cautious, and conservative, interpretation

of indirect effects was presented in the text.

237

 

 

APPENDIX F
ALTERNATIVE MODELING OF INTERACTION TERMS FOR HOUSING
STABILITY AND EMPLOYMENT STABILITY

Tables 30 through 32 present alternative modeling of interaction terms for

housing stability and employment stability70. Negative binomial regression was used to
replicate full conditional models with interaction terms. The negative binomial models
were estimated with robust standard errors (see Appendix B for a brief background on the
negative binomial technique and robust standard errors).

The dichotomized interaction term used in the text referred to conditional median
distributions of housing instability (i.e., 4 or more housing moves) and employment
instability (i.e., employed for less than half of their supervision term). Three additional
dummy variable interaction terms were constructed based upon the remaining
combinations of conditional median distributions (Hardy, 1993). The ﬁrst interaction
term (U SH*SE) represented those who experienced housing instability (i.e., 4 or more
housing moves) and employment stability (i.e., employed for more than half of their
supervision term). The second interaction term (SH*USE) represented those who had
stable housing (i.e., 3 or fewer housing moves), but also experienced employment
instability (i.e., employed for less than half of their supervision term). The ﬁnal
interaction term (SH*SE) represented those who had stable housing and stable
employment (i.e., experienced 3 or fewer housing moves and were employed for more

than half of their supervision term). In modeling these 3 alternative interaction terms, the

 

O Caution is needed with the interpretation of these regression models since the models are likely to be
underpowered and may only be able to identify medium or large effect sizes.

238

 

 

combination of housing instability and employment instability used in the text serves as
the reference category.

Table 30 presents the full conditional negative binomial models of treatment
processes on independent, control, and interaction variables. Overall, the initial ﬁndings
presented in the text were replicated. The joint contributions of housing stability and
employment stability were not related to substance abuse treatment dose or absconding.
However, the interaction between the primary independent variables was associated with
treatment program violations.

The ﬁndings conﬁrm that housing and employment instabilities led to higher
treatment pro gram violations relative to those individuals who have relatively stable
housing with or without stable employment. This suggested that housing stability may be
the driving factor that inﬂuences treatment program violations, with stable housing
leading to fewer violations. Those who have unstable housing, but stable employment
have similar levels of program violations compared to those with unstable housing and

unstable employment.

239

Table 30: Summary of Full Regressions of Treatment Processes on Independent and
Control Variables with Alternative Interaction Terms (n=511).

 

 

 

Treatment Dose Treatment Violations Absconding
b (SE)1RR b (SE) IRR b (SE) IRR

Age .00 (.00) 1.00 .00 (.01) 1.00 -.01 (.01) 99*
White“ -.17 (.07) .84** .29 (.13) 134* -.02 (.10) .98
(313133 -.01 (.06) .99 -.05 (.12) .95 .05 (.10) 1.05
HS Grad pm,8 .03 (.09) 1.03 -.21 (.18) .81 -.07 (.15) .93
Divorce/Widowb .12 (.09) 1.13 .01 (.17) 1.01 .08(.14)1.08
Mamedb -.01 (.09) .99 .06 (.21) 1.06 -.08 (.17) .92
pit-o” Commit" -.07 (.07) .93 .15(.14)1.16 .15(.10)1.17
Prior prison" -.10(.06) .90 .01 (.12) 1.01 .27 (.11) 1.31*
permn Crimec -.09 (.07) .91 .31 (.17) 1.36 .12 (.13) 1.12
property Crimec —.10 (.08) .90 .35 (.17) 1.42* .22 (.13) 1.25
PSafety Crimec -.11 (. 10) .90 .15 (.21) 1.16 .05 (.16) 1.06
Past TX Services -.02 (.02) .98 .07 (.03) 1.08" .06 (.02) 1.06”
Sup Intensity -.01 (.00) 1.01*** .00 (.00) 1.00 -.00 (.00) 1.00
Grad ofo“ .51 (.06) 1.66*** -.78 (.15) .45*** -.59 (.12) .55***
TX Groupd .55 (.06) 1.73*** .73 (.14) 2.08*** .21 (.09) 1.23*
Hospitalizede .14 (.06) 1.15* .06 (.12) 1.06 .03 (.10) 1.04
MH Treatmente .05 (.09) 1.05 —.53 (.18) .59** -.00 (.15) 1.00
H Edue .05 (.07) 1.05 -.17 (.18) .84 -.23 (.16) .79
House Stability .06 (.02) 1.06** .15 (.03) 1.16*** .11 (.03) 1.12***
Emp Stability .02 (.01) 1.02** -.03 (.02) .97 -.04 (.01) .96**
USHatSEf -.09 (.12) .91 -.41 (.29) .66 .18 (.22) 1.20
SH*USEf -.11 (.10) .89 -.68 (.20) .51*** -.05 (.16) .95
SH*SEf -.18 (.14) .83 -1.16 (.43) .31*** -.14 (.28) .86
-2LL 1419.67 475.93 370.57
Model Chi2 403.57*** 376.34*** 211.28***
df 23 23 23

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses and followed by a standardized coefﬁcient (IRR); IRR refers to incident rate ratio.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Contrasted against the reference category of
individuals who experienced 4 or more housing moves (unstable housing) and who were employed for less
than half of their supervision term (unstable employment).

Table 31 provides the full conditional negative binomial models of outcome

indicators on independent, control, and interaction variables. Table 32 presents the full

conditional negative binomial models of outcome indicators on independent, treatment

processes, control, and interaction variables. Both sets of models provide similar results
and suggest that the joint contribution of housing stability and employment stability were
not related to relapse or re-arrest likelihood. These models do suggest that the measure of
housing and employment stability interaction used in the text may have masked important
relationships with re-incarceration likelihood.

Tables 31 and 32 suggested that the joint contribution of housing stability and
employment stability on re-incarceration was most prominent on the extremes.
Individuals with relatively stable housing and stable employment were nearly three times
more likely than those with unstable housing and unstable employment to be re-
incarcerated. Additionally, those with unstable housing and stable employment were just
over two times more likely than those with unstable housing and unstable employment to
be re-incarcerated. Those with unstable employment, but stable housing had similar
levels of re-incarceration risk relative to those with unstable housing and unstable
employment. These ﬁndings suggested that employment stability may be the driving
factor that inﬂuenced re-incarceration likelihood. The direct effect of stable employment
on re-incarceration likelihood was negative. However, when combined with levels of

housing stability, the negative effect of stable employment on re-incarceration likelihood

was negated and switches directions7|

 

7] This ﬁnding is unexpected given the previous ﬁndings of this study regarding the effect of employment
stability. It is possible that the change in expected direction was due to a rrrixture of measurement error in

the direct effect of employment stability, measurement error in the interaction terms, the distribution of the
measures, and statistical power considerations that may have led to the error of rejecting a null hypothesis

when it is true (i.e., type 1 error).

241

Table 31: Summary of Full Regressions of Outcome Indicators on Independent and
Control Variables with Alternative Interaction Terms (n=511).

 

 

 

Relapse Re-Arrest Re-Incarceration
b £13) IRR b (SE) IRR b (SE) IRR

Age -.01 (.01) .99 -.02 (.01) 98* -.02 (.01) 98*
Whit; -.26 (.12) .77* -04 (.14) .96 -.06 (.12) .94
GED“ -.03 (.10) .97 .32 (.17) 1.38 .37 (.14) 1,45...
HS Grad Plus“ --20 (.16) .82 .48 (.20) 1.62* .53 (.16) 1.69***
Divorce/Widowb .03 (.15) 1.03 .22 (.18) 1.25 .17 (.15) 1.18
Mamedb -.08 (.18) .92 .05 (.22) 1.06 .13 (.16) 1.34
pn-o, J Commite -.01 (.12) .99 -.10 (.16) .91 .24 (.12) 1.27
Prior prison“ .26 (.10) 1.30** .23 (.16) 1.26 .17 (.12) 1.18
person cﬁmec -.15 (.13) .86 -.02 (.18) .98 .32 (.15) 1.38
property Crimec -.14 (.13) .87 .13 (.18) 1.14 .02 (.16) 1.02
PSafety Crimec -. 17 (.16) .84 -.19 (.23) .82 -04 (.18) .96
Past Tx Services -.02 (.03) .98 -.05 (.04) .95 .00 (.03) 1.00
Sup Intensity .00 (.00) 1.00 -.02 (.00) .98*** -.02 (.00) 98***
Grad ofo“ --22 (.10) .80* -.36 (.19) .69 -93 (.24) .40***
TX Groupd .16(.10)1.17 .14(.14)1.15 .22(.11)1.25*
Hospitalizede .21 (.09) 124* .02 (.18) 1.02 -.17 (.17) .84
MH Treatmente -.13 (.13) .88 -.22 (.26) .80 .15 (.18) 1.17
H Edtf -.23 (. 13) .80 -.18 (.22) .84 -.08 (.18) .92
House Stability .04 (.03) 1.04 .03 (.05) 1.03 .01 (.05) 1.01
Emp Stability -.01 (.01) .98 -.02 (.02) .98 -.12 (.02) .39"...
USH*SEf -.11 (.23) .89 .27 (.32) 1.31 .80 (.33) 2.22**
SH*USEf -.20 (.16) .81 -.19(.26) .83 .16 (.24) 1.17
SH"‘SEf --41 (.27) .66 -.02 (.36) .98 1.09 (.30) 2.97***
-2LL 253.30 303.81 284.10
Model ChiZ ll4.63*** 66.15“” 14599:”:
df 23 23 23

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses and followed by a standardized coefﬁcient (IRR); IRR refers to incident rate ratio.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, c. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Contrasted
against the reference category of the non-instance, f. Contrasted against the reference category of
individuals who experienced 4 or more housing moves (unstable housing) and who were employed for less
than half of their supervision term (unstable employment).

242

Table 32: Summary of Full Regressions of Outcome Indicators on Independent,
Treatment Processes, and Control Variables with Alternative Interaction Terms (n=511).

 

 

 

 

Relapse Re-Arrest Re-Incarceration
b (SE) IRR b (SE) IRR b (SE) IRR

Age -.01 (.01) .99 -.02 (.01) .98 -.01 (.01) 98*
Whit; -27 (.12) .76* -.05 (.14) .95 -.06 (.12) .94
513133 -.01 (.10) .99 .29 (.17) 1.34 .35 (.13) 1.42**
HS Grad Plus“ -.14 (.16) .87 .46 (.20) 158* .54 (.16) 1,71“...
Divorce/Widowb -.03 (.15) .97 .27 (.17) 1.31 .19 (.14) 1.21
Mantedb -09 (.18) .91 .07 (.23) 1.07 .18 (.17) 1.19
Prior J Commit? —.02 (.12) .98 -.11 (.16) .89 .21 (.12) 1.24
prior prisone .28 (.10) 1.32** .16 (.16) 1.17 .06 (.12) 1.07
person cm; -.20 (.13) .82 .00 (.18) 1.00 .26 (.15) 1.29
property Crimec -.19 (.13) .83 .12 (.18) 1.13 -.05 (.15) .95
PSafety Crimec -.17 (. 16) .84 -22 (.22) .80 -.10 (.17) .90
Past Tx Services -.04 (.03) .96 -.03 (.05) .97 -.01 (.03) .99
Sup Intensity -.00 (.00) 1.00 -.01 (.00) .99" -.Ol (.00) .99***
Grad ofo6 -.26 (.11) .77* —.17 (.20) .84 -.63 (.25) .53*
TX Groupd -.09 (.11) .91 .42 (.16) 1.52** .33 (.12) 1.39**
Hospitanzede .19 (.09) 121* .05 (.18) 1.05 -.19 (.17) .82
MH Treatments -.08 (.13) .92 -.30 (.25) .74 .22 (.17) 1.24
H Educ -25 (.13) .78 -.13 (.21) .88 .01 (.17) 1.01
House Stability -.01 (.03) .99 .07 (.06) 1.07 .00 (.05) 1.00
Emp Stability -.02 (.01) .97 -.01 (.02) .99 -.11 (.02) .90***
USHssef .03 (.22) 1.03 .15 (.32) 1.16 .76 (.31) 215*
SH*USEf -.09 (.16) .91 -.26 (.27) .77 .13 (.24) 1.13
SH*SEf ~20 (.26) .91 -.13 (.36) .87 1.00 (.30) 2.71***
Dose of SATX .05 (.01) 1.05*** -.05 (.02) .95** -.04 (.02) 96*
SATX P Violations .12 (.04) 1.16*** -.28 (.09) .75** -03 (.06) .97
Absconde .12 (.12) 1.13 .50 (.17) 1.65** .56 (.12) 1.75**
-2LL 249.61 293.45 277.26
Model Chi2 142.67*** 111.13*** 348.19"...
df 26 26 26

 

*p<.05, **p<.01, ***p<.001; NOTE: Unstandardized coefﬁcients are presented with robust standard errors
in parentheses and followed by a standardized coefﬁcient (IRR); IRR refers to incident rate ratio.

a. Contrasted against reference category of less than a high school diploma (and no GED attainment), b.
Contrasted against reference category of single relationship status, 0. Contrasted against reference category
of drug crime conviction, (1. Contrasted against the reference category of the control group, e. Conﬁasted
against the reference category of the non-instance, f. Contrasted against the reference category of
individuals who experienced 4 or more housing moves (unstable housing) and who were employed for less
than half of their supervision term (unstable employment).

243

BIBLIOGRAPHY

Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency.
Criminology, 30, 47-87.

Agresti, A. (2002). Categorical data analysis (2nd Ed.). Hoboken, NJ: John Wiley and
Sons, Inc.

Allen, M. P. (1997). Understanding regression analysis. New York, NY: Plenum Press.

Andrews, D. A. & Bonta, J. (2003). The psychology of criminal conduct. Cincinnati, OH:
Anderson Corporation.

Andrews, D., I. Zinger, R. Hoge, J. Bonta, P. Gendreau and F. Cullen (1990). Does
Correction Treatment Work? A Clinically Relevant and Psychologically Informed
Meta-Analysis. Criminology 28(3): 369-404.

Anglin, D. M. & Hser, Y. I. (1992). Drug abuse treatment. In R. R. Watson (Ed.), Drug
abuse treatment (pp. 1-36). Totowa, NJ: Humana Press.

Anglin, M. D. & Hser, Y. I. (1990). Treatment of drug abuse. In M. Tonry & J. O.
Wilson (Eds), Drugs and Crime. Chicago, IL: University of Chicago Press.

Antonowicz D. H. and R. R. Ross (1994). Essential Components of Successful
Rehabilitation Programs for Offenders. International Journal of Offender Therapy
and Comparative Criminology 38: 97-104.

A08, 8., Miller, M, & Drake, E. (2006). Evidence-based adult corrections programs:
What works and what does not. Olympia, WA: Washington State Institute for
Public Policy.

Aos, S., P. Phipps, R. Bamoski, and R. Lieb. (1999). The Comparative Costs and Beneﬁts

of Programs to Reduce Crime: A Review of National Research Findings with
Implications for Washington State. Olympia, WA: Washington State Institute for
Public Policy

Baron, R. M. & Kenny, D. A. (1986). The moderator-mediator variable distinction in
social psychological research: Conceptual, strategic, and statistical considerations.
Journal of Personality and Social Psychology, 51(6), 1 173-1 182.

Bean, P. & Nerrritz, T. (2004). Drug treatment; what works? An overview. In P. Bean &
T. Nemitz (Eds), Drug Treatment: What Works? (pp. 1-18). New York, NY:
Routledge.

Beck, A. J. (2000). State and federal prisoners returning to the community:
Findings from the Bureau of Justice Statistics. Paper presented at the First

244

Reentry Courts Initiative Cluster Meeting, Washington, DC. Retrieved 15
November 2008 from http://www.oip.usdoi.gov/bis/pub/pdf/sfprc.pdf

Beck, A. J. & Shipley, B. E. (1989). Recidivism of Prisoners Released in 1983.
Bureau of Justice Statistics Special Report (N CJ 116261). Washington, DC: US
Department of Justice, Bureau of Justice Statistics.

Berg, M. T. & Huebner, B. M. (2009). Reentry and the ties that bind: An examination of
social ties, employment, and recidivism. Manuscript submitted for publication.

Berk, R. A. (2004). Regression analysis: A constructive critique. Thousand Oaks, CA:
Sage Publications.

Berk, R. A., Lenihan, K. J., & Rossi, P. H. (1980). Crime and poverty: Some
experimental evidence from ex-offenders. American Sociological Review, 45,
766-786.

Bernard, T. J. (1984). Control criticisms of strain theories: An assessment of theoretical
and empirical adequacy. Journal of Research in Crime and Delinquency, 21(4),
353-372.

Berry, W. D. & Feldman, S. (1985). Multiple regression in practice. Newbury Park, CA:
Sage Publications.

Bloom, D. (2006). Employment-focused programs for err-prisoners: What have we
learned, what are we learning, and where should we go from here? Ann Arbor,
MI: National Poverty Center, Gerald R. Ford School of Public Policy.

Bloom, D., Redcross, C., Zweig, J ., & Azurdia, G. (2007). Transitional jobs for ex-
prisoners: Early impacts from a random assignment evaluation of the center for
employment opportunities (CEO) prisoner reentry program. New York, NY:
Manpower Demonstration Research Corporation (MDRC).

Blumer, H. (1969). Symbolic interactionism: Perspective and method. Berkeley, CA:
University of California Press.

Blumstein, A., & Beck, A. J. (1999). Population growth in US. prisons, 1980—1996. In
M. Tonry & J. Petersilia (Eds.), Prisons (pp. 17-62). Chicago, IL: University of
Chicago Press.

Boruch, R. F. (1997). Randomized experiments for planning and evaluation: A practical
guide. Thousand Oaks, CA: Sage Publications.

Bottoms, A., Shapland, J ., Costello, A., Holmes, D., & Muir, G. (2004). Towards

desistance: Theoretical underpinnings for an empirical study. Howard Journal of
Criminal Justice, 43(4), 368-389.

245

 

Braman, D. & Wood, J. (2003). From one generation to the next: How criminal sanctions
are reshaping family life in urban America. In J. Travis & M. Waul (Eds.),
Prisoners Once Removed: The Impact of Incarceration and Reentry on Children,
Families, and Communities (pp. 157-188). Washington, DC: The Urban Institute.

Brennan, T., Dieterich, W., & Ehret, B. (2009). Evaluating the predictive validity of the
COMPAS risk and needs assessment system. Criminal Justice and Behavior,
36(1), 21-40.

Brennan, T. & Oliver, W. L. (2000). Evaluation of reliability and validity of COMPAS
scales: National aggregate sample. Traverse City, MI: Northpointe Institute for
Public Management.

Brooks, L. E., Visher, C. A., & Naser, R. L. (2006). Community residents ' perceptions of
prisoner reentry in selected Cleveland neighborhoods. Washington, DC: Urban
Institute, Justice Policy Center.

Brown, B. S., O’Brady, K. E., Battjes, D., Farrell, E. V., Smith, N. P., & Nurco, D. N.
(2001). Effectiveness of a stand—alone aftercare program for drug-involved

offenders. Journal of Substance Abuse Treatment, 21, 185-192

Brown, B. S., Wienckowski, L. A., & Stolz, S. B. (2004). Behavior modiﬁcation:
Perspective on a current issue. In P. C. Kratcoski (Ed.), Correctional Counseling
and Treatment (5th Ed.) (pp. 367-404). Long Grove, IL: Waveland Press, Inc.

Burnett, R. & Maruna, S. (2004). So ‘Prison Works’, does it? The criminal careers of 130
men released from prison under home secretary, Michael Howard. Howard
Journal of Criminal Justice, 43(4), 390-404.

Bursik, R. J. & Grasmick, H. G. (1993). Neighborhoods and crime: The dimensions of
eﬁ'ective community control. Lanharn, MD: Lexington Books.

Butzin, C. A., Martin, s. S., & Inciardi, J. A. (2002). Evaluating component effects of a
prison-based treatment continuum. Journal of Substance Abuse Treatment, 22, 63-
69.

Byme, J. M. (2009). What works in prisoner reentry: A review of evidence: Hearings
before the Subcommittee on Commerce, Justice, Science, and Related Agencies of
the House Committee on Appropriations, 111th Cong., (chaired by Alan
Mollohan).

Byme, J. M. (2008). The social ecology of community corrections — understanding the
link between individual and community change. Criminology and Public Policy,
7(2), 263-274.

246

Byme, J. (2004). Reentry — the emperors new clothes. Federal Probation, 68, 1-2.

Byme, J. M. & Taxman, F. S. (2005). Crime (control) is a choice: Divergent perspectives
on the role of treatment in the adult corrections system. Criminology and Public
Policy, 4(2), 291-310.

Cameron, A. C. & Trivedi, P. K. (1998). Regression analysis of count data. New York,
NY: Cambridge University Press.

Carver, J. A. (2004). Drug testing: A necessary prerequisite for treatment and for crime
control. In P. Bean & T. Nemitz (Eds), Drug Treatment: What Works? (pp. 142-
177). New York, NY: Routledge.

Chambliss, W. J. & Nagasawa, R. H. (1969). On the validity of ofﬁcial statistic: A
comparative study of white, black, and Japanese high-school boys. Journal of
Research in Crime and Delinquency, 6, 71-77.

Chanhatasilpa, C ., MacKenzie, D., & Hickman, L. (2000). The effectiveness of
community-based programs for chemically dependent offenders: A review and
assessment of the research. Journal of Substance Abuse Treatment, 19, 383-393.

Checinski, K. & Ghodse, H. (2004). Types of treatment for types of patients. In P. Bean
& T. Nemitz (Eds), Drug Treatment: What Works? (pp. 31-57). New York, NY:

Routledge.

Cheliotis, L. K. (2006). Demystifying risk management: A process evaluation of the
prisoners’ home leave scheme in Greece. Criminology and Criminal Justice, 6(2),
163-195.

Clear, T. R. (2007). Imprisoning communities: How mass incarceration makes
disadvantaged neighborhoods worse. New York, NY: Oxford University Press.

Clear, T. & Latessa, E. (1993). Intensive supervision: Surveillance versus treatment.
Justice Quarterly, 10(3), 441-459.

Cloward, R. A. (1959). Illegitimate means, anomie, and deviant behavior. American
Sociological Review, 24, 164-176.

Cloward, R. A. & Ohlin, L. (1960). Delinquency and opportunity: A theory of delinquent
gangs. New York, NY: The Free Press.

Cohen, A. K. (1955). Delinquent boys: The culture of the gang. New York, NY: The Free
Press.

Cohen, J. (1988). Statistical power for the behavioral sciences (2nd Ed.). Hillsdale, NJ:
Lawrence Erlbaum Associates, Inc.

247

 

Cohen, L. E. & F elson, M. (1979). Social change and crime rate trends: A routine activity
approach. American Sociological Review, 44, 588-608.

Coleman, J. (1988). Social capital and the creation of human capital. American Journal of
Sociology, 94, 95-120.

Compton, W., Cottler, L., Jacobs, J ., Ben-Abdallah, A., & Spitznagel, E. (2003). The role
of psychiatric disorders in predicting drug dependence treatment outcomes.
American Journal of Psychiatry, 160, 890-895.

Campbell, D. T. & Stanley, J. C. (1966). Experimental and quasi-experimental designs
for research. Boston, MA: Houghton Mifﬂin.

Corbett, R. P. (2008). A promising future for community corrections. Criminology and
Public Policy, 7(2), 303-310.

Corsaro, N., & McGarrell, E. F. (2009). Testing a promising homicide reduction strategy:
Re-assessing the impact of Indianapolis “pulling levers” intervention. Journal of
Experimental Criminology, 5, 63-82.

Cullen, F. T. (1988). Were Cloward and Ohlin strain theorists? Journal of Research in
Crime and Delinquency, 25(3), 214-241.

Cullen F. T. & Gendreau, P. (2000). Assessing correctional rehabilitation: Policy,
practice, and prospects. In J. Homey (Ed.), Policies, Processes, and Decisions of
the Criminal Justice System (pp. 109-175). Washington, DC: National Institute of
Justice.

Decker, S. H. (Ed.). (2003). Policing gangs and youth violence. Belmont, CA:
Thompsom Wadsworth.

DiIulio, J. J. & Piehl, A. M. (1991). Does prison pay? The stormy national debate over
the cost-effectiveness of imprisonment. The Brookings Review, 9(4), 28-35.

Dishion, T. J ., McCord, J ., & Poulin, F. (1999). When interventions harm: Peer groups
and problem behavior. American Psychologist, 54, 755-764.

Donaldson, L. (1999). The normal science of structural contingency theory. In s. Clegg
& C. Hardy (Eds), Studying Organization: Theory and Method (pp. 51-70).
London, UK: Sage Publications.

Dowden, C., Antonowicz, D., & Andrews, D. A. (2003). The effectiveness of relapse
prevention with offenders: A meta-analysis. International Journal of Oﬂender

Therapy and Comparative Criminology, 4 7(5), 516-528.

Eddy, J. M. & Reed, J. B. (2003). The adolescent children of incarcerated parents. In J.

248

Travis & M. Waul (Eds), Prisoners Once Removed: The Impact of Incarceration
and Reentry on Children, Families, and Communities (pp. 233-258). Washington,
DC: The Urban Institute.

Eicker, F. (1967). Limit theorems for regression with unequal and dependent regressors.
In L. LeCam & J. Neyman (Eds), Fifth Berkeley Symposium on Mathematical
Statistics and Probability (pp. 59-82). Berkeley, CA: University of California
Press.

F arabee, D. (2005). Rethinking rehabilitation: Why can 't we reform our criminals?
Washington, DC: American Enterprise Institute Press.

Farrall, S. & Bowling, B. (1999). Structuration, human development and desistance from
crime. British Journal of Criminology, 39(2), 253-268.

Farrington, D.P. (2003a). A Short History of Randorrrized Experiments in Criminology:
A Meager Feast. Evaluation Review, 27, 218-227.

Farrington, D. P. (2003b). Methodological quality standards for evaluation. Annals of the
American Academy of Political and Social Science, 58 7, 49-68.

F arrington, D. P., Gallagher, B., Morley, L., St. Ledger, R. J., & West, D. J. (1986).
Unemployment, school leaving, and crime. British Journal of Criminology, 26,
335-356.

Fass, T. L., Heilbrun, K., Dematteo, D., & Fretz, R. (2008). The LSI-R and the
COMPAS: Validation data on two risk-needs tools. Criminal Justice and
Behavior, 35(9), 1095-1108.

Feeley, M. M. & Simon, J. (1992). The new penology: Notes on the emerging strategy of
corrections and its implications. Criminology, 30(4), 449-474.

Finn, P. (1998). Texas 's Project RIO (re-integration of oﬂenders) (NCJ 193427).
Washington, DC: National Institute of Justice.

Foucault, M. (1977). Discipline and punish: The birth of the prison. New York, NY:
Vintage.

Fox, J. (1991). Regression diagnostics. Newbury Park, CA: Sage Publications.

Gaes, G., Flanagan, T., Motiuk, L., & Stewart, L. (1999). Adult correctional treatment. In
M. Tonry & J. Petersilia (Eds), Prisons (pp. 361-426). Chicago, IL: University of
Chicago Press.

Gibbons, J. J. & Katzenbach, N. B. (2006). Conﬁonting conﬁnement: A report of the

commission on safety and abuse in America 's prisons. New York, NY: Vera
Institute of Justice.

249

Giordano, P. C., Cemkovich, S. A., & Holland, D. D. (2003). Changes in friendship
relations over the life course: Implications for desistance from crime.
Criminology, 41(2), 293-328.

Giordano, P. C., Cemkovich, S. A., & Rudolph, J. L. (2002). Gender, crime, and
desistance: Toward a theory of cognitive transformation. American Journal of
Sociology, 107(4), 990-1064.

Glaser, D. (1964). The effectiveness of a prison and parole system. Indianapolis, IN:
Bobbs-Merrill

Glaze, L. E. (2003). Probation and parole in the United States, 2002. Bureau of
Justice Statistics Bulletin (NCJ 201135). Washington DC: US Department of
Justice, Bureau of Justice Statistics.

Glaze, L. E. & Palla, S. (2005). Probation and parole in the United States, 2004.
Washington, DC: Bureau of Justice Statistics.

Gofﬁnan, E. (1963). Stigma: Notes on the management of spoiled identity. New York,
NY: Simon and Schuster.

Goffrnan, E. (1959). The presentation of self in everyday life. New York, NY:
Doubleday.

Gottfredson, S. D. & Gottfredson, D. M. (1994). Behavioral prediction and the problem
of incapacitation. Criminology, 32(3), 441-474.

Gove, W., Hughes, M., & Geerken, M. (1985). Are uniform crime reports a valid
indicator of index crimes? An afﬁrmative answer with minor qualiﬁcations.
Criminology, 23, 451-501.

Greenwald, A. G. & Draine, S. C. (1998). Distinguishing unconscious from conscious
cognition -— reasonable assumptions and replicable ﬁndings: Reply to Merikle and
Reingold (1998) and Dosher (1998). Journal of Experimental Psychology:
General, 127(3), 320-324.

Grommon, E. & Devitt, C. (2003). Sharing criminal history record information: The
Interstate Identiﬁcation Index. Illinois Criminal Justice Information Authority
Research Bulletin (V 01. 2, No. 8). Chicago, IL: Illinois Criminal Justice
Information Authority.

Haggard, U-., Gumpert, C. H., & Grann, M. (2001). Against all odds: A qualitative

follow-up study of hi gh-risk violent offenders who were not reconvicted. Journal
of Interpersonal Violence, 1 6(10), 1048- l 065.

250

Hairston, C. F. (2003). Prisoners and their families: Parenting issues during incarceration.
In J. Travis & M. Waul (Eds), Prisoners once removed: The impact of
incarceration and reentry on children, families, and communities (pp. 259-282).
Washington, DC: The Urban Institute.

Hardin, J. & Hilbe, J. (2001). Generalized linear models and extensions. College Station,
TX: Stata Press.

Harding, D. J. (2003). Jean Valj ean’s dilemma: The management of ex-convict identity in
the search for employment. Deviant Behavior, 24(6), 571-595.

Hardy, M. A. (1993). Regression with dummy variables. Newbury Park, CA: Sage
Publications.

Harris, M. K. (2005). In search of common ground: The importance of theoretical
orientations in criminology and criminal justice. Criminology and Public Policy,
4(2), 311-328.

Harrison, L. D. (2001). The revolving prison door for drug-involved offenders:
Challenges and opportunities. Crime and Delinquency, 47(3), 462-485.

Hechter, M. & Kanazawa, S. (1997). Sociological rational choice theory. Annual Review
of Sociology, 23, 191-214.

Helfgott, J. (1997). Ex-offender needs versus community opportunity in Seattle,
Washington. Federal Probation, 10, 64-79.

Hepburn, J. R. & Grifﬁn, M. L. (2004). The effect of social bonds on successful
adjustment to probation: An event history analysis. Criminal Justice Review,

29(1), 46-75.

Hilbe, J. M. (2007). Negative binomial regression. New York, NY: Cambridge
University Press.

Hiller, M. L., Knight, K., & Simpson, D. D. (1999). Prison-based substance abuse
treatment: Residential aftercare and recidivism. Addiction, 94(6), 833-842.

Hindelang, M. J ., Hirschi, T., & Weis, J. G. (Eds) (1981). Measuring delinquency.
Newbury Park, CA: Sage.

Hirschi, T. (1969). Causes of delinquency. Berkeley, CA: University of California Press.

Holzer, H. J. (1996). What employers want: Job prospects for less-educated workers.
New York, NY: Russell Sage.

Holzer, H. J ., Raphael, S., & Stoll, M. A. (2004). Will employers hire former offenders?:

251

Employer preferences, background checks, and their determinants. In M. Pattillo,
D. Weiman, & B. Western (Eds), Imprisoning America: The Social Eﬂects of
Mass Incarcertion (pp. 205-243). New York, NY: Russell Sage Foundation.

Homey, J. D., Osgood, D. W., & Marshall, I. H. (1995). Criminal careers in the short-
terrn: Intra-individual variability in crime and its relation to local life
circumstances. American Sociological Review, 60(5), 655-673.

Hser, Y., Stark, M. E., Paredes, A., Huang, D., Anglin, M. D., & Rawson, R. (2006). A
12-year follow-up of a treated cocaine dependent sample. Journal of Substance
Abuse Treatment, 30(3), 219-226.

Huber, P. J. (1967). The behavior of maximum likelihood estimates under nonstandard
conditions. In L. LeCam & J. Neyman (Eds), Fifth Berkeley Symposium on
Mathematical Statistics and Probability (pp. 221-233). Berkeley, CA: University
of Califonria Press.

Huebner, B. M. (2005). The effect of incarceration on marriage and work over the life
course. Justice Quarterly, 22, 281-303.

Hughes, T. & Wilson, D. (2003). Reentry trends in the United States. Washington, DC:
Bureau of Justice Statistics. Available from
http://www.oip.usdoi.gov/bis/reentry/reentryhtm

Hughes, T. A., Wilson, D. J., & Beck, A. J. (2001). Trends in state parole, 1990-2000.
Bureau of Justice Statistics, Special Report (N CJ 184735). Washington, DC:
National Institute of Justice, Ofﬁce of Justice Programs, Bureau of Justice
Statistics.

Inciardi, J. A., Martin, S. S., & Butzin, C. T. (2004). F ive—year outcomes of therapeutic
community treatment of drug-involved offenders after release from prison. Crime
and Delinquency, 50, 88-107.

Irwin, J. (1970). The Felon. Englewood Cliffs, NJ: Prentice Hall.

J accard, J. (2001). Interaction eﬂects in logistic regression. Thousand Oaks, CA: Sage
Publications.

J accard, J. Tunisi, R., & Wan, C. K. (1990). Interaction effects in multiple regression.
Newbury Park, CA: Sage Publications.

Jacobs, E. & Western, B. (2007). Report on the evaluation of the comalert prisoner
reentry program. Unpublished manuscript.

Justice Research Associates. (2005). Genesee parolee reentry program assessment: First
year progress report. Williarnston, MI: Justice Research Associates.

252

 

 

Karburg, J. C. & James, D. J. (2005). Substance dependence, abuse, and treatment of jail
inmates, 2002. Bureau of Justice Statistics, Special Report (NCJ 209588).
Washington, DC: Us Department of Justice, Ofﬁce of Justice Programs, Bureau
of Justice Statistics.

Katz, C. M. (2001). The establishment of a police gang unit: An examination of
organizational and environmental factors. Criminology, 39(1), 37-73.

Katz, C. M., Maguire, E. R., & Roncek, D. W. (2002). The creation of specialized police
gang units: A macro-level analysis of contingency, social threat and resource

dependency explanations. Policing: An International Journal of Police Strategies
and Management, 25(3), 472-506.

Kline, R. B. (2005). Principles and practices of structural equation modeling (2nd Ed.).
New York, NY: The Guildford Press.

Kling, J. R. (2006). Incarceration length, employment, and earnings. American Economic
Review, 96, 863-876.

Komhauser, R. R. (1978). Social sources of delinquency. Chicago, IL: University of
Chicago Press.

Landenberger, N. A. & Lipsey, M. (2006). The positive effects of cognitive-behavioral
programs for offenders: A meta-analysis of factors associated with effective
treatment. Journal of Experimental Criminology, 1(4), 43 5-450.

Langan, P. A. & Levin, D. J. (2002). Recidivism of prisoners released in 1994. Bureau of
Justice Statistics, Special Report (NCJ 193427). Washington, DC: Us Department
of Justice, Ofﬁce of Justice Programs, Bureau of Justice Statistics.

Latessa, E. J. (2004). The challenge of change: Correctional programs and evidence-
based practices. Criminology and Public Policy, 3(4), 547-560.

Lattimore, P. K., Brumbaugh, S., Visher, C., Linquist, C. H., Winterﬁeld, L., Salas, M.,
& Zweig, J. (2004). National portrait of the serious and violent oﬂender reentry
initiative (RTI Report 8805). Research Triangle Park, NC: Research Triangle
Institute.

Lattimore, P. K. & Visher, C. A. (2009). Assessment of the serious and violent oﬂender
reentry initiative: Hearings before the Subcommittee on Commerce, Justice,

Science, and Related Agencies of the House Committee on Appropriations, 111th
Cong., (chaired by Alan Mollohan).

Lattimore, P. K., Visher, C. A., & Lindquist, C. (2005). Implementation of prisoner
reentry programs: Findings from the serious and violent offender reentry initiative

253

 

 

multi-site evaluation. Justice Research and Policy, 7(2), 87-109.

Laub, J. H., Nagin, D. S., & Sampson, R. J. (1998). Trajectories of change in criminal
offending: Good marriages and the desistance process. American Sociological
Review, 63(2): 225-238.

Laub, J. H. & Sampson, R. J. (2003). Shared beginnings, divergent lives: Delinquent
boys to age 70. Cambridge, MA: Harvard University Press.

Laub, J. H. & Sampson, R. J. (2001). Understanding desistance from crime. In M. Tonry
(Ed.), Crime and justice: A review of research (vol. 28, pp. 1-69). Chicago, IL:
University of Chicago Press.

Laub, J. & Sampson, R. (1993). Turning points in the life course: Why change matters to
the study of crime. Criminology, 31(3), 301-326.

LaVigne, N. & Cowan, J. (2005). Mapping prisoner reentry: An action research
guidebook. Washington, DC: Urban Institute.

LaVigne, N. & Kachnowski, V. (2003). A portrait of prisoner reentry in Maryland.
Washington, DC: The Urban Institute.

LaVigne, N. G., Mamalian, C. A., Travis, J., & Visher, C. (2003). A portrait of
prisoner reentry in Illinois. Washington, DC: Urban Institute, Justice Policy
Center.

LaVigne, N. G., Visher, C., & Castro, J. (2004). Chicago prisoners ’ experiences
returning home. Washington, DC: The Urban Institute.

Lazowski, L. E. & Miller, F. G. (1998). Efﬁcacy of the substance abuse subtle screening
inventory-3 (SASSI-3) in identifying substance dependence disorders in clinical
settings. Journal of Personality Assessment, 71(1), 1 14-128.

Levitt, S. D. (2004). Understanding why crime fell in the 1990s: Four factors that explain
the decline and six that do not. Journal of Economic Perspectives, 18(1), 163-190.

Lipsey, M. W., Chapman, G. L., & Landenberger, N. A. (2001). Research ﬁndings from
prevention and intervention studies: Cogrritive-behavioral programs for offenders.
Annals of the American Academy of Political and Social Science, 578, 144-157.

Lipsey, M. W. & Cordray, D. S. (2000). Evaluation methods for social intervention.
Annual Review of Psychology, 51, 345-375.

Lipton, D., Martinson, R., & Wilks, J. (1975). The eﬂectiveness of correctional

treatment: A survey of treatment evaluation studies. New York, NY: Praeger
Publishers.

254

 

Listwan, S. J ., Cullen, F. T., & Latessa, E. J. (2006). How to prevent prisoner re-entry
programs from failing: Insights from evidence-based corrections. Federal
Probation, 70(3), 19-25.

Loeber, R. & LeBlanc, M. (1990). Towards a developmental criminology. In M. Tonry &
N. Morris (Eds), Crime and justice: A review of research (vol. 12, pp. 375-473).
Chicago, IL: University of Chicago Press.

Long, J. S. (1997). Regression models for categorical and limited dependent variables.
Thousand Oaks, CA: Sage Publications.

Long, J. S. & Ervin, L. H. (2000). Using heteroscedasticity consistent standard errors in
the linear regression model. American Statistician, 54, 217-224.

Longshore, D., Prendergast, M. L., & Farabee, D. (2004). Coerced treatment for drug-
using criminal offenders. In P. Bean & T. Nemitz (Eds), Drug Treatment: What
Works? (pp. 110-122). New York, NY: Routledge.

Lowenkamp, C. T. & Latessa, E. J. (2005). Increasing the effectiveness of correctional
programming through the risk principle: Identifying offenders for residential
placement. Criminology and Public Policy, 4(2), 263-290.

Lowenkamp, C. T. & Latessa, E. J. (2002). Evaluation of Ohio ’3 community based
correctional facilities and halfway house programs. Cincinnati, OH: University of

Cincinnati, Division of Criminal Justice, Center for Criminal Justice Research.

Lurigio, A. J. (2005). Taking stock of community corrections programs. Criminology and
Criminal Justice, 4(2), 259-262.

Lynch, J. P. (2006). Prisoner reentry: Beyond program evaluation. Criminology and
Public Policy, 5(2), 401-412.

Lynch, J. P. & Sabol, W. J. (2001). Prisoner reentry in perspective. Washington, DC:
Urban Institute, Justice Policy Center.

MacKenzie, D. L. (2006). What works in corrections: Reducing the criminal activities of
oﬂenders and delinquents. New York, NY: Cambridge University Press.

MacKenzie, D. L. (2005). The importance of using scientiﬁc evidence to make decisions
about correctional programming. Criminology and Public Policy, 4(2), 249-258.

MacKenzie, D. L. (2000). Evidence-based corrections: Identifying what works. Crime
and Delinquency, 46, 457-471.

MacKenzie, D. L. & Brame, R. (2001). Community supervision, prosocial activities, and

255

f

recidivism. Justice Quarterly, 18, 429-448.

MacKinnon, D. P. & Dwyer, J. H. (1993). Estimating mediated effects in prevention
studies. Evaluation Review, 17, 144-158.

MacKinnon, D. P., Krull, J. L., & Lockwood, C. M. (2000). Equivalence of the
mediation, confounding, and suppression effect. Prevention Science, 1(4), 173-
1 81 .

Mallik-Kane, K. & Visher, C. A. (2008). Health and prisoner reentry: How physical,
mental, and substance abuse conditions shape the process of reintegration.
Washington, DC: The Urban Institute, Justice Policy Center.

Maltz, M. (1984). Recidivism. Orlando, FL: Academic Press.

Marlowe, D. B. (2006). When “what works” never did: Dodging the “scarlet M” in
correctional rehabilitation. Criminology and Public Policy, 5(2), 339-346.

Martin, S. S., Butzin, C. A., Saum, C. A., & Inciardi, J. A. (1999). Three-year outcomes
of therapeutic community treatment for drug-involved offenders in Delaware:
From prison to work release to aftercare. The Prison Journal, 79, 294-320.

Martinson, R. (1979). New ﬁndings, new views: A note of caution regarding sentencing
reform. Hoﬂstra Law Review, 7(2), 242-258.

Martinson, R. (1974). What works? Questions and answers about prison reform. Public
Interest, 10, 22-54.

Maruna, S. (2001). Making good: How ex-oﬂenders reform and reclaim their lives.
Washington, DC: American Psychological Association.

Maruna, S., Immarigeon, R., & LeBel, T. P. (2004a). Ex-offender reintegration: Theory
and practice. In S. Maruna & R. Immarigeon (Eds), After crime and punishment:
Pathways to offender reintegration (pp. 3-26). Portland, OR: Willan Publishing.

Maruna, S., LeBel, T. P., Naples, N., & Naples, M. (2004b). Pygrnalion in the
reintegration process: Desistance from crime through the looking glass.
Psychology, Crime, and Law, 10(3), 271-281.

Marvell, T. B. (2009). Prison population and crime. Paper presented at the DeVoe Moore
Center Symposium on the Economics of Crime at Florida State University,

Tallahassee, Florida.

Marvell, T. B. & Moody, C. (1994). Prison population growth and crime reduction.
Journal of Quantitative Criminology, 10(2), 109-140.

256

Maslow, A. H. (1943). A theory of human motivation. Psychological Review, 50(4), 370-
396.

Matza, D. (1964). Delinquency and drift. New York, NY: John Wiley.

McCord, J. (2003). Cures that harm: Unanticipated outcomes of crime prevention
programs. Annals of the American Academy of Political and Social Science, 58 7,
16-30.

McCleary, R. (1978). Dangerous men: The sociology of parole. Beverly Hills, CA: Sage.
McCleary, R. (1977). How parole ofﬁcers use records. Social Problems, 24(5), 576-589.

McCleary, R. (1975). How structural variables constrain the parole ofﬁcer’s use of
discretionary powers. Social Problems, 23(2), 209-225.

McGarrell, E. F., Giacomazzi, A. L., & Thurman, Q. C. (1999). Reducing disorder, fear,
and crime in public housing: A case study of place-speciﬁc crime prevention.
Justice Research and Policy, 1 (2), 61-87.

McGarrell, E. F ., Hipple, N., & Banks, D. (2003). Community meetings as a tool in
inmate reentry. Justice Research and Policy, 5 (2), 5-32.

McGarrell, E. F., Hipple, N. K., Corsaro, N., Bynum, T. S., Perez, H., Zimmerman, C.
A., & Garmo, M. (2009). Project safe neighborhoods — a national program to
reduce gun crime: Final project report (NCJ 226686). East Lansing, MI:
Michigan State University, School of Criminal Justice.

McGarrell, E. F . & Zimmerman, C. (2005). Reentry: A new lens for a persistent problem.
International Journal of Comparative and Applied Criminal Justice, 29(1), 3-4.

McGarrell, E. F ., Zimmerman, C., Hipple, N. K., Corsaro, N. & Perez, H. (2005). The
roles of the police in the offender reentry process. International Journal of
Comparative and Applied Criminal Justice, 29(1), 53-78.

McNeil], F. (2006). A desistance paradigm for offender management. Criminology and
Criminal Justice, 6(1), 39-62.

Mears, D. P., Wang, X., Hay, C., & Bales, W. D. (2008). Social ecology and recidivism:
Implications for prisoner reentry. Criminology, 46(2), 301-340.

Merton, R. K. (1968). Social theory and social structure. Glencoe, IL: The Free Press.

Metraux, S. & Culhane, D. P. (2006). Recent incarceration history among a sheltered
homeless population. Crime and Delinquency, 52, 504-517.

Metraux, S. & Culhane, D. P. (2004). Homeless shelter use and reincarceration following

257

prison release. Criminology and Public Policy, 3(2), 139-160.

Miller, W. R. & Rollnick, S. (1991). Motivational interviewing: Preparing people to
change addictive behavior. New York, NY: Guilford Press.

Miller, W. R. & Tonigan, J. S. (1996). ‘Assessing drinkers’ motivations for change: The
stages of change readiness and treatment eagerness scale (SOCRATES).
Psychology of addictive behaviors, 10(2), 81-89.

Mobley, A. (2005). From “weeds” to “seeds”: A commentary. International Journal of
Comparative and Applied Criminal Justice, 29(1), 97-107.

Morris, C. W. (Ed.). (1962). Mind, self and society: From the standpoint of a social
behavioralist. Chicago, IL: University of Chicago Press.

Multi-Site Evaluation of SVORI. (2008). The multi-site evaluation of the serious and
violent offender reentry initiative. Retrieved 14 July 2008 from http://www.svori-
evaluationorgl

Mumola, C. J. (1999). Substance abuse and treatment, state and federal prisoners, 1997.
Bureau of Justice Statistics, Special Report (NCJ 172871). Washington, DC: US
Department of Justice, Ofﬁce of Justice Programs, Bureau of Justice Statistics.

Mumola, C. J. & Karberg, J. C. (2006). Drug use and dependence, state and federal
prisoners, 2004. Bureau of Justice Statistics, Special Report (NCJ 213530).
Washington, DC: US Department of Justice, Ofﬁce of Justice Programs, Bureau
of Justice Statistics.

Naser, R. L. & Visher, C. A. (2006). Family members’ experiences with incarceration
and reentry. Western Criminology Review, 7(2), 20-31.

National Institute of Corrections. (2008). Transition from prison to community initiative.
Retrieved 04 April 2008 from httpzl/nicicgov/TPCModel

National Research Council. (2008). Parole, desistance from crime, and community
integration. Committee on community supervision and desistance from crime,
committee on law and justice, division of behavioral and social sciences and
education. Washington, DC: National Academy Press.

Nelson, M., Deess, P., & Allen, C. (1999). The ﬁrst month out: Post incarceration
experiences in New York City. New York, NY: Vera Institute of Justice.

Newburn, T. & Jones, T. (2005). Symbolic politics and penal populism: The long shadow
of Willie Horton. Crime Media Culture, 1(1), 72-87.

Onifade, E., Grommon, E., Campbell, C., & Davidson, W. (2008). Evaluation design and
substance abuse treatment effectiveness with offenders: A meta-analysis.

258

Manuscript submitted for publication.

Pager, D. (2007). Marked: Race, crime, and ﬁnding work in an era of mass
incarceration. Chicago, IL: University of Chicago Press.

Pager, D. (2003). The mark of a criminal record. American Journal of Sociology, 108,
93 7-975.

Palmer, T. (1995). Programmatic and nonprogrammatic aspects of successful
intervention: New directions for research. Crime and Delinquency, 41, 101-131.

Palmer, T. (1994). A profile of correctional effectiveness and new directions for research.
Albany, NY: State University of New York Press.

Palmer, T. (1992). The re-emergence of correctional intervention. Newbury Park, CA:
Sage.

Palmer, T. (1983). The “effectiveness” issue today: An overview. Federal Probation, 47,
3-10.

Palmer, T. (1975). Martinson revisited. Journal of Research in Crime and Delinquency,
12, 133-152.

Pampel, F. C. (2000). Logistic regression: A primer. Thousand Oaks, CA: Sage
Publications.

Paparozzi, M. A. & Gendreau, P. (2005). An intensive supervision program that worked:
Service delivery, professional orientation, and organizational supportiveness. The
Prison Journal, 85, 445-466.

Parke, R. D. & Clarke-Stewart, K. A. (2003). The effects of parental incarceration on
children: Perspectives, promises, and policies. In J. Travis & M. Waul (Eds),

Prisoners Once Removed: The Impact of Incarceration and Reentry on Children,
Families, and Communities (pp. 189-232). Washington, DC: The Urban Institute.

Petersilia, J. (2004). What works in prisoner reentry? Reviewing and questioning the
evidence. Federal Probation, 68(2), 4-8.

Petersilia, J. (2003). When prisoners come home: Parole and prisoner reentry. New
York, NY: Oxford University Press.

Petersilia, J. (1999). Parole and prisoner reentry in the United States. In M. Tonry & J.
Petersilia (Eds), Prisons (pp. 479-529). Chicago, IL: University of Chicago Press.

Petersilia, J. & Turner, S. (1991). An evaluation of intensive probation in California.
Journal of Criminal Law and Criminology, 82(3), 610-658.

259

 

Petrosino, A., Turpin-Petrosino, C., & Fickenauer, J .O. (2000). Well-meaning Programs
Can Have Harmful Effects! Lessons from Experiments of Scared Straight. Crime
and Delinquency, 46(3), 354-379.

Pew Center. (2009). One in 31: The long reach of American corrections. Washington,
DC: The Pew Center on the States, The Pew Charitable Trusts.

Pew Center. (2008). One in 100: Behind bars in America, 2008. Washington, DC: The
Pew Center on the States, The Pew Charitable Trusts.

Piliavin, I. & Gartner, R. (1981). The impact of supported work for ex-ojfenders. New
York, NY: Manpower Demonstration Research Corporation.

Piquero, A. R., MacDonald, J. M., & Parker, K. F. (2002). Race, local life circumstances,
and criminal activity. Social Science Quarterly, 83, 654-670.

Prendergast, M. L., Farabee, D., Cartier, J ., & Henkin, S. (2002). Involuntary treatment
within a prison setting: Impact on psychosocial change during treatment. Criminal
Justice and Behavior, 29(5), 5-26.

President’s Commission on Law Enforcement and Administration of Justice. (1967). The
challenge of crime in a free society. Washington, DC: US Government Printing
Ofﬁce.

Reiss, A. J. (1951). Delinquency as the failure of personal and social controls. American
Sociological Review, 16, 196-207.

Rhine, E. E., Mawhorr, T. L., & Parks, E. C. (2006). Implementation: The bane of
effective correctional programs. Criminology and Public Policy, 5(2), 347-358.

Roman, C. G. (2004). A roof is not enough: Successful prisoner reintegration requires
experimentation and collaboration. Criminology and Public Policy, 3(2), 161-168.

Roman, C. G. & Travis, J. (2004). Taking stock: Housing, homelessness, and prisoner
reentry. Washington, DC: Urban Institute, Justice Policy Center.

Rose, D. & Clear, T. (2003). Incarceration, reentry, and social capital: Social networks in
the balance. In J. Travis & M. Waul (Eds), Prisoners once removed: The impact
of incarceration and reentry on children, families, and communities (pp. 313-
341). Washington, DC: Urban Institute Press.

Rosenfeld, R. (2008). Recidivism and its discontents. Criminology and Public Policy,
7(2), 311-318.

Rossi, P. H., Freeman, H. E., & Lipsey, M. W. (2004). Evaluation: A systematic

260

approach (7th ed.). Thousand Oaks, CA: Sage Publications.

Rossman, S. B. (2003). Building partnerships to strengthen offenders, families, and
communities. In J. Travis & M. Waul (Eds), Prisoners once removed: The impact

of incarceration and reentry on children, families, and communities (pp. 343-
379). Washington, DC: Urban Institute Press.

Rossman, S. B. & Roman, C. G. (2003). Case managed reentry and employment: Lessons
from the opportunity to succeed program. Justice Research and Policy, 5(2), 75-
100.

Sampson, R. & Laub, J. (1993). Crime in the making: Pathways and turning points
through life. Cambridge, MA: Harvard University Press.

Sampson, R. J. & Laub, J. H. (1990). Crime and deviance over the life course: The
salience of adult social bonds. American Sociological Review, 49, 607-627.

Sampson, R. J ., Laub, J. H., & Wimer, C. (2006). Does marriage reduce crime? A
counterfactual approach to within-individual causal effects. Criminology, 44, 465-
508.

Schlay, A. B. & Rossi, P. H. (1992). Social science research and contemporary studies of
homelessness. Annual Review of Sociology, 18, 129-160.

Seiter, R. P., & Kadela, K. R. (2003). Prisoner reentry: What works, what does not, and
what is promising. Crime and Delinquency, 49(3), 360-388.

Sevigny, E. L. & Caulkins, J. P. (2004). Kingpins or mules: An analysis of drug offenders
incarcerated in federal and state prisons. Criminology and Public Policy, 3, 401-
434.

Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-
experimental designs for generalized causal inference. Boston, MA: Houghton
Mifﬂin Company.

Shaw, C. R. & McKay, H. D. (1969). Juvenile delinquency in urban areas (2"d Ed.).
Chicago, IL: University of Chicago Press.

Sherman, L. W. (2007). The power few: Experimental criminology and the reduction of
harm. Journal of Experimental Criminology, 3, 299-321.

Sherman, L.W., Gottfredson, D., MacKenzie, D., Eck, J ., Renter, P., and Bushway, S.
(1997). Preventing Crime: What Works, What Doesn ’t, What ’s Promising.
Washington, DC: National Institute of Justice.

Shover, N. & Thompson, C. Y. (1992). Age, differential expectations, and crime

261

desistance. Criminology, 30(1), 89-104.

Shrout, P. E. & Bolger, N. (2002). Mediation in experimental and nonexperimental
studies: New procedures and recommendations. Psychological Methods, 7(4),
422-445.

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

Simpson, D. D., Joe, G. W., Rowan-Szal, G., & Greener, J. (1995). Client engagement
and change during drug abuse treatment. Journal of Substance Abuse, 7, 117-134.

Sobel, M. E. (1986). Some new results on indirect effects and their standard errors in
covariance structure models. In N. Tuma (Ed.), Sociological Methodology (pp.
159-186). Washington, DC: American Sociological Association.

Solomon, A. L., Johnson, K. D., Travis, J ., McBride, E. C. (2004a). From prison to work:
The employment dimensions of prisoner reentry, a report of the reentry
roundtable. Washington, DC: Urban Institute, Justice Policy Center.

Solomon, A., Waul, M., van Ness, A., & Travis, J. (2004b). Outside the walls: A national
snapshot of community-based prisoner reentry programs. Washington, DC: The
Urban Institute.

Spelman, W. (2005). Jobs or jails? The crime drop in Texas. Journal of Policy Analysis,
24(1), 133-165.

Spelman, W. (2000). The limited importance of prison expansion. In A. Blumstein & J.
Wallman (Eds), The crime drop in America (pp. 97-129). Cambridge, MA:
Cambridge University Press.

Stephan, J. J. (2008). Census of state and federal correctional facilities, 2005. Bureau of
Justice Statistics, Special Report (NCJ 222182). Washington, DC: US Department
of Justice, Ofﬁce of Justice Programs, Bureau of Justice Statistics.

Stephan, J. J. (2004). State prison expenditures, 2001. Bureau of Justice Statistics,
Special Report (NCJ 202949). Washington, DC: US Department of Justice, Ofﬁce

of Justice Programs, Bureau of Justice Statistics.

Stephan, J. J. (1997). Census of state and federal correctional facilities, 1995. Bureau of
Justice Statistics, Special Report (NCJ 164266). Washington, DC: US Department
of Justice, Ofﬁce of Justice Programs, Bureau of Justice Statistics.

Steurer, S., Smith, L., & Tracy, A. (2002). Three state recidivism study. Lanham, MD:
Correctional Educational Association.

262

Tashokkori, A. & Teddlie, C. (1998). Mixed methodology: Combining qualitative and
quantitative approaches (Applied Social Research Methods). Thousand Oaks,
CA: Sage.

Taxman, F. S. (2008). No illusions: Offender and organizational change in Maryland’s
proactive community supervision efforts. Criminology and Public Policy, 7(2),
275-302.

Taxman, F. S. (2005). Brick walls facing reentering offenders. International Journal of
Comparative and Applied Criminal Justice, 29(1), 5-18.

Taxman, F. S. (2004). Research and relevance: Lessons from the past, thoughts for the
future. Criminology and Public Policy, 3(2), 169-180.

Taxman, F. 8., Young, D., & Byme, J. M. (2003). From prison safety to public safety:
Best practices in oﬂender reentry. Washington, DC: US Department of Justice,
National Institute of Justice.

Taxman, F. 8., Young, D., Byme, J. M., Holsinger, A., & Anspach, D. (2002). From
prison safety to public safety: Innovations in oﬂender reentry (N CJ 196464).
College Park, MD: University of Maryland.

Toch, H. (2002). Everything works. International Journal of Offender Therapy and
Comparative Criminology, 46, 119-122.

Toch, H. (1997). Corrections: A humanistic approach. Guilderland, NY: Harrow &
Heston.

Travis, J, & Petersilia, J. (2001). Reentry reconsidered: A new look at an old question.
Crime and Delinquency, 4 7(3), 291-313.

Travis, J ., Solomon, A. L., & Waul, M. (2001). From prison to home: The dimensions
and consequences of prisoner reentry. Washington, DC: The Urban Institute.

Tonry, M. (1995). Malign neglect: Race, crime and punishment in America. New York,
NY: Oxford University Press.

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

Uggen, C. (2000). Work as a turning point in the life course of criminals: A duration
model of age, employment, and recidivism. American Sociological Review, 67,
529-546.

Uggen, C., Manza, J ., & Behrens, A. (2004). Less than average citizen: Stigma, role
transition, and the civic reintegration of convicted felons. In S. Maruna & R.

263

Immarigeon (Eds), After crime and punishment: Pathways to offender
reintegration (pp. 25 8-290). Portland, OR: Willan Publishing.

Uggen, C. & Manza, J. (2004). Lost voices: The civic and political views of
disenfranchised felons. In M. Pattillo, D. Weiman, & B. Western (Eds),
Imprisoning America: The Social Effects of Mass Incarcertion (pp. 165—204).
New York, NY: Russell Sage Foundation.

Urban Institute (2008). Decision points ’08. Retrieved 04 October 2008 from:
http://www.urban.org/decisionpointsO8/08prisonerreentrv.cfm

 

US. Department of Justice. (2008). U. S. Department of Justice, Oﬂice of Justice
Programs, Reentry. Retrieved 20 February 2008 from

http://www.oip.usdoi.gow’reentrﬂ

 

Useem, B. & Piehl, A. M. (2008). Prison state: The challenge of mass incarceration.
Cambridge, NY: Cambridge University Press.

Vieraitis, L. M., Kovandzic, T. V., & Marvell, T. B. (2007). The criminogenic effects of
imprisonment: Evidence from state panel data, 1974-2002. Criminology and
Public Policy, 6(3), 589-622.

Visher, C. A. (2006). Effective reentry programs. Criminology and Public Policy, 5(2),
299-302.

Visher, C. & Farrell, J. (2005). Chicago communities and prisoner reentry.
Washington, DC: Urban Institute, Justice Policy Center.

Visher, C. A., & Travis, J. (2003). Transitions from prison to community:
Understanding individual pathways. Annual Review of Sociology, 29, 89-113.

Visher, C. A., Winterﬁeld, L., & Coggeshall, M. B. (2005). Ex-offender employment
programs and recidivism: A meta-analysis. Journal of Experimental Criminology,
1(3), 295-315.

Vold, G. B., Bernard, T. J ., & Snipes, J. B. (2002). Theoretical criminology (5th Ed.).
New York, NY: Oxford University Press.

Wacquant, L. (2001). Deadly symbiosis: When ghetto and prison meet and mesh.
Punishment and Society, 3(1), 95-134.

Wacquant, L. (2000). The new ‘peculiar institution’: On the prison as surrogate ghetto.
Theoretical Criminology, 4(3), 377-389.

Warr, M. (1998). Life course transitions and desistance from crime. Criminology, 36,
1 83-216.

264

Weisburd, D. (2000). Randomized Experiments in Criminal Justice Policy: Prospects and
Problems. Crime and Delinquency, 46(3), 181-193.

Wells, E. A., Hawkins, J. D., & Catalano, R. F. (1988). Choosing drug measures for
treatment outcome studies: The inﬂuences of measurement approach on treatment
results. International Journal of the Addictions, 23, 851-873.

Western, B. (2002). The impact of incarceration on wage mobility and inequality.
American Sociological Review, 67, 526-546.

Western, B. & Beckett, K. (1999). How unregulated is the US labor market? The penal
system as a labor market institution. American Journal of Sociology, 104, 1030-
1060.

Western, B., Kling, J. R., & Weiman, D. F. (2001). The labor market consequence of
incarceration. Crime and Delinquency, 47(3), 410-427.

Western, B., Lopoo, L. M., & McLanahan, S. (2004). Incarceration and the bonds
between parents in fragile families. In M. Pattillo, D. Weiman, & B. Western
(Eds), Imprisoning America: The Social Effects of Mass Incarcertion (pp. 21-45).
New York, NY: Russell Sage Foundation.

Western, B. & Pettit, B. (2005). Black-white wage inequality, employment rates, and
incarceration. American Journal of Sociology, 111, 553-5 78.

Wexler, H. K. (1995). The success of therapeutic communities for substance abusers in
American prisons. Journal of Psychoactive Drugs, 27, 57-66.

White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a
direct test for heteroskedasticity. Econometrica, 48, 817-838.

Wilson, W. J. (1996). When work disappears: The world of the new urban poor. New
York, NY: Knopf.

Wilson, W. J. (1990). The truly disadvantaged: The inner city, the underclass, and public
policy. Chicago, IL: University of Chicago Press.

Wilson, J. A. & Davis, R. C. (2006). Good intentions meet hard realities: An evaluation
of the project greenlight reentry program. Criminology and Public Policy, 5 (2),
303-338.

Winterﬁeld, L. & Castro, J. (2005). Returning home Illinois policy brief: Treatment
matching. Washington, DC: Urban Institute.

Winterﬁeld, L., Lattimore, P. K., Steffey, D. M., Brumbaugh, S., & Lindquist, C. (2006).
The serious and violent offender reentry initiative: Measuring the effects on

265

service delivery. Western Criminology Review, 7(2), 3-19.

Wooldridge, J. M. (2009). Introductory econometrics: A modern approach (4‘h Ed.).
Cincinnati, OH: South-Western College Publishing.

Zamble, E. & Quinsey, V. L. (1997). The criminal recidivism process. New York, NY:
Cambridge University Press.

Zhang, S. X. & Zhang, L. (2005). An experimental study of the Los Angeles County
repeat offender prevention program: Its implementation and evaluation.
Criminology and Public Policy, 4(2), 205-236.

Zhang, Z., Friedmann, P. D., & Gerstein, D. R. (2003). Does retention matter? Treatment
duration and improvement in drug use. Addiction, 98, 673-684.

 

266

Mllllllllllllll llllll llllllllllllllllllllllllllles
3 1293 03063 5464