An Evaluation of Seven Second Chance Act DOJ 2017

The Habeas Citebook: Prosecutorial Misconduct - Header

An Evaluation of Seven Second Chance Act DOJ 2017

• Locations: United States of America • Topics: Sentencing, Parole, Parole Conditions, Alternative Sentencing

Download original document:
Brief thumbnail

This text is machine-read, and may contain errors. Check the original document to verify accuracy.

The author(s) shown below used Federal funding provided by the U.S.
Department of Justice to prepare the following resource:
Document Title:

An Evaluation of Seven Second Chance Act
Adult Demonstration Programs: Impact
Findings at 18 Months

Author(s):

Ronald D’Amico, Christian Geckeler, Hui
Kim

Document Number: 251139
Date Received:

September 2017

Award Number:

2010-RY-BX-0003

This resource has not been published by the U.S. Department of
Justice. This resource is being made publically available through the
Office of Justice Programs’ National Criminal Justice Reference
Service.
Opinions or points of view expressed are those of the author(s) and
do not necessarily reflect the official position or policies of the U.S.
Department of Justice.

AN EVALUATION OF SEVEN SECOND CHANCE
ACT ADULT DEMONSTRATION PROGRAMS:
IMPACT FINDINGS AT 18 MONTHS
18-Month Impact Report
June 2017

Prepared by:
Ronald D’Amico
Christian Geckeler
Hui Kim

This project was supported by Award No. 2010-RY-BX-0003, awarded by the National Institute of
Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and
conclusions or recommendations expressed in this publication are those of the authors and do not
necessarily reflect those of the Department of Justice.

This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Acknowledgements
This study was supported by a grant from the U.S. Department of Justice (DOJ) National
Institute of Justice (NIJ). We are very grateful for NIJ’s support. Special thanks go to our project
officer, Dr. Marie Garcia, who provided invaluable advice and support throughout the project.
Many hurdles would have been far more difficult to overcome without her help. Thanks also to
Dr. Angela Moore Parmley, who shepherded the project through its final stages, and others at
NIJ who provided very helpful input along the way. We also are grateful for the assistance
provided by DOJ’s Bureau of Justice Assistance (BJA), particularly Gary Dennis, who helped us
obtain the grantees’ cooperation. DOJ’s information technology team was also of great help in
facilitating our access to the National Directory of New Hires.
We are heavily indebted to the study participants, who agreed to share their experiences with
us through the participant survey and the study’s other data collection activities. We also much
appreciate the cooperation of the seven grantees who participated in this study. They are:
1.

Allegheny County (PA) Department of Human Services

Kentucky Department of Corrections

Marion County (OR) Sheriff’s Office

Oklahoma Department of Corrections

San Francisco (CA) Department of Public Health

San Mateo County (CA) Division of Health and Recovery Services

South Dakota Department of Corrections

Each agency complied with the study’s requirements and their staff graciously answered our
many questions. This study would not have been possible without their help. We also
acknowledge help from state and local agencies who provided arrest, conviction, and jail and
prison incarceration data for the study sample.
We are grateful to the many members of the study team. In particular, we thank our partners
on this project from MDRC, especially Dan Bloom and Dr. Erin Jacobs Valentine, who provided
excellent advice on the study’s design and analysis. From NORC, Dr. Carrie Markowitz, Pamela
Loose, and the rest of the survey team are responsible for the high survey response rate. From
SPR, we thank Mary Hancock for her exceptional programming support, and Trace Elms and
Eduardo Ortiz for assistance collecting administrative data.
Finally, many thanks to those who provided valuable input on an earlier draft of this report.
These individuals include Drs. Parmley and Garcia, of NIJ; staff members from BJA and the
Council of State Governments Justice Center; Dan Bloom and Dr. Valentine, of MDRC; Dr.
i
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Markovitz, of NORC; Dr. Marian Negoita and Dr. Andrew Wiegand, of SPR; and two outside
reviewers.
All these individuals, and others, were instrumental in bringing this report to conclusion.

ii
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Table of Contents
ABSTRACT .................................................................................................................. VII
LIST OF ACRONYMS.................................................................................................... IX
EXECUTIVE SUMMARY ............................................................................................... XI
I.

INTRODUCTION ..................................................................................................... 1
Background ..................................................................................................................... 1
About the Evaluation ....................................................................................................... 3
Design and Implementation of the Study ..................................................................... 3
Data Collection for the Impact Evaluation .................................................................... 5
Estimating Program Impacts ....................................................................................... 7
Sample Sizes and Statistical Power .......................................................................... 11
Limitations of the Study ............................................................................................. 11
About Study Participants ............................................................................................... 13
Roadmap to the Report ................................................................................................. 15

II.

ABOUT THE SCA PROGRAMS ........................................................................... 19
Grant Funding and Use of Funds................................................................................... 19
Managing the Programs and Partners ........................................................................... 23
Types of SCA Grantees ............................................................................................ 23
Service Models and Partnerships .............................................................................. 24
Screening and Enrollment ............................................................................................. 25
Subgroups Targeted for Project Participation ............................................................ 26
Assessment .............................................................................................................. 27
Case Management Services.......................................................................................... 28
Other Services............................................................................................................... 30
Types of Services ..................................................................................................... 31
Delivering Services to Participants ............................................................................ 32
Services for the Control Group ...................................................................................... 32
Pre-Release Services for the Control Group ............................................................. 33
Post-Release Services for the Control Group ............................................................ 34
Summary ....................................................................................................................... 35

III.

PARTICIPANTS’ EXPERIENCES IN SCA ........................................................... 39
Services Received ......................................................................................................... 39
The Duration of Participation ......................................................................................... 42
iii
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

IV. IMPACTS ON SERVICES ..................................................................................... 45
Impacts on Services Received ...................................................................................... 45
Unmet Needs ................................................................................................................ 48
Impacts for Subgroups .................................................................................................. 49
Summary ....................................................................................................................... 50

IMPACTS ON RECIDIVISM .................................................................................. 57
Impacts Overall ............................................................................................................. 57
Impacts on Time to Release and Time at Risk .......................................................... 57
Impacts on Recidivism for the Full Sample ................................................................ 59
Impacts on Recidivism by Time at Risk ..................................................................... 63
Impacts for Subgroups .................................................................................................. 63
Summary ....................................................................................................................... 65

VI. IMPACTS ON EMPLOYMENT AND EARNINGS ................................................. 71
Impacts Overall ............................................................................................................. 71
Impacts for Subgroups .................................................................................................. 73
Summary ....................................................................................................................... 73

VII. IMPACTS ON OTHER OUTCOMES ..................................................................... 77
Impacts Overall ............................................................................................................. 77
Impacts for Subgroups .................................................................................................. 79
Summary ....................................................................................................................... 79

VIII. SUMMARY AND CONCLUSIONS ........................................................................ 85
About the Grantees and Their Programs ....................................................................... 85
Summary of Impact Findings ......................................................................................... 86
Why Were There No Impacts on Recidivism?................................................................ 87
Conclusions and Caveats .............................................................................................. 89
Next Steps ..................................................................................................................... 90

REFERENCES .............................................................................................................. 91
APPENDIX A. IMPLEMENTING RANDOM ASSIGNMENT ....................................... 97
Changes to Eligibility to Accommodate Random Assignment ........................................ 97
The Random Assignment Process ................................................................................ 97
Providing an Orientation and Obtaining Consent ....................................................... 98
The Mechanics of Random Assignment .................................................................... 98

APPENDIX B. SURVEY METHODS ........................................................................ 101
Locating Respondents ................................................................................................. 101

iv
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Conducting the Interviews ........................................................................................... 102
Weighting .................................................................................................................... 103

APPENDIX C. METHODS FOR DATA ANALYSIS.................................................. 105
Statistical Methods Used ............................................................................................. 105
Difference in Means ................................................................................................ 106
Regression Analysis................................................................................................ 106
Hierarchical Linear Modeling ................................................................................... 107
Survival Analysis ..................................................................................................... 108
Results of the Sensitivity Analysis ............................................................................... 109

APPENDIX D. DEFINING RISK SUBGROUPS ....................................................... 113

v
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Abstract
This report describes the impacts of seven programs that were awarded grants under the
Second Chance Act (SCA) Adult Demonstration Program to reduce recidivism by addressing the
challenges faced by adults returning to their communities after incarceration. In estimating
impacts, the evaluation used a randomized controlled trial, whereby 966 individuals eligible for
SCA were randomly assigned to either a program group whose members could enroll in SCA, or
a control group whose members could not enroll in SCA but could receive all other services
generally available. Using survey and administrative data, each study participant was measured
on a range of outcomes 18 months after random assignment.
Using their SCA funds, the grantees improved their partnerships with community agencies and
strengthened the connection between pre-release and post-release services. All used their SCA
funds to provide services after individuals were released from incarceration, and most also
enhanced pre-release services. Services included education and training, employment
assistance, substance abuse treatment, mental health services, cognitive behavioral therapy,
housing assistance, and supportive services. Grantees provided some of these services using
their SCA funds and others through unfunded referrals to community partners. Case
management was a common service element; case managers were either parole officers who
had reduced caseloads or staff members from social services agencies or community-based
organizations. Because case management was the focal point of most grantees’ efforts, the
impact study primarily represents the influence of this service. Nonetheless, given the diversity
of approaches taken by the grantees, this study does not provide a test of a single program
model.
Impact findings show that those assigned to the program group were significantly more likely
than those assigned to the control group to have received help with re-entry and were more
likely to have had an individual case plan. They were also more likely to have received cognitive
behavioral therapy, help with looking for a job, substance abuse treatment, housing assistance,
and mentoring. However, many control-group members also received these services, and, at
the end of 18 months, SCA participants were just as likely as those in the control group to
report that additional services would have been helpful.
Being assigned to the program group did not reduce involvement with the criminal justice
system in the 18 months after random assignment. Whether recidivism was measured using
survey or administrative data, those in the program group were no less likely to be re-arrested,
reconvicted, or re-incarcerated; their time to re-arrest or re-incarceration was no shorter; and
they did not have fewer total days incarcerated (including time in both prisons and jails). Those

vii
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

in the program group were somewhat more likely to have had probation or parole revoked and
to have new convictions.
Being assigned to the program group also did not significantly improve employment outcomes
and had no effect on other outcomes, including the adequacy of housing, health status, or the
ability to meet child-support obligations.
One reason why impacts were not greater is that, although SCA significantly increased access to
a wide range of services, the difference in service receipt between the program group and the
control group was modest. Furthermore, SCA funds did not seem adequate to meet the many
and complex needs of those returning from incarceration. Finally, most grantees emphasized
case management as the key service strategy, and prior research has suggested that casework
alone is not very successful as a re-entry approach.
The grantees in this study were among the first to receive SCA funding. Grant requirements
were substantially tightened for grantees that received funding in subsequent rounds of
competition. Further research is needed to determine whether these enhanced requirements
led to programs that were effective in reducing recidivism.

viii
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

List of Acronyms
ASI =Addiction Severity Index
BIF = Baseline Information Form
BJA = Bureau of Justice Assistance, U.S. Department of Justice
CAIS = Correctional Assessment and Intervention System
DOC = Department of Corrections
DOJ = U.S. Department of Justice
FY = Fiscal year
GED = General Educational Development
HHS = U.S. Department of Health and Human Services
HLM = Hierarchical linear modeling
ICC = Intraclass correlation coefficient
IRB = Institutional Review Board
LS/CMI = Level of Service/Case Management Inventory
LSI-R = Level of Service Inventory-Revised
MDD = Minimum detectable difference
MIS = Management information system
NCIC = National Crime Information Center
NDNH = National Directory of New Hires
NIJ = National Institute of Justice, U.S. Department of Justice
OCSE = Office of Child Support Enforcement, U.S. Department of Health and Human Services
PO = Probation or parole officer
RA = Random assignment
RCT = Randomized controlled trial
SCA = Second Chance Act of 2007
UI = Unemployment Insurance
URICA = University of Rhode Island Change Assessment Scale

ix
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Executive Summary
The Second Chance Act (SCA), signed into law in 2008 with widespread bipartisan support,
authorizes grants to government agencies and nonprofit organizations to reduce the recidivism
of individuals being released from prisons and jails. Thus far, more than 600 grants have been
awarded for programs serving adults under various categories of competition.1 This report
describes the 18-month impacts of seven programs that were awarded grants through the first
round of funding under the SCA Adult Demonstration Program. The Adult Demonstration
program represents only one of a number of separate grant programs authorized through SCA.
Because these seven programs were purposively selected and were drawn from only one grant
program, this study’s findings cannot be generalized to other grantees that received Adult
Demonstration funds or to SCA as a whole.

About the Evaluation
This evaluation uses a random assignment (RA) design and administrative and survey data to
study the impacts of these seven SCA programs. For the impact study, 966 individuals eligible
for SCA were assigned to either:
•

A program group whose members could enroll in SCA, or

•

A control group whose members could receive all services otherwise available but could
not enroll in SCA.

RA for the impact study commenced in the last week of 2011 and continued through March
2013. Of the 966 study participants, 63 percent were randomly assigned to the program group
and 37 percent to the control group. Data on study participants are from a number of sources.

•

Baseline Information Form (BIF). Just before RA, all study participants completed a onepage BIF; this form asked about the individual’s background and criminal history.

•

Data extracted from grantees’ management information systems (MISs). The grantees
provided the study team with data showing which pre-release and post-release services
program-group members received as part of their participation in SCA.

•

Administrative data from state and local criminal justice agencies. State and local
criminal justice agencies provided data on arrests, convictions, and prison and jail
incarcerations for the 10 years prior to each individual’s RA date and the 18 months
following RA.

Cited from the Office of Justice Programs, Bureau of Justice Assistance website, accessed at
https://www.bja.gov/ProgramDetails.aspx?Program_ID=90#horizontalTab2 on July 8, 2016.
xi
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

•

Administrative data from the National Directory of New Hires (NDNH). NDNH, built up
from states’ quarterly Unemployment Insurance (UI) program wage and claimant files,
federal employment files, and the Directory of New Hires, provides information on study
participants’ employment and earnings.

•

A follow-up survey. The research team administered a follow-up survey to cover the 18
months following RA. All study participants were in the survey sampling frame and
interviews were completed with 82.3 percent of them (82.2 percent of the program
group and 82.6 percent of the control group).

The research team also conducted site visits to the grantees to learn about program
implementation, and a separate report describes those findings.
Using the survey and administrative data on study participants, the study team estimated the
impacts of being assigned to the SCA program for the full sample as well as for five subgroups:
those defined by gender, age (under age 30 versus ages 30 or more), risk of recidivism (lower
versus higher risk), length of time from random assignment to release from incarceration (RA
was more than 30 days prior to release from custody versus within 30 days of release or after
release), and type of grantee (a criminal justice agency versus a social service or health agency).
In estimating impacts, the study uses an intent-to-treat framework by comparing the outcomes
of those assigned to the SCA program group to the outcomes of those assigned to the control
group. Some program-group members might not have received all the SCA services intended
for them and, conversely, control-group members could have received very similar services
from sources other than SCA.
At least two of the seven grantees used a portion of SCA funds for general system
improvements. The control-group members could have benefited from these improvements,
just as any other individual returning from incarceration. Therefore, this study assesses the
impacts of the personalized services that program-group members received as part of SCA and
not of these system improvements.

The implementation study revealed several key points about the grantees and their programs.
SCA funds helped expand re-entry services. The grantees reported that their SCA grants helped
them fill gaps in their existing re-entry services and expand service capacity. Partly through
their grants, the grantees improved their partnerships with other community agencies and
strengthened the connection between pre-release and post-release services.
The grantees were a diverse group and targeted different populations. Three of the seven
grantees were state departments of corrections (DOCs), one was a sheriff’s office, and three
were local government social services or health agencies. Some recruited SCA participants
exclusively from prisons, others exclusively from jails, and others from both prisons and jails.
Some grantees served only females, some served only males, and others served both females
and males.
The emphasis on pre-release services was greater in some sites than others. Three grantees
delivered fairly extensive pre-release services as part of their SCA programs, and, therefore,
required participants to have an extended period of incarceration remaining at the time of SCA
enrollment. Others relied heavily on existing programming in institutions for pre-release
services and focused on using their SCA funds for transition planning and post-release services;
they generally enrolled participants in SCA nearer to release and, sometimes, after release.
Overall, approximately 55 percent of participants were enrolled in SCA three or more months
prior to release, 28 percent within three months of release, and 17 percent after release.
Case management was a key service. Case management was a central feature of all the
grantees’ programs except one. Across grantees, the goal of case management was to help
prevent recidivism by providing individualized support and coordinating access to services
based on identified needs and risk factors. These case managers were either probation or
parole officers (POs) who commonly had reduced caseloads and extra training provided
through the grant, or came from social services agencies and had more traditional case
management backgrounds (i.e., social workers, counselors). In the latter case, SCA participants
might also have been required to report to a PO after release, but this individual was not the
SCA case manager.
Grantees provided other services to SCA participants directly and through referrals. The
grantees made a range of services available: education and training, employment assistance,
substance abuse treatment, mental health services, cognitive behavioral therapy, pro-social
services, housing assistance, and supportive services. The grantees provided some of these
services directly. Other services were provided through a network of partners, sometimes on a
fee-for-service basis but often through unfunded referrals. Where unfunded referrals were
used, coordination with the SCA program was typically weak and case managers could not
readily track whether participants received the services to which they were being referred.

xiii
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Given the centrality of case management to the grantees’ service models, the impact study
primarily represents the influence of this service. However, given the diversity of approaches
taken by the grantees, SCA as it operated in these sites does not represent a single program
model.

Types of Services Received by SCA Participants
Grantees provided MIS data for those assigned to the program group. These data capture the
services provided to SCA participants that the grantees knew about and entered into their data
systems.
Just over one-third of those assigned to the SCA program group received both pre-release and
post-release SCA services following their enrollment in the program. According to the
grantees’ MIS data, 36 percent of those in the SCA program group received both pre-release
and post-release services as part of SCA, 40 percent received only post-release service, and 24
percent received only pre-release services. Note that those not receiving pre-release or postrelease services as part of SCA could have received those services from other sources.
Employment assistance, cognitive behavioral therapy, and substance abuse treatment were
the most common services provided through SCA, both before and after release. Nearly onehalf of the SCA program group received employment assistance and cognitive behavioral
therapy as part of SCA while they were still incarcerated, and more than one-third received
substance abuse treatment. These three services were also the most common ones provided
through SCA after release.
The length of participation in SCA varied greatly. Approximately 26 percent of those assigned
to the program group participated in SCA for more than one year, and another 35 percent
participated for more than six months. A little less than 40 percent participated for up to six
months.

Impacts on Services
The logic underlying the SCA grant program is that SCA funds are used to provide more
comprehensive and coordinated re-entry services than would be available in the absence of
SCA, and that these services will, in turn, improve desistance and lead to other desirable
outcomes. An important step in the evaluation, therefore, was to assess whether, in fact, those
assigned to the SCA program group received more services than those assigned to the control
group. Service receipt was measured through a follow-up survey administered to both the
program and control groups and covered the 18 months after RA.

xiv
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

The program group was significantly more likely to receive help with re-entry. Those assigned
to the program group were significantly more likely to report getting help with re-entry, and
they were more likely to have an individual case plan. They were also more likely to report that
they had someone who went out of the way to help them and to whom they could turn for
advice.
SCA significantly increased a wide range of other re-entry services. Those assigned to the
program group were significantly more likely to receive cognitive behavioral therapy, help with
finding a job, substance abuse treatment, housing assistance, and mentoring.
SCA could not provide all the services that participants desired. Despite the fact that SCA had
a significant impact on services received, the program group reported having many unmet
service needs 18 months after RA. In fact, their needs for additional services were no less than
the control group’s needs. For example, approximately two-thirds of both groups reported
wanting additional housing assistance and job placement assistance, and more than half
wanted additional health services, educational services, and job training. More than one-third
wanted family reunification services, substance abuse treatment, and mental health services.

Impacts on Recidivism
The study measured recidivism as involvement with the criminal justice system in the 18
months after RA that led to re-arrest, reconviction, or re-incarceration. These outcomes were
measured using both administrative and survey data.
As of 18 months after random assignment, increased access to services for SCA participants
did not lead to increased desistance. Whether recidivism was measured using survey or
administrative data, those in the program group were not less likely than those in the control
group to be re-arrested, reconvicted, or re-incarcerated; their time to re-arrest or reincarceration was no shorter; and they did not have fewer total days incarcerated (including
time in both prisons and jails). There is some evidence that those in the program group were
somewhat more likely to be convicted of a new crime or have probation or parole revoked; this
higher incidence may have occurred because enhanced case management for those in the
program group could have increased the likelihood of catching new offenses and violations of
terms of parole or probation when they occurred.

Impacts on Other Outcomes
There were no program impacts on employment-related outcomes. In the seven grantee sites
participating in this study, assignment to the program group did not improve the probability of
being employed in the follow-up period. In the last six months of follow-up, those in the

xv
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

program group earned an average of approximately $3,200 and those in the control group
approximately $3,000, but the difference between the groups is not statistically significant.
SCA may have improved income adequacy. Study participants were asked about their income
for the last month of the 18-month follow-up period. Those in the program group were more
likely than those in the control group to report that they had enough income to support
themselves during that month.
There were no effects on a range of other outcomes. Using survey data, the study measured
the adequacy of housing, health status, the self-reported incidence of illegal drug use and
excessive alcohol consumption, and the ability to meet child-support obligations. Assignment to
the program group had no effect on any of these outcomes.

Other Analyses
There are, at best, modest differences across subgroups. As an exploratory analysis, the study
estimated program impacts separately across the different subgroups mentioned previously.
Although there were some modest differences, impacts of assignment to the program group
were about the same for all the subgroups.
The study’s major findings are robust to alternative model specifications and data sources.
Program impacts were calculated as a simple difference in means between the program and
control groups and using more complex statistical models. For recidivism, the key outcome of
interest, program impacts were also estimated using both administrative data and survey data,
which provide independent estimates of desistance. The findings summarized above are robust
to these alternative model specifications and data sources.

Conclusions
SCA represented a substantial infusion of funds for these seven grantees, and this study has
demonstrated that this led to a statistically significant increase in service receipt for the
program group. Why did these additional services not improve desistance? A number of
general reasons can be suggested (although not every reason applies to each grantee).
1.

Control-group members accessed many of the same services that program-group
members did, both before and after release. Although SCA significantly increased
access to a wide range of services, the difference in service receipt between the
program group and the control group was modest—at most, the program group was 25
percentage points more likely to receive a given service than the control group. For
example, 61 percent of the program group reported getting help with job-finding skills,
but 40 percent of the control group also reported receiving this service. Even if the
services were effective, the gap in service receipt between the groups might not be
xvi
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

large enough to translate into differences in recidivism or other outcomes. There are
several reasons why so many control group members were able to access services.
a. Control-group members had access to services available in prisons and jails after RA
but while still incarcerated. Most institutions had courses and workshops available to
their inmates without regard to SCA eligibility. Depending on the institution, these
services included substance abuse treatment, adult literacy instruction, employment
assistance, cognitive behavioral therapy, and others. RA generally occurred while
individuals were incarcerated, so the control group, just as the program group, had
access to these services. Although SCA case managers who worked with SCA
participants prior to release might have made special efforts to encourage programgroup members to take advantage of these services, the services were generally
available to those in the control group without restriction.
b. A substantial proportion of control group members got help with re-entry from a PO
or case manager. According to the participant survey, 59 percent of the control
group reported that they got help with re-entry (compared to 78 percent of the
program group). Whether this help was provided by a traditional PO or someone
else, this individual could have provided many of the same services that SCA case
managers did: assessing service needs, offering advice, and providing referrals. From
qualitative findings, we know that SCA case managers and POs were more involved
than traditional POs were in brokering services, but the difference was one of
degree.
c. Grantees and their partners had other sources of funding, which were, in many
cases, quite substantial. All the grantees were required to leverage funds from
multiple sources, which could include state and local funds and grants from
philanthropic organizations or other sources. Similarly, the grantees’ partners were
existing organizations with their own funding sources and pre-existing outreach
mechanisms. SCA funds, while much appreciated and valued by all the grantees and
the partners the grantees funded, were often not the largest share of the
organizations’ budgets. These other sources of funds were not specifically
earmarked for SCA participants and could have been used to serve control group
members and others in need.
2.

Given available funding, there were limitations to what the grantees could do. Those
returning from incarceration face challenges to re-entry that are many and complex. The
grantees could not help participants fully overcome these challenges.
a. Funds were not adequate to directly fund all participants’ needs. Due to resource
constraints, all the grantees relied on unfunded partners and informal referrals to
provide some services. For services that were not SCA funded, program-group
members did not have priority access over anyone else who sought services.
b. At the end of 18 months, SCA participants had many unmet needs. Perhaps because
of the funding constraints, those in the program group reported many unmet service
needs 18 months after RA, including the need for housing assistance, job placement

xvii
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

assistance, job training, health services, and educational services. In fact, at the end
of the 18-month follow-up period, program-group members were just as likely to
express the need for additional services as those in the control group.
3.

There were inherent limitations to the projects that grantees developed. Although the
grantees used evidence on what works in developing their programs, there were
limitations to their program models.
a. Case management, even with reduced caseloads, has not been demonstrated to be
effective. All but one of the grantees emphasized case management as part of their
SCA programs. For several grantees, this case management was provided by POs
who were given reduced caseloads; for others, it was provided by staff members
from a social services agency or community-based organization. However, in their
review of correctional rehabilitation approaches, Cullen and Gendreau (2000) cite
evidence that “casework” has not been demonstrated to be very successful as a reentry approach. Others have concluded that giving POs reduced caseloads does not
by itself appear to reduce recidivism, and the increased supervision can increase
revocation rates (Petersilia 1999, Jalbert et al. 2011).
b. It is hard to ensure that participants get the services they need through unfunded
referrals. As noted, many services were provided through unfunded referrals. Using
unfunded referrals to provide services had some clear advantages: this strategy
conserved limited project resources and enabled grantees to draw on a wide
network of community agencies experienced at addressing the many complex needs
of those returning from incarceration. However, its limitation was that there was
often no way for the grantee to ensure that participants would seek out the services
to which they were referred. Moreover, the quality of services provided by loosely
connected partners can be uncertain.
c. Developing strong programs based on the risk-need-responsivity (RNR) framework is
difficult. Programs that address criminogenic needs have been shown to be effective
in reducing recidivism (e.g., Latessa and Lowenkamp 2006). However, Bonta and
Andrews (2007) argue that taking the RNR framework out of a tightly controlled
setting and trying to widely use its principles in the real world tends to make the
model much less effective. In their systematic review of the literature, Weisburd et
al. (2017) note that, while we generally know what works in reducing recidivism, the
specific guidance that practitioners need to convert principles into practice is often
lacking. In short, implementing evidence-based practices and taking them to scale is
not easy.

Despite these limitations, the SCA grants were meaningful; they helped the grantees enhance
their existing programs and capacity and strengthen partnerships. Absence of evidence that
these funds reduced recidivism to some degree highlights a well-known limitation of impact
studies: if there are alternative sources of funds for services, then each source is important in
expanding a community’s capacity but no one source is singularly impactful when compared
against all the others (Heckman et al. 2000).
xviii
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

At the same time, modifications to the service models that the grantees developed might help
improve outcomes. Even before these impact findings were made available, the Department of
Justice learned from the experiences of the grantees in this study and others that received early
funding through the Adult Demonstration Program. Based on what it learned from the
grantees’ implementation experiences, it tightened requirements for grantees that received
subsequent waves of grant funding under the Adult Demonstration program (now called Smart
Reentry). For example:
•

To ensure adherence to evidence-based practices and the provision of meaningful reentry services, grantees are now required to complete a planning process before being
approved for implementation funds. During this time, they are to work with a technical
assistance provider to improve their program models.

•

Grantees are required to establish a memorandum of understanding with providers to
ensure that there is a mechanism for follow-up when referrals are made.

•

Grantees must engage with participants prior to release.

•

Grantees must ensure adequate dosage of cognitive-based interventions.

With these modifications to grant requirements, this next generation of Smart Reentry holds
significant promise for yielding meaningful benefits.

Next Steps for the Evaluation
The findings described in this report cover 18 months after RA. This represents a relatively short
observation period. Many SCA participants in this study were enrolled in SCA while they were
incarcerated and were not released from custody for six or more months after RA. Therefore,
the post-release period covered by this study is much shorter than 18 months for many
individuals. As a consequence, there was limited time during the post-release observation
period for program- and control-group members to differentiate themselves. It is possible that
these programs will turn out to be effective with a longer post-RA observation period.
The next step for the evaluation is to estimate impacts measured 30 months after RA, which
may shed additional light on the programs’ effectiveness. A report based on these data is
forthcoming.

xix
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

xx
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Introduction

The annual number of inmates being released from incarceration increased dramatically in
recent decades, and rates of recidivism for the formerly incarcerated have been disturbingly
high. In light of these facts, significant resources have been devoted to improving the success of
re-entry for persons making the transition from incarceration to the community. This report
describes the impacts of re-entry services provided by seven grantees that received an early
round of funding through the Second Chance Act (SCA) Adult Demonstration Program. Impacts
are estimated on services received, recidivism, employment and earnings, family stability, and
other outcomes measured for the 18 months after individuals were determined eligible to
participate in SCA. This chapter establishes the context for the study, presents the evaluation
design, discusses limitations, and describes study participants. Subsequent chapters present the
study’s findings.

Background
At the end of 2014, approximately 6.85 million individuals were under some form of supervision
by the U.S. adult correctional system, representing about 1 in 36 adults in the U.S. (Kaeble et al.
2016). The total figure includes more than 1.5 million adults held in state or federal prisons,
approximately 745,000 confined in local jails, and more than 4.7 million under community
supervision.2 Although the total figure represents a substantial decline since the peak in 2007,
about three and a half times as many adults were under some form of supervision as in 1980,
when national estimates first became available (Glaze 2010). Moreover, flows are substantial—
each year, there are more than 450,000 entries to parole and several million entries to
probation (Kaeble and Bonczar 2016). These figures suggest that the burden on the nation’s
correctional system is extraordinary.
Adding to the challenge, those released from incarceration face substantial obstacles to
successful re-entry. Substantial numbers lack a high school degree or equivalent (Harlow 2003)
and many have problems with substance abuse and mental health or physical impairments
(Petersilia 2003, James and Glaze 2006, Hammett et al. 2001, Mumola and Karberg 2006). Upon
release, they have difficulty finding jobs for these reasons, and because of the stigma that
comes with their status as having been incarcerated (Pager 2003, Holzer et al. 2004, Raphael
2014). Moreover, the formerly incarcerated tend to be released into a relatively small number

The sum of the components exceeds the total because some individuals had multiple correctional statuses (see
Kaeble et al. 2016).
1
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

of urban neighborhoods that are characterized by high rates of poverty and other social
problems (Travis et al. 2001, La Vigne and Kachnowski 2003).
Not surprisingly given these challenges, about two-thirds of those released from state prisons
are re-arrested and nearly half are returned to prison within three years of release, either for
violations of parole conditions or new crimes (Durose et al. 2014, Pew Center on the States
2011). This cycle of imprisonment and re-entry has tremendous personal consequences for the
men and women who churn in and out of the criminal justice system and costs that extend to
many spheres of public policy and community life. High rates of recidivism impose a financial
drain on federal and state governments, impair public safety, strain community resources, and
impose hardship on the families of those who are imprisoned. Reducing recidivism is therefore
critical, both as a means of reducing corrections costs and as a strategy for addressing the
interrelated problems of low-income families and vulnerable communities.
In recognition of the gravity of the situation and the urgency of the need, SCA was signed into
law on April 9, 2008, with widespread bipartisan support. Since then, more than $475 million
has been awarded through over 600 grants to government agencies and nonprofit
organizations under various categories of competitions.3
One category of grant awards consists of Adult Demonstration Programs, from which the
grantees included in this study are drawn. The U.S. Department of Justice (DOJ) Bureau of
Justice Assistance (BJA) has made annual awards in this category since fiscal year (FY) 2009,
with more than 150 grants thus far awarded to state and local governments and federally
recognized Indian tribes for planning and implementing strategies to address the challenges
faced by adults returning to their communities after incarceration.4 Grantees are expected to
use validated and dynamic risk and needs assessments for purposes of delivering evidencebased services. Far removed from a time when it seemed that “nothing works,” there is now
considerable evidence that well-designed re-entry programs can make a difference.5 Grantees
are expected to draw on this evidence in designing their programs.

These figures are cited from the Office of Justice Programs, Bureau of Justice Assistance website (accessed at
https://www.bja.gov/ProgramDetails.aspx?Program_ID=90 on October 19, 2016) and include grants awarded
as part of the Adult Demonstration Program as well as other SCA grant competitions, such as the Reentry
Program for Adults with Co-occurring Disorders, the Adult Mentoring Program, and the Reentry Courts
Program, among others.

The number of awards in the SCA Adult Demonstration Program is as of January 2016, as cited at
https://csgjusticecenter.org/wp-content/uploads/2014/08/SCA_Fact_Sheet.pdf, accessed on October 19, 2016.

See the reviews by Seiter and Kadela (2003), Drake et al. (2009), and Cullen and Gendreau (2000). While
lauding prior work for its insights, Petersilia (2004) notes the paucity of impact studies using rigorous methods.
2
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

About the Evaluation
This evaluation uses a random assignment (RA) design and administrative and survey data to
study SCA Adult Demonstration grantees that were selected by BJA to participate in the study.

Design and Implementation of the Study
DOJ’s National Institute of Justice (NIJ) awarded a grant to Social Policy Research Associates
(SPR) and its partners, MDRC and NORC at the University of Chicago (NORC), to evaluate seven
grantees awarded FY 2009 SCA Adult Demonstration funding. These grantees were awarded
their funds in late summer 2009 and began enrolling participants several months after that.
Some of the grantees are state departments of corrections (DOCs); others are local government
agencies, including a sheriff’s office and public
Grantees Selected by BJA for the Study
health and social services agencies.
State Agencies

As part of the evaluation of these programs,
the research team was to:
1.

Conduct an implementation study of
the seven programs to learn grantees’
strategies for developing their
programs and the challenges they
encountered in providing re-entry
services.

Kentucky Department of Corrections
[Kentucky]

Oklahoma Department of Corrections
[Oklahoma]

South Dakota Department of
Corrections [South Dakota]

Local Agencies

Estimate the impacts of the grantees’
programs on participants’ recidivism,
employment, and other outcomes,
and calculate program costs.

For the implementation study, the research
team reviewed documents and conducted site
visits to each of the grantees. During the site
visits, research team members interviewed
program administrators and line staff and
conducted focus groups with program
participants. Results from the implementation
study are summarized in Chapter II of this
report and described in more detail in a
separate report (D’Amico et al. 2013).

Allegheny County (PA) Department of
Human Services [Allegheny County]

Marion County (OR) Sheriff’s Office
[Marion County]

San Francisco (CA) Department of
Public Health [San Francisco]

San Mateo County (CA) Division of
Health and Recovery Services [San
Mateo County]

_______________________
Note: The shorthand names by which grantees are
referred to in this report are shown in brackets.

In estimating impacts, those determined eligible for SCA were randomly assigned to either:
•

A program group whose members could enroll in SCA, or

3
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

•

A control group whose members could not enroll in SCA but could receive all services
otherwise available.

RA for the impact study commenced in the last week of 2011 (or approximately two years after
the grantees began operating their SCA programs) and continued through March 2013. The
timeline for the grantees and the study is displayed in Exhibit I-1. The exact date when RA
started varied by grantee, and was contingent on each grantee’s readiness to participate in the
study. RA ceased when each grantee reached its enrollment target or by the end of March
2013, whichever occurred sooner.6 All the grantees conducted RA for at least eight months;
during study intake, 966 individuals were randomly assigned.
Exhibit I-1: Timeline for SCA and Study Implementation
for the Study’s Grantees
SCA grant funds awarded
(fall 2009)

2009

2010

Start of random assignment
(Dec 23, 2011)

2011

2012

Grantees began enrolling in SCA
(late 2009/early 2010)

2013

End of random assignment
(March 28, 2013)

In addition to varying the dates when RA started and stopped for each grantee, the study team
adapted RA procedures for each grantee in other ways so that the study would be assessing
SCA as it was intended to operate in each site. For example, given their different funding levels
and service designs, each grantee was assigned a different enrollment target. Further, the rate
of random assignment to the program group varied. A condition of obtaining approval for the
research design from the study’s Institutional Review Board (IRB) was that no grantee could
have unfilled SCA program slots. Given the expected sizes of their applicant pools in relation to
their funding levels, most grantees were able to assign approximately 60 percent of those
eligible for SCA to the program group and 40 percent to the control group. However, for two
grantees, the study team randomly assigned approximately 75 percent of applicants to the
program group. Exhibit I-2 shows the numbers that each grantee enrolled in the program and
control groups. Appendix A provides more information about the mechanics of random
assignment.

The first grantee to start RA had its first applicant randomly assigned on December 23, 2011. The last grantee
started on May 8, 2012. Random assignment ceased for some grantees in December 2012, and for the others
by March 28, 2013.
4
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

The study team also let each grantee determine when in the transition from incarceration to
release that RA occurred. Given their standard program designs, some grantees enrolled
individuals in SCA six or more months before expected release from incarceration and provided
pre-release services during that time. By contrast, other grantees generally relied on the
institutions’ pre-existing pre-release services and began what were uniquely SCA services only
as the release date neared or after release. Regardless, RA always occurred just before an
individual’s intensive and personalized involvement with SCA was expected to occur. Given the
grantees’ varying program designs, this means that some study participants were randomly
assigned well before they were released, others near the date of release, and still others after
release.
Finally, the study allowed each grantee to establish its own criteria for determining who was
eligible for SCA and what services would be provided, subject to the requirements of their
grants (eligibility and service strategies are discussed in the next chapter).
Exhibit I-2: Number of SCA Participants in the Study,
by Grantee and Group
Total

Program Group

Allegheny County

133

105

Kentucky

187

113

Marion County

119

Oklahoma

134

San Mateo County

114

South Dakota

202

120

Total

966

606

360

San Francisco

Control Group

Source: Random assignment system.

Data Collection for the Impact Evaluation
The study team collected data for the implementation study through multi-day visits to each
grantee site. For the impact study, data were collected from five additional sources:
1.

Baseline Information Forms. All study participants completed a one-page Baseline
Information Form (BIF) just before RA. The form asked about the individual’s
background and criminal history (e.g., gender, age, race and ethnicity, level of
education, employment history, type of crime for which the most recent incarceration
occurred, length of sentence). Additionally, the program applicant was asked to provide
identifying information, such as a social security number and prison or jail identification
numbers (IDs).

5
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Data extracted from the grantees’ management information systems. We asked each
grantee to provide us with data extracted from its management information system
(MIS) on the services that SCA program participants received. The data elements we
requested represented a subset of those that grantees needed to report in the
Performance Management Tool to meet the quarterly reporting requirements of their
grants as stipulated by BJA. These elements included each participant’s date of SCA
enrollment and date of last service and indicators for which pre-release and postrelease services that SCA participants received, including substance abuse treatment,
mental health services, and employment services, among others. These data are only
available for SCA participants, and not those assigned to the control group.

Administrative data from state and local criminal justice agencies. We forwarded
participants’ identifying information collected on the BIFs to state and local criminal
justice agencies for matching with agency records. These agencies included
departments of corrections, departments of justice, offices of the courts, sheriff’s
offices, and others. Depending on each agency’s data system, the matching was
conducted using criminal justice IDs, social security numbers, names and birthdates, or
combinations of these. We requested participant data from each agency twice, once for
data covering a period beginning at least 10 years prior to the start of RA up through
September 2014, and again covering the period through September 2015. Using these
data, we created measures of arrests, convictions, and incarcerations (both prison and
jail) benchmarked to the RA date. One set of measures covered the ten-year period
prior to each individual’s RA date and was used to describe the sample’s criminal history
and create subgroups used in the analysis. Another set covered the period from the RA
data through 18 months after RA; this set constitutes key outcomes used in this analysis.
(Data for the period through September 2015 were still being collected at the time this
report was written. These data are used to create 30-month outcome variables; 30month impacts will be described in a subsequent report.)

Administrative data from the National Directory of New Hires. The National Directory of
New Hires (NDNH) is maintained by the Office of Child Support Enforcement (OCSE) of
the U.S. Department of Health and Human Services (HHS). NDNH is built up from states’
quarterly Unemployment Insurance (UI) program wage and claimant files, federal
employment files, and the Directory of New Hires, and includes information on covered
workers’ dates of hire, quarterly employment and earnings, and UI claimant benefit
amounts. The database is maintained to assist states in enforcing child-support
obligations for noncustodial parents, but can be used for research purposes under
strictly defined circumstances. Through an agreement between HHS and DOJ, the study
team gained access to NDNH data to calculate study participants’ employment and
earnings for the period following RA.

A follow-up survey. The research team administered a follow-up survey to study
participants to cover the 18 months following RA. All study participants were included in
the survey sampling frame and interviews were completed with 82.3 percent of them
(82.2 percent of the program group and 82.6 percent of the control group). The survey
covered pre-RA characteristics (e.g., demographics, criminal history); services received
6
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

since RA (whether from the SCA program or other sources); and outcomes. The latter
includes recidivism (arrests, convictions, and incarcerations), employment (whether
worked since RA, whether currently employed, wages), health status, housing status,
family status, substance abuse, payment of child-support obligations, and other topics.
Appendix B presents more detail on survey administration.
Having multiple data sources allows us to take advantage of the best characteristics of each. For
example, administrative data provide an objective source for measuring key outcomes and are
not subject to recall error or respondent reporting bias. On the other hand, survey data cover a
much broader set of outcomes and provide greater depth about each topic. Administrative and
survey data used together provide the opportunity to corroborate key findings using
independent sources of evidence.

Estimating Program Impacts
This report presents the estimated impacts of the grantees’ programs measured for the 18
months after each individual’s date of random assignment.
General Approach
The study uses an intent-to-treat framework in that we compare the outcomes of those
randomly assigned to the SCA program group to the outcomes of those assigned to the control
group. RA is considered the “gold standard” for estimating program impacts because it is the
best way of ensuring that there are no pre-existing differences between the program group and
those to whom they are being compared. Through RA, we can assume that program-group
members are, on average, like those in the control group on observable and unobservable
characteristics; for example, they are not more motivated than those in the control group and
the two groups will have similar criminal histories and criminogenic needs. Because of the preRA equivalence between the groups, estimation methods can be relatively simple: we can
attribute the mean difference in the outcomes between the groups to the effects of being
assigned to SCA.7
We built off this simple approach in several ways. First, we have weighted the sample to
account for the fact that the probability of assignment to the program group is not constant
across the grantee sites. The weight used is the inverse of the probability that an individual was
assigned to his or her observed study group. When using survey data, we also weighted

By contrast, alternative approaches under the broad category of quasi-experimental designs use statistical
methods to define a comparison group to which the outcomes of the program group can be compared. Their
disadvantage is that one cannot confidently rule out the possibility that any observed difference in outcomes
between the groups is due to unobserved pre-existing differences rather than being the effect of the
intervention.
7
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

analyses to account for potential nonresponse bias (details of the survey weighting are
described in Appendix B).
Because RA effectively neutralizes the impact of pre-existing characteristics, we calculate
impacts as the simple difference in means between the program and control groups. However,
we calculate whether these differences are statistically significant by using ordinary least
squares regression models (for outcomes that are continuous variables) or logit models (for
outcomes that are dichotomous), which take into account individuals’ observed baseline
characteristics, such as gender, age, and criminal history. Observed mean differences in
outcomes provide an unbiased estimate of the treatment effect, but regression adjustment
improves statistical precision by reducing the variance of the estimates.
This approach is used predominantly. However, some outcomes (e.g., date of first re-arrest
following random assignment) are based on elapsed time to an event. For outcomes of this
type, we used survival analysis, which is more appropriate for analyzing duration data.
We also conducted additional analyses that are refinements to this general approach to test the
sensitivity of the results to alternative model specifications. For example, we estimated
hierarchical linear models that take into account the fact that study participants are nested
within grantees. These models yielded very similar conclusions to the ones from the simpler
models just described; to avoid needless complexity, the simpler models are predominantly
used in this report. Statistical methods and results from the additional models are described in
Appendix C.
Subgroup Analysis
We estimated impacts for the full sample, but also separately for subgroups that were deemed
a priori to be of substantive or policy interest. The subgroup analysis is designed to “unpack the
black box,” by identifying whether impacts varied depending on the types of participants served
or program design features. These subgroups were of three types: one type was based on preexisting characteristics of participants, a second was defined based on a key program design
feature, and a third was based on grantee characteristics.
Subgroups based on participant characteristics. Prior research has shown that the risk of
recidivism and the impacts of re-entry services may be different for different subsets of the
formerly incarcerated (e.g., Lipsey and Cullen 2007). Based on this research, we have identified
the following key subgroups, each defined by study participants’ pre-RA characteristics:
•

Gender. Adult men have significantly different criminal behaviors than do adult women
and are at higher risk of recidivism. Moreover, women have very different criminogenic
needs than men and face different transition challenges, suggesting the need for reentry services that are gender specific (Berman 2005, Bloom et al. 2003). A program’s

8
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

ability to respond to these needs may mean that these programs were more or less
effective for women than men.
•

Age. Although explanations for the relationship abound, it has been well established
that crime rates peak in early adulthood and decline steeply thereafter (Hirschi and
Gottfredson 1983). Further, interventions aimed at increasing desistance can be more
effective for those who are older (Uggen 2000). To test whether the programs were
more effective for study participants of different ages, we defined two subgroups: those
less than age 30 and those ages 30 or older.

•

Risk of Recidivism. Gender and age are two well established predictors of recidivism, but
there are others, including criminal history and dynamic factors that are indicators of
criminogenic need. Some researchers have found that interventions can be more
effective for higher-risk individuals and that, in fact, programs targeted to those at
lower-risk can increase failure rates in some instances (Latessa and Lowenkamp 2006,
Lipsey and Cullen 2007).8 The study’s SCA grantees determined risk by using validated
assessment instruments (see Chapter II). We do not have access to those scores, but
instead follow the example of Kemple and Snipes (2001) in using simulations estimated
on the control group to divide the sample into lower-risk and higher-risk individuals (see
Appendix C for details). All those eligible for SCA were supposed to be at medium or
high risk of recidivism, so this classification represents a relative ranking within a
truncated range.9

Subgroups based on program design. Research also shows that recidivism is highest shortly
after release from incarceration (Durose et al. 2014), suggesting that interventions can be most
effective if they are applied before the transition from incarceration to release rather than after
release (Petersilia 2003). The SCA grant solicitation recognizes this, by defining successful reentry as something that requires “delivery of a variety of evidenced-based program services in
both a pre- and post-release setting designed to ensure that the transition from prison or jail to
the community is safe and successful” (U.S. Department of Justice 2009, p. 2). Accordingly, we
define a subgroup that captures the potential importance of the timing of SCA enrollment.
•

Timing of SCA Entry. To measure the possibility that the programs’ impacts are greater
when individuals are enrolled well before release rather than later, we define two

However, see Wilson and Zozula (2011) as an example of an evaluation that found contrary evidence.

As a condition of their grants, the study grantees were to target those at medium or high risk of recidivism.
Thus, the risk scale we created does not capture the full variation of risk level across the population of adults
who are scheduled for release from incarceration, but merely provides a relative ranking (based on static risk
factors) of individuals in this sample.
9
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

groups: those randomly assigned at least 31 days before release and those randomly
assigned no more than 30 days prior to release or after release.10
Subgroups based on grantee type. Some grantees were associated with the criminal justice
system (i.e., DOCs or a sheriff’s office), while others were health or social services agencies. For
the first group, the key point of contact for participants accessing SCA services after release was
generally a probation or parole officer (PO), whereas health and social services agencies
assigned a case manager apart from the PO. This difference had important implications for the
way re-entry services were delivered (D’Amico and Geckeler 2014) and represented a
fundamental difference in program designs across the seven grantees in this study. Further, the
two categories of agencies had different types of pre-existing partnerships, which had
implications for their ability to leverage resources for different kinds of re-entry services (these
differences are discussed in the next chapter). Accordingly, we define the subgroup below.
•

Grantee Type. This categorization divides the grantees into those that were state DOCs
or local sheriff’s offices versus those that were local social services or health agencies.

Confirmatory and Exploratory Analyses
We measure impacts on a range of outcomes relating to services received, recidivism,
employment, family stability, and others. These impacts are measured for the full sample and
for the various subgroups described above. With so many comparisons, at conventional
thresholds for determining statistical significance we are likely to find some impacts simply by
chance (that is, even if true impacts are zero). This is known in the literature as the multiple
testing problem.
Statistical adjustments that have been proposed for dealing with multiple testing typically
reduce the threshold for determining statistical significance. These approaches decrease the
likelihood of false positives (that is, of claiming that there is an impact when in fact the
difference between groups occurred purely by chance). However, as a consequence, these
methods reduce statistical power and increase the likelihood of false negatives (that is, of
failing to conclude that a difference between groups is real even when it is).
To avoid this loss of statistical power, we adopt an approach recommended by a panel of
experts (Schochet 2008) that treats a main analysis as confirmatory and other analyses as
exploratory. Our confirmatory analysis considers re-incarceration for the full sample anytime
within 18 months after random assignment as the main outcome of interest, and considers

Generally, the date of RA is coincident with or very shortly before the date of SCA enrollment for those study
participants assigned to the program group. Note that even when RA occurred after release, it does not mean
that the individual did not receive pre-release services; however, these services would have been provided
apart from enrollment in SCA.
10
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

other analyses as exploratory. Further, we focus on patterns of effects rather than isolated
impacts.

Sample Sizes and Statistical Power
Statistical power refers to the ability of a significance test to confidently detect an effect when
in fact an effect exists. Among the factors that determine statistical power, two of the most
important are the study’s sample size and the size of the effect one is trying to detect. In a
study with 966 study participants, split unevenly between the program and control groups, we
can confidently detect a difference between the program and control groups on a binary
variable if the difference is at least nine percentage points.11 Thus, this study is powered to
detect effects that are approximately that large.
Some analyses for this study are conducted on smaller sample sizes and therefore will have
weaker statistical power. For example, results estimated from survey data are based on the 82
percent of study participants who responded to the survey. Additionally, some analyses are
conducted on subsets of participants, and there is a very modest amount of missing data on
some items.12

Limitations of the Study
The impact estimates presented later in this report represent differences in outcomes
experienced by those randomly assigned to the SCA program group in comparison to those
assigned to the control group. Although the research design ensures a rigorous and unbiased
estimate of intent to treat, the interpretation of findings is subject to certain limitations and
cautions.
•

Control-group members were allowed to access re-entry services. Those randomly
assigned to the program group were able to access the full range of SCA services, while
those assigned to the control group could access other re-entry services but not SCA
services. The study thus represents a comparison of the effectiveness of SCA in these
sites relative to services otherwise available, and not in comparison to no services
whatsoever. In service-rich environments, control-group members could have accessed
significant services, even ones very comparable to SCA, from other sources.

This calculation assumes a 95 percent confidence level for a two–tailed test, an 80 percent level of power, that
the outcome variable has an observed value of approximately 50 percent (the worst-case scenario), and that
the test of the difference is run unweighted without controlling for covariates. A change to any of these
parameters can change the minimum detectable difference (MDD); for example, MDDs will be considerably
better for variables that are more skewed.

Results for an outcome calculated from administrative data are not reported if more than five percent of the
group that should have data is missing on that item.
11
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

•

Some SCA funds were spent on control-group members. Two of the grantees
participating in this study used a portion of their SCA funds for general system
improvements that could have benefited all those returning from incarceration
(including the control group) to some degree. For example, they used a portion of their
funds to improve prison programming, such as by modifying pre-release classes or
workshops that all those who were incarcerated could access on an equal footing
whether or not they were SCA eligible. Because these changes were general system
improvements, it was not practical to deny control-group members access to them.13
The study thus captures the effect of the personalized services that SCA provided, but
not the general system improvements.

•

The individual SCA program models varied in important ways. Because of small sample
sizes in each grantee site, it is not practical to estimate grantee-specific impacts.
Accordingly, for our main analyses, we pool observations across the seven grantees.
However, as will be described in more detail in Chapter II, grantees used their SCA funds
to implement somewhat different program models.

•

We cannot generalize findings beyond the study sample. The seven grantees included in
this study were purposively selected by BJA from a larger group of 15 grantees that
received FY 2009 funding, because BJA believed that the seven were best able to
participate in a rigorous evaluation. Because the grantees were purposively selected, we
cannot generalize findings to the larger pool of FY 2009 grantees. Moreover, BJA made
SCA awards in subsequent fiscal years and under different categories of competition; we
cannot generalize this study’s results to those other grantees.

•

Outcomes are measured imperfectly. As is inevitable with studies of this nature,
outcomes are not always measured with perfect accuracy. For example:


Recidivism data provided by state and local agencies were collected only from the
jurisdictions in which individuals were most likely to have involvement—arrest,
conviction and prison incarceration data were collected from the states in which
these SCA programs operated, and jail incarceration data from the counties to
which study participants were expected to be released. This means that the
administrative data used in the study miss involvement with the criminal justice
system that occurred outside these jurisdictions.14

Only modest amounts were spent on general system improvements and, in most sites, no funds were spent in
this way at all.

Because we do not have data from other jurisdictions, this study likely under-reports recidivism to some
degree. We attempted to access recidivism data through the National Crime Information Center (NCIC)
maintained by the Federal Bureau of Investigation (FBI) to overcome this limitation. This would have provided
national coverage and a more uniform source of recidivism data. The FBI denied the data request, so this
approach was not possible. However, even though recidivism is measured from separate state and local
sources, it is measured consistently for program and control group members in each jurisdiction, which
minimizes the effect of potential measurement bias on the estimates of program impacts.
12
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice



Administrative data do not fully measure all criminal activity that occurred.
Administrative data only capture events available in the state and local agencies’
data systems; some events may have occurred without being recorded (for
example, if the event did not come to the attention of the criminal justice system),
and some records may have been expunged.



Matching with criminal justice databases could be incomplete if the identification
information the study participant provided on the BIF were faulty or incomplete.
As a result, criminal justice data may be missing for some individuals.15



NDNH does not capture employment and earnings from self-employment and
selected other sources.



Survey data are self-reported and are therefore subject to recall or other
respondent bias.

Fortunately, key outcomes are measured using both survey and administrative data,
providing an opportunity to test the robustness of conclusions. Furthermore, outcomes
are measured in the same way for both the program and control groups, which
minimizes the role of reporting bias on the estimation of impacts.

About Study Participants
To be eligible to participate in the SCA Adult Demonstration Program, individuals had to:
•

Be 18 years of age or older,

•

Be convicted as an adult,

•

Have been imprisoned in a state, local, or tribal prison or jail, and

•

Be classified as being at medium or high risk of recidivism.

Within this pool, grantees were expected to identify the specific subset that their programs
intended to target, which could include, among others:
•

A specific demographic group (e.g., based on age or gender)

•

Those returning to a specific community or neighborhood.

Once determined eligible for SCA participation according to whatever criteria each grantee
established, the individual was provided an orientation to SCA program services and to the

We assumed that everyone in the sample should have been arrested, convicted, and incarcerated in the ten
years prior to the RA date. If an individual had no evidence of arrest, conviction, or incarceration in this period
based on the administrative data we were provided, we assumed that the agency could not successfully match
this individual to its records. In these cases, the corresponding measures of recidivism in the post-RA period
were set to missing. This is a conservative assumption that prevents us from falsely assuming recidivism did not
occur when in fact there was a problem with the match. Fortunately, the incidence of missing data of this
nature is very small.
13
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

random assignment study. Those who agreed to participate in both the program and the study
were randomly assigned using an online random assignment system maintained by the study
team, thus becoming study participants. Those who did not provide written consent to
participate in the study were not randomly assigned and could not enroll in SCA (see
Appendix A).
As was discussed, 966 individuals are included in the study, with 606 (62.7 percent) assigned to
the program group and 360 (37.3 percent) assigned to the control group.16 Exhibit I-3 shows key
characteristics of program- and control-group members measured at the time of RA. By virtue
of random assignment, we would expect those in the program and control groups to have very
similar characteristics on average, and the exhibit shows that indeed they do. Results show
that:
•

Approximately 80 percent of study participants in each group are male.

•

Approximately half are white and one-third are African-American.

•

Approximately half of study participants in each group are 30 years old or less, and
approximately one-fourth are older than 40.

•

Approximately one-fourth had not obtained a high school diploma or general
educational development degree (GED), and just under half achieved a GED. Very small
percentages attended college.

•

Nearly all had been employed at some time in their lives prior to RA. Approximately half
were employed at the time of incarceration that preceded RA, usually full time, and the
remaining half were not employed.

•

Just over 10 percent had a disability (self-reported and defined as a condition limiting
one’s physical activity or kind of work).

•

Nearly all spoke English as their primary language.

Importantly, there are almost no statistically significant differences between the program and
control groups on the characteristics shown here. The one exception is the modest difference in
the percent who worked sometime prior to RA.
Exhibit I-4 reports the criminal history of study participants in the period before RA. Programand control-group members were arrested and incarcerated a similar number of times in the
period prior to RA, and their offense categories are comparable. The length of their most recent
prior sentence was comparable, with just over one-half of both groups serving more than a
two-year sentence. One difference is that 88 percent of program-group members were
randomly assigned while still incarcerated, while the figure for the control group is 83 percent.

In actuality, 973 individuals were randomly assigned. However, after random assignment, one grantee lost the
signed consent forms for seven individuals. These individuals were dropped from the study.
14
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Taken as a whole, these findings demonstrate that, despite some minor differences that
occurred by chance, random assignment succeeded in defining two equivalent groups.

Roadmap to the Report
The remaining chapters describe the SCA programs that are being studied and present the
impact estimates.
•

Chapter II presents a summary of findings from the implementation study. It covers the
amounts of SCA Adult Demonstration funding the grantees received, their grant
management and partnerships, their targeting and enrollment practices, and the
services they provided with their SCA funds. It also discusses how SCA changed the
typical services that were available.

•

Chapter III presents results tabulated from the grantees’ MIS data, showing the types of
services that SCA program-group members received and their duration of participation.

•

Chapter IV presents impacts on the services that study participants received. It covers
differences between the program and control groups on receipt of case management
services, cognitive change therapy, employment and education services, housing
assistance, substance abuse treatment, and other services.

•

Chapter V covers impacts on recidivism. It measures arrests, convictions, and jail and
prison incarcerations, with alternative measures created using survey data collected
from study participants and administrative data collected from state and local criminal
justice agencies.

•

Chapter VI presents impacts on employment and earnings, using both survey data and
NDNH data.

•

Chapter VII presents impacts on a range of other outcomes, including physical and
mental health, self-reported substance abuse, housing stability, and ability to meet
child-support obligations.

•

Chapter VIII presents a summary and conclusions.

The chapters presenting impacts (i.e., Chapters IV through VII) are structured in a similar way:
we first present impacts for the full sample, and then for the various subgroups identified
earlier in this chapter. Appendices present technical material, including details on random
assignment, the survey administration and weighting, and estimation techniques.

15
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit 1-3: Background Characteristics of
Program and Control Groups
Program

Control

Difference

Gender
Female
Male

21.8
78.2

19.9
80.1

1.9
-1.9

Race and Ethnicitya
White
Black
American Indian/Alaska Native
Hispanic
Hawaiian Native/Pacific Islander
Asian

52.3
31.2
13.2
10.2
1.8
0.9

49.0
33.8
15.6
9.2
2.4
1.7

3.3
-2.7
-2.4
1.1
-0.6
-0.8

Age
18 to 21
22 to 25
26 to 30
31 to 35
36 to 40
41 to 50
51 or more

8.4
17.8
23.7
15.7
8.2
18.6
7.5

9.3
20.6
23.3
12.9
11.2
17.6
5.2

-0.8
-2.8
0.4
2.8
-2.9
1.0
2.3

Highest Degree Attained
Less than high school degree or GED
GED
High school diploma
Some college
Employment-Related

25.0
44.9
24.4
5.7

23.2
43.4
27.1
6.2

1.8
1.4
-2.8
-0.5

Worked sometime in the past

93.0

88.8

Employment status at time of most
recent incarceration prior to RA
Was employed full time
Was employed part time
Was not employed
Other Characteristics

32.7
14.4
52.9

33.3
15.4
51.3

-0.6
-1.0
1.6

Has a disability

13.6

11.6

2.0

English is primary languageb

98.7

98.8

-0.1

Demographic Characteristics

4.2**

Notes: Numbers in the first two columns represent the percentage of study participants with the
characteristics in question; the third column represents the difference between the two (program-group
value minus control-group value). Estimates were weighted to equalize the odds of selection into the groups
and, where appropriate, to account for potential survey response bias.
Source: Baseline Information Form, except where noted.
a

The sum across the categories exceeds 100 percent, because some individuals indicated being of more than one
race or being of Hispanic ethnicity and at least one race.
b

Based on survey data

*/**/*** Statistically significant at the .1/.05/.01 level.

16
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit 1-4: Criminal History of
Program and Control Groups
Program

Control

14.8
21.4
26.3
37.4

15.2
23.3
25.4
36.1

-0.4
-1.9
0.9
1.4

Most serious arrest offense in the 10 years prior to RAa
Violent
52.0
Property
35.4
Drug
10.6
Public order
2.0

52.0
33.5
12.7
1.8

-0.0
1.9
-2.1
0.2

Number of separate times incarcerated in
prison or jail any time prior to RAb
1 time
2 to 4
5 or more

11.1
38.3
50.5

13.3
34.2
52.5

-2.2
4.2
-2.0

Type of crime for which most recently
incarcerated prior to RAb #
Violent
Property
Drug
Public order

19.8
34.5
43.9
26.9

19.5
29.9
49.5
26.9

0.2
4.6
-5.5
-0.0

Length of current or most recent sentence prior to RAb
Less than 90 days
3.6
At least 90 days but less than 6 months
6.6
At least 6 months but less than 12 months
14.1
1 year to 2 years
21.0
More than 2 years
54.7

4.8
7.3
13.2
20.2
54.5

-1.2
-0.7
0.8
0.8
0.3

Total days incarcerated in prison or jail in 10
years prior to RAa
Up to 1 year
1 to 3 years
3 to 5 years
More than 5 years

28.4
36.0
16.9
18.7

25.8
33.7
19.7
20.8

2.6
2.3
-2.8
-2.1

Incarcerated on the date of RAc

87.8

83.1

Number of separate times arrested in the 10
years prior to RAa
1 or 2
3 to 5
6 to 10
11 or more

Difference

4.6**

Notes: Numbers in the first two columns represent the percentage of study participants with the
characteristics in question; the third column represents the difference between the two (program-group
value minus control-group value). Estimates were weighted to equalize the odds of selection into the groups.
Types of crime were coded according to Durose et al. (2014).
Source: a=Administrative data; b=Baseline Information Form; c=Both administrative data and the study’s
random assignment system.
#

The sum across the categories exceeds 100 percent, because multiple types could have been recorded.

*/**/*** Statistically significant at the .1/.05/.01 level.

17
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

II.

About the SCA Programs

In 2013, SPR prepared an implementation study report that described the SCA programs
developed by the seven impact-study grantees (D’Amico et al. 2013).17 This chapter summarizes
the findings of that report. It describes the overall organization of these seven grantees and the
funding that supported them, their administrative structures and partnerships, their service
models, the ways the grantees screened and enrolled participants, and the types of services
that the programs provided. Finally, the chapter describes services that control-group members
may have had available and how these services compared to those available to the program
group.

Grant Funding and Use of Funds
BJA first awarded SCA Adult Demonstration funding to the study’s grantees in FY 2009. It also
awarded supplemental funding in FY 2010 and FY 2012, contingent upon the grantees’
participation in the impact study. Exhibit II-1 identifies the seven grantees’ grant amounts.
However, funds for SCA programming were in actuality much greater than the amounts shown
in the exhibit, because, following statutory requirements, the BJA FY 2009 grant solicitation
specified a 100-percent matching requirement. At least 50 percent of the match needed to be
made up of cash, and the rest of in-kind contributions. Sources for the match could include
state or local government funds, grantee or partner contributions, or other public or foundation
funds.
Exhibit II-1:
Grantees and their SCA Adult Demonstration Grant Awards
Grantee

Total

FY 2009

FY 2010

FY 2012

Allegheny County

$2,653,339

$608,339

$825,000

$1,220,000

Kentucky

$3,250,000

$750,000

$1,000,000

$1,500,000

Marion County

$1,502,768

$302,768

$400,000

$800,000

Oklahoma

$3,250,000

$750,000

$1,000,000

$1,500,000

San Francisco

$2,600,000

$600,000

$800,000

$1,200,000

San Mateo County

$2,937,674

$677,674

$900,000

$1,360,000

South Dakota

$3,249,749

$749,749

$1,000,000

$1,500,000

Source: Data provided by the Bureau of Justice Assistance.

That report also included implementation study findings on three other grantees funded in FY 2009, but which
did not receive continuation funding and which are not part of the impact study.
19
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

BJA expected grantees to use their SCA grants to fill gaps in their community’s re-entry services.
Following this directive, grantees blended their SCA grant funds with their matched amounts
for three general purposes:
•

To provide service coordination for SCA participants. Much of the grantees’ funding was
used to provide individualized case management and service planning for SCA
participants.

•

To provide other services specifically or primarily for SCA participants. Grantees used a
portion of their funds to provide post-release services and, in some cases, pre-release
services that were restricted to SCA participants or for which SCA participants had
priority access.

•

To make general system improvements. Two of the grantees (Kentucky and Marion
County) used their SCA funds to make changes to, or provide partial support for, prerelease or post-release services that were generally available to anyone returning from
incarceration. For example, one grantee used a portion of its funds to modify prerelease classes that anyone in the institutions could access. Because these changes were
general system improvements, these services could have been accessed by those in the
control group and their effects are not captured by the impact study.

Exhibit II-2 shows, for each grantee, the types of activities that SCA grant funds supported. The
right-hand column of the exhibit highlights SCA program activities whose effects are being
assessed through the impact study; these include all SCA-funded activities to which SCA
participants had exclusive or priority access, but exclude the general system improvements in
the two sites mentioned above.
The exhibit also shows that, broadly, case management with service coordination was the focal
service for six of the seven grantees. As will be discussed, there were important differences
among the six; for example, they differed in which type of organization provided the case
management and which other services were funded with SCA grant funds. Nonetheless,
brokering services through case management was the centerpiece in these six sites.
The seventh grantee (Marion County) operated quite differently. Although it provided case
management, central to its program model was a structured set of classes that took place full
time Monday through Friday during the 12 weeks after release. Classes covered cognitive
behavioral therapy, employment assistance, substance abuse treatment, and life skills, among
other topics, and were provided either by the lead agency or through partners. Case managers
were available to meet individually with participants as needed, but, because classes were
structured and occurred virtually full time throughout the week, case management was more
peripheral.

20
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit II-2: Use of Funds and Focus of the Impact Study
Grantee
Allegheny
County

What SCA Grant Funded (in whole or part)
• Re-entry staff, who provided SCA participants with assessments and service coordination, both
pre-release and post-release, and developed participants’ re-entry plans, and family support
specialists, who worked to prepare families for the release of the inmate

• Pre-release and post-release classes, including job readiness training, family support, cognitive

Kentucky

behavioral therapy, and job placement assistance (SCA participants had priority for these
classes, but others could participate on a space-available basis)
• Reentry Parole Officers (RPOs), who provided parole supervision to SCA participants after
release; RPOs had smaller caseloads than regular POs, but reporting requirements were the
same

• Bus vouchers that were provided to RPOs for disbursement to SCA participants as needed

Captured by
Impact Study
√
√
√
√

• Prison re-entry coordinators, who coordinated home placements
• General system improvements, such as upgrading instruments used for assessments and
enhancing pre-releases classes, such as cognitive behavioral therapy and parenting
Marion County

• SOAR, a 12-week, full time, post-release course with modules addressing topics of cognition,
substance abuse, family support, job preparedness, and life skills; after the course, there are 12
weeks of "aftercare" classes, in 1-2 hour sessions

• Quest for Change House, a living facility for SOAR participants who needed housing

√
√

• Expanded operating hours for a resource center open to those released from incarceration (use is
not restricted to SCA participants)

• Enhanced "reach-in" classes, available to incarcerated individuals nearing release
Oklahoma

• Program specialists, who coordinated pre-release services for SCA participants

√

• Transition coordinators, who created transition plans

√

• Community specialists, who provided service coordination after release to those released without
supervision requirements; for SCA participants released with supervision requirements, service
coordination was provided by regular POs, who also were funded through SCA, with the
expectation that case management would be more proactive than normal

• Pre-release and post-release classes, including program slots for job readiness training,
vocational training, cognitive behavioral therapy, and substance abuse treatment; SCA
participants have priority access

• Vouchers for housing, transportation, and other needs after release

21
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

√

√
√

Exhibit II-2 (concluded)
Grantee
San Francisco

What SCA Grant Funded (in whole or part)
• Case managers from a community-based organization, who first met with SCA participants prior
to release to begin transition planning and were then available to meet regularly after release to
coordinate services; case managers were trained in Motivational Interviewing

√

• IRIS Center, an outpatient mental health and substance abuse treatment program

√

• SF Clean City, which offered work experience (street cleaning) with work readiness training to a

√

• Case managers from a community-based organization, who first met with SCA participants prior
to release to begin transition planning and were then available to meet regularly after release to
coordinate services

√

• Mentorship program available to SCA participants

√

• Post-release classes, including program slots reserved for SCA participants for job readiness

√

training, transitional employment, and substance abuse treatment, and transitional housing

• Transportation vouchers for SCA participants
South Dakota

√

• Homeless Prenatal Program, which helped participants navigate the child welfare system

small number of SCA participants
San Mateo

Captured by
Impact Study

√

• Post-release case management, provided by “enhanced” POs (intensive pre-release case
management is also provided strictly to SCA participants, but is funded by the match
requirement)

√

• Cognitive behavioral therapy, both pre-release and post-release

√

• Additional post-release services for participants with special needs (e.g., chemical dependency or

√

substance abuse treatment, transitional housing, health and mental health care)

• Supportive services for program participants, including transportation assistance

22
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

√

Managing the Programs and Partners
The SCA grantees were different types of organizations and relied heavily on partnerships for
delivering services. They worked with their partners using different service models.

Types of SCA Grantees
As is illustrated in Exhibit II-3, the grantees represented both different levels of government and
different types of government agencies.
•

Levels of government. Four grantees were county or municipal agencies, while three
were state agencies.

•

Type of organization. Four grantees were criminal justice agencies, including three state
departments of corrections and a sheriff’s office, while three grantees were agencies
responsible for administering health and human services programs.
Exhibit II-3:
SCA Grantees by Governmental Level and Type of Organization

State Government Agencies

Local Government Agencies

Criminal Justice

• Kentucky DOC
• Oklahoma DOC
• South Dakota DOC

• Marion County Sheriff's Office

Health and Human Services
•Allegheny County Department of Human Services
•San Francisco Department of Public Health
•San Mateo County Division of Health and Recovery Services

These agencies were the formal grant recipients, responsible for fiscal management and
reporting to BJA on the use of program funds and outcomes for participants. Most of them also
designated persons within their organizations to manage their SCA programs, with
responsibility for overseeing and coordinating implementation and operation. However, two

23
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

grantees (San Francisco and San Mateo County) subcontracted out much of the project
management and operational responsibilities to nonprofit organizations.

Service Models and Partnerships
The grantees were not capable of providing all program services themselves, so they depended
upon linkages with numerous partner organizations. With their partners, the programs
delivered services in three ways.
•

In the direct service approach, the agency that primarily operated the SCA program
provided the service.

•

With formal partnerships, the SCA program operator arranged with other providers
(typically 2 to 5 such partners per grantee) to deliver services. The grantees paid for
these services on a fee-for-service basis, in a lump sum to increase that provider’s
capacity, or through some other formal agreement with a provider to ensure that SCA
participants had priority of service.

•

Using informal partnerships, SCA staff members made unfunded referrals to various
community organizations (upwards of 10 such agencies per grantee) to deliver services.

These three approaches are profiled in Exhibit II-4. Each grantee used all three approaches, but
they varied in terms of which services they provided directly versus through informal or formal
partnerships.
Given limitations of available funding and the grantee’s own expertise, formal and informal
partnerships were used extensively by each grantee. These partnerships were built upon
previous re-entry efforts within the states or local areas, but existing partnerships were
strengthened and new partnerships were developed as part of the grant. Partnerships were
important in two distinct ways.
•

Partnerships strengthened grantees’ capacity for service delivery. Grantees created
formal agreements with new providers to deliver services and expanded their
knowledge of agencies to which they could make unfunded referrals. Some of these
partnerships were with community-based organizations, while others were with public
agencies, including those responsible for (among other things) public assistance, alcohol
and drug treatment, mental health services, and education. Overall, these partnerships
were important for building the capacities of grantees to serve participants during the
grant and into the future. These partnerships also helped agencies and organizations
share ideas and approaches to service delivery, expanding the perspectives and
knowledge of grantee and partner staff.

•

Partnerships allowed for increased coordination between pre-release and post-release
services. An increase in the continuity of services was generally viewed by program staff
as one of the more important successes of their grant activities. Through partnerships,
case managers who generally worked with participants after release were able to begin

24
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

meeting participants who were still incarcerated so that they could cement
relationships, gain participants’ trust and begin planning for post-release services. It also
allowed those who worked within jails or prisons to help share their knowledge about
participants with those who could continue care after release. Among grantees that
were correctional agencies, this strategy involved strengthening the relationships
between jail or prison staff and POs. For grantees that were health and human services
agencies, this coordination often led to new or stronger partnerships with correctional
system agencies.
Regardless of the context, maintaining quality partnerships required strong communication
among management staff for clarifying policies and procedures, and among service delivery
staff for communicating participants’ progress and needs.
Exhibit II-4:
Three Approaches for Delivering Program Services
Service Approach
Direct Service

Formal Partnership

Informal Partnership

Nature of
Agreement

Directly provided by the
organization operating
the SCA program

Grantee has formal
arrangement to
provide services

No specific terms or
agreement; SCA program staff
provide referrals

Treatment of
SCA
participants

The service is exclusively
for SCA participants

SCA participants given
priority over others

SCA participants are like all
others seeking services

Advantages

Specifically tailored to
SCA participants;
grantee controls access
and engagement

Provides SCA
participants with
priority access;
services coordinated
by the program

Most flexible, least costly, and
allows SCA staff members to
use any service provider
available in the community

Limitations

Grantee lacks resources
and expertise to deliver
all services directly

Typically costs the
program money

SCA participants have the
same access to services as
others; little formal follow-up
on participant involvement in
services

Screening and Enrollment
Grantees varied in their recruitment and enrollment procedures, including who was eligible to
enroll in SCA and how eligibility was determined.

25
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Subgroups Targeted for Project Participation
Chapter I described the SCA Adult Demonstration Program’s eligibility requirements, but noted
that grantees were expected to add their own targeting criteria.18 All the grantees used
location-based targeting and took risk of recidivism into account.
•

Location-based targeting. Eligibility was restricted to individuals being released to
specific cities, metropolitan areas, or counties.

•

Risk of recidivism. Each of the grantees targeted individuals assessed as being at
significant risk for recidivism. However, they used different assessment instruments and
drew the threshold for eligibility at different places.

Beyond these general similarities, grantees used evidence of service gaps in their communities
to define additional targeting criteria that differed among them.
•

Gender. Allegheny County, Kentucky and South Dakota allowed participation by both
men and women, while Marion County and Oklahoma served men exclusively, San
Francisco served women exclusively, and San Mateo County gave priority to women.

•

Age. Six of the grantees included adults of any age. South Dakota capped the age of
enrollment at 30.

•

Type of incarceration facility. Allegheny County and San Mateo County targeted
individuals scheduled for release from county jails. Oklahoma and Marion County
targeted individuals scheduled for release from state prisons. Kentucky and San
Francisco started targeting individuals released from prison but expanded their
enrollment to those released from jails. South Dakota targeted individuals from a range
of state prisons, county jails, and tribal detention facilities.

•

Expected time to release. Grantees had different ideas about when it was best to
screen, recruit and enroll participants. Allegheny County, Oklahoma and South Dakota
had more extensive pre-release services as part of their SCA programs, and typically
enrolled individuals more than three months prior to release. The remaining grantees
tended to rely on institutions’ existing programming for pre-release services and used
their SCA funds to concentrate on the post-release period; therefore, they enrolled
participants in SCA within a few months prior to release or, sometimes, after release.
The range of time between RA (a proxy for the SCA recruitment, screening and
enrollment process) and release is shown in Exhibit II-5.19

The study did not specify eligibility criteria. In order to be enrolled in the study, an individual had to be eligible
to participate in SCA based on whatever criteria each grantee used and give consent to be in the study.

Generally, random assignment occurred just after eligibility for SCA was established and just before intensive
and personalized SCA-funded services were expected to begin. The control group is used for the calculation in
the exhibit because the date of release is an endogenous variable for the program group, and can be affected
by SCA.
26
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit II-5:
Time Elapsed between Random Assignment and Release from Incarceration
RA occurred more than 6 months prior to
release

44%

RA occurred 3 to 6 months prior to release

11%

RA occurred within 3 months prior to
release

28%

RA occurred after release

17%
0%

10%

20%

30%

40%

50%

Notes: Numbers represent the time elapsed between the date of random assignment and the date of release
from incarceration for members of the control group. Four percent of those in the control group were never
released in the 18 months following RA; these individuals are included in the category of those randomly
assigned more than six months prior to release.
Source: Administrative data and the study’s random assignment system.

Assessment
Once targeting criteria were established, grantees identified and enrolled eligible individuals.
One key step in this process was relying on an assessment instrument to identify individuals
who met the risk threshold the program had established. Different types of instruments were
used for this purpose, varying in their complexity. They were administered either by SCA staff
members or correctional system staff members. The simplest such instrument was a proxy
indicator constructed from age at first arrest, number of arrests, and age. Other instruments
included the Level of Service Inventory-Revised (LSI-R), which scores potential participants on
54 risk items; the Level of Service/Case Management Inventory (LS/CMI), which includes a case
management module; and the Correctional Assessment and Intervention System (CAIS).
Once risk levels were initially determined, SCA staff members went about identifying eligible
individuals. Many grantees began by generating lists of potentially eligible participants from the
full roster of inmates on a weekly, semi-weekly, or monthly basis. Additionally, grantees that
enrolled some participants after release distributed flyers about SCA or relied on word-ofmouth or referrals from partners, especially probation and parole, to generate interest. Then,
staff members conducted information sessions, made final eligibility determinations, and

27
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

obtained consent to participate.20 Some grantees conducted eligibility screening and orientation
sessions on an ongoing basis, whereas others held orientation sessions only when program
slots opened up. Despite some initial concerns expressed by grantees about their ability to
meet impact-study recruitment targets, only one had difficulty meeting its enrollment target
and its initial target was revised down. The other six grantees met their enrollment targets.
Assessments were used not only to establish program eligibility but also to customize services
and update the service plan over time. Assessments were administered periodically for the
latter purpose and sometimes different instruments were used at different times. For example,
one grantee used a proxy indicator (based on age, age at first arrest, and number of arrests) to
establish program eligibility, but then, once
the individual was enrolled in SCA,
Examples of Assessment Instruments
administered the LSI-R to develop a re-entry
Used by Grantees
plan. Another used LSI-R at the outset, but
• Addiction Severity Index
also used Starting Point: My Personal
• Correctional Assessment and Intervention
Assessment for ongoing case planning. Still
System
another used CAIS to establish eligibility, but,
• Level of Service Inventory-Revised
for service planning, supplemented it with the • Level of Service/Case Management Inventory
Addiction Severity Index (ASI) and the
• University of Rhode Island Change Assessment
Scale
University of Rhode Island Change Assessment
Scale (URICA), among other instruments.

Case Management Services
As noted, case management was the cornerstone of the SCA program model for six of the seven
grantees. The goal of case management was to help prevent recidivism by providing
individualized support and coordinating services based on identified needs and risk factors and,
in some cases, by promoting compliance with parole or probation terms. Case managers
functioned as mentors, enforcers, and brokers of SCA program services, and ensured that
participants’ risk assessments were used to guide service plans. While many services accessed
by SCA participants were not reserved exclusively for them, case managers endeavored to
make service access more likely, more efficient, and in keeping with a holistic view of
participants’ situations and needs.
As shown in Exhibit II-6, case management differed across the grantees in two key ways. One
distinction was when SCA case management began: all seven grantees assigned participants to
20

The study required each grantee to offer eligible individuals the option of declining study participation before
random assignment occurred. Prior to the study, SCA program participation had sometimes been mandatory.
The introduction of the impact study made SCA participation optional.
28
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

a case manager who worked with each participant after release, but six grantees also provided
some form of case management prior to release with SCA funding. The pre-release aspect of
the work generally involved having the case managers work with participants on transition
planning, although in a few cases it went beyond that by including the coordination of prerelease services.21
Exhibit II-6:
Number of Grantees Providing Pre- and Post-Release SCA Case
Management from Various Sources

• Provided by corrections staff (3 grantees)
Provided Pre-Release
Case Management

• Provided by social services agencies (3 grantees)
• Not provided as part of SCA (1 grantee)

• Provided by corrections staff (4 grantees)
Provided Post-Release
Case Management

• Provided by social services agencies (3 grantees)
• Not provided as part of SCA (none)

Source: Site visits conducted as part of the evaluation.

A second way in which case management services varied across grantees was the position or
role of the individual who served as the case manager. Correctional system grantees were more
likely to provide pre-release case management through existing jail or prison staff, and postrelease case management through “enhanced POs”—that is, POs that might have had special
training and reduced caseloads. In contrast, grantees that were health and human services
agencies designated separate individuals as SCA case managers. These case managers came
from more traditional case management backgrounds (i.e., social workers, psychologists) within
the social services system, and generally also had prior experience working with the formerly
incarcerated. They developed transition plans, provided post-release case management, and

Program participants in the one SCA program without pre-release case management worked with prison or jail
staff around service planning, but these efforts were not coordinated with the SCA case manager or supported
through the grant. Thus, they were no different from what any incarcerated individual would have received.
29
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

sometimes coordinated with correctional system staff members around the delivery of prerelease services or worked with participants while still incarcerated. Participants served by
these grantees were very likely required to report to a PO after release, but this individual was
separate from the SCA case manager.22
The two approaches to staffing the SCA case management function had both advantages and
disadvantages. On the one hand, blending the roles of formal supervision and case
management under an enhanced PO could help head off potential conflicts between POs and
separate case managers, as occurred in at least one program where these two roles were split
across two individuals. Also, having a PO as a case manager was beneficial for promoting
participant retention in SCA because project participants knew they might face revocation if
they did not show up for post-release appointments and services. On the other hand,
participants often had negative perceptions of POs due to their own past associations with the
criminal justice system, and this bias could adversely affect their ability to fully benefit from SCA
(and from case management specifically).
Another challenge was that correctional system staff, either enhanced POs or other
correctional system staff working with the program, sometimes needed to embrace a new
approach to their work, one focused more on rehabilitation and less on enforcement. This
latter point was a significant growing pain for some DOCs, but led to an important cultural shift,
or change in mindset—a point that was echoed across respondents in the implementation
study. This shift involved having correctional system staff members downplay the prevailing
view that their role was about monitoring and enforcing and instead embrace a rehabilitative
philosophy designed to support participants throughout the re-entry process using evidencebased approaches. To support them in this transition, the grantees typically gave these
enhanced POs special training on the use of risk assessment tools and skills, or techniques, such
as motivational interviewing. They also had smaller caseloads, typically around half to a third of
their typical load, meaning they could meet both longer and more frequently with participants
than could traditional POs.

Other Services
In addition to case management, each grantee coordinated and delivered other services that
participants needed.

According to the participant survey, approximately 97 percent of study participants served by health or social
services agencies were on probation or parole after release from incarceration.
30
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Types of Services
BJA established seven categories of services that grantees were expected to provide and on
which they were to report participant involvement. These seven categories are outlined in
Exhibit II-7. Although all grantees offered access to each of these services, they did not provide
identical packages of services to SCA participants and did not deliver the services in the same
way. As with case management, the type of grantee had some influence on how services were
delivered. While criminal justice agency grantees had strong relationships within the criminal
justice system, they often had to forge more new relationships with agencies to ensure the
availability of the full range of services that participants might need after release. Health and
human services agencies, on the other hand, often had to work hard to develop partnerships
with correctional agencies, but had more community partners in place. They also provided
many services themselves.
Exhibit II-7:
Seven Categories of SCA Program Services
Category

Service Description

Education and Training

GED preparation and testing, vocational training, and community
college education

Employment Assistance

Job search and placement assistance, employment opportunities,
soft-skills training, and resume and interviewing skills
development

Substance Abuse
Treatment

Intensive, outpatient, 12-step or change-model substance abuse
treatment administered by licensed specialists

Mental Health Services

Mental health screenings and referrals to mental health services

Cognitive Behavioral
Therapy

Psychotherapeutic approach that addresses dysfunctional
emotions, maladaptive behaviors/cognitive processes and
contents through a number of goal-oriented, explicit systematic
procedures

Pro-Social Services

Stress and anger management services, peer support, leisure
activities, family and parenting classes, and mentoring

Housing Assistance and
Other Supportive
Services

Subsidized housing, housing placement services, vouchers for
food, transportation, and other needs

31
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Another important element of variability is that not all participants within a program received
the same mix of services. SCA program staff placed a strong emphasis on “needs-based” service
planning; therefore, participants accessed only some of the services a grantee offered, based
upon that individual’s service plan.23

Delivering Services to Participants
Grantees provided some services (in addition to case management) themselves and used
partners to deliver other services. Overall, grantees made widespread use of informal
partnerships and made limited and more targeted use of the direct service and formal
partnership approaches—both of which required the expenditure of SCA funds. Exhibit II-8
shows how grantees used these latter two approaches for pre-release and post-release
services. In this exhibit, a solid dot signifies a service that the lead agency provided directly
while a hollow dot signifies a service provided through a formal partnership. Absence of a dot
means that the service was delivered primarily through informal partnerships or existing
institutions.
As the exhibit shows, with the exceptions of Allegheny County and Oklahoma, grantees relied
more often on existing systems to provide services for participants still in jail or prison. SCA
program staff members gave two main reasons for relying so heavily on existing systems for
pre-release services: services of various types were generally already available as part of
existing jail or prison programs, and it could be difficult to integrate unique SCA service
components into the jail or prison environment.
For the provision of post-release services, by contrast, grantees commonly used the formal
partnership models to provide the range of services, with some use of direct delivery for
housing, cognitive behavioral therapies and pro-social services. Even when services were
provided directly or through formal partnerships, limited SCA grant funding may have led the
grantee to make extensive use of informal partnerships to increase capacity.

Services for the Control Group
Critical to the impact study is that SCA services were distinct from whatever services the control
group may have received. Therefore, in addition to describing SCA programs and the services
grantees provided, the implementation study learned about the services that were available to
control-group members and the extent to which these differed from SCA program services.

The grantee that provided the 12-week structured course is to some degree an exception, in that all
participants generally were expected to attend the same classes during the 12-week structured program.
32
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Housing

○

Pro-Social
Services

Substance
Abuse Treatment

○

Cognitive
Behavioral
Therapy

Employment
Assistance

○

Mental Health
Services

Education and
Training

Exhibit II-8: Pre- and Post-Release SCA Services Delivered
by Direct Service or Formal Partnership

Pre-Release

○

Kentucky

(Did not coordinate any pre-release services in SCA)

Marion County

(Did not coordinate any pre-release services in SCA)

Oklahoma

○

●/○

○

●

San Francisco

○

San Mateo

(Not applicable)

Allegheny

●

South Dakota
Post-Release

○

Kentucky

○
○

Marion County

○

Allegheny

Oklahoma

○

San Mateo

○

San Francisco

○

○
○

○

○
○

South Dakota

●/○

○
●/○

○

●

○

●

Notes: ● denotes services that the grantee provided directly; ○ denotes services provided through
formal partnership. Other services were provided through informal partnerships or existing
institutions.

Pre-Release Services for the Control Group
As was shown in Exhibit II-5, 83 percent of participants were randomly assigned while still
incarcerated, and 55 percent spent more than three months in jail or prison before being
released. Program-group members were able to use this time to access SCA-provided services
in addition to what was provided within the jails or prisons. For the most part, this pre-release
period primarily involved having SCA case managers work with participants to develop
transition plans, which included navigating and taking advantage of existing prison or jail
services and preparing for the period after release. While control-group members often had
access to a correctional system staff person who was responsible for helping inmates navigate
in-facility services, the time and attention these staff members had to assist control-group
members was generally less than what was available to assist program group members.

33
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

In some sites, and most especially Allegheny County and Oklahoma (see Exhibit II-8), SCA funds
were also used to develop in-institution workshops or other courses to which the SCA program
group was given priority access. For example, in Allegheny County, services in the jail could
include drug and alcohol treatment, cognitive behavioral therapy, education, mentoring, family
support, and employment planning. Court-ordered inmates and SCA program participants were
given priority access. Others (including control-group members) could participate only on a
space-available basis. With the exception of Allegheny County and Oklahoma, though, most ininstitution programming (apart from case management) was equally available to program and
control-group members, so the service differential prior to release was not that great.

Post-Release Services for the Control Group
Although the incidence of release to supervision varies markedly by state (Pew Charitable
Trusts 2014), most participants in this study were assigned to a PO after release from prison or
jail. As described above, grantees assigned program-group members either to an enhanced PO
(in lieu of a traditional PO) or to a social services case manager (to complement the work of a
traditional PO). This means that most study participants—program and control-group
members—should have received some form of “case management” after release, whether it
was from a PO, a social service agency, or both.
However, case management provided to control-group members by traditional POs was
expected to be less extensive and of a different quality than what SCA case managers (including
enhanced POs) typically provided.
•

SCA caseloads were smaller. Grantee and correctional system staff indicated that
traditional POs had caseloads that were approximately two to three times greater than
SCA case managers. Staff indicated that SCA case managers were available to meet with
participants weekly or biweekly (and more often, as needed), for about 30-60 minutes.
Traditional POs reportedly met with control-group members for 15-30 minutes once or
twice a month. This difference was further magnified in some areas, like California,
where AB 109 and AB 117, known as “Public Safety Realignment,” placed additional
burdens on probation as prison populations were released to local supervision,
lessening the time probation officers could spend with individuals. One staff member
indicated that probation sessions in the San Francisco area might sometimes last as little
as five minutes.

•

SCA case managers may have approached case management from a different
perspective than traditional POs. SCA case managers received specialized training
around evidence-based practices and approaches for addressing participants’
criminogenic needs. While SCA case managers often focused on needs assessments,
case planning, and service delivery, SCA program staff indicated that traditional POs
were more focused on monitoring, ensuring that those they supervised were meeting
the terms of their release, finding work, staying clean, and not committing new crimes.

34
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

•

SCA case managers had access to a greater array of programs and services. SCA case
managers could use providers committed to serving SCA participants. While many of
these SCA service delivery partners also provided services to a broader pool, SCA
funding or service delivery agreements ensured there were positions available for
program-group participants, while control-group members had to compete with others
for spaces. Furthermore, because of their SCA case manager, program-group members
had support and guidance in learning about and accessing these services.

Overall, because SCA grantees provided both case management and other services, the service
contrast between program and control-group members should have been fairly extensive.
However, there were circumstances that caused the service contrast to be diminished in some
ways. In Marion County and Allegheny County, the grantees, either through SCA funding or
some additional funding, created drop-in centers that provided traditional POs with an easy
one-stop referral source for their supervisees, thus increasing the likelihood that control-group
members might have received at least some of the services also available to program-group
members. Kentucky also used some SCA funding to support classes and workshops to which
traditional POs may have referred control-group members. In San Francisco and San Mateo
County, the nonprofit organizations that primarily operated the SCA programs also provided
numerous other services that were available to the public. Since control-group members would
have interacted with these agencies as part of the study enrollment process, they would have
likely known about and had access to many of these services.

Summary
The FY 09 grant announcement for the Adult Demonstration Program issued by BJA (U.S.
Department of Justice 2009) drew attention to the need for a continuity of services from
incarceration to release and stipulated that funded projects:
1.

Should use validated and dynamic assessment tools to determine risks and needs

Should provide “offenders … with all necessary services, including: (1) educational,
literacy, vocational, and job placement services … (2) substance abuse treatment … and
(3) coordinated supervision and comprehensive services … including housing and mental
and physical health care to facilitate reentry … and which, to the extent applicable, are
provided by community-based organizations entities” (pp. 2-3).

Site visits to the grantees occurred shortly after RA commenced, which was more than two
years into the grantees’ implementation of their grant program. The evaluation found that, at
that time, grantees had relatively well established programs, with features as outlined below.
•

Assessment guided service planning. All the grantees used validated assessment
instruments to identify criminogenic needs.

35
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

•

Coordinated supervision was a central feature. Each grantee’s service delivery model
relied on case managers, who generally planned and coordinated a wide range of
services for participants.

•

SCA grew partnerships. Partnerships were crucial for service delivery, as the grantees
lacked the capacity to deliver the full range of services themselves. These partnerships
increased the linkages between pre-release and post-release services and improved
coordination and service delivery between correctional system agencies and other
governmental and social service system agencies.

•

Services became more comprehensive. On the whole, the grant created an increased
continuity of services, both across partners and especially pre-release and post-release
environments. Staff benefited from training, learning to use various assessments and
case planning tools, and had access to a greater array of services for participants than
they had before the grant.

•

A new approach to re-entry services emerged. Based on what we were told during the
site visits, grantees that were DOCs developed new partnerships, training, and service
delivery approaches, and moved away from a policing and enforcement mindset toward
a rehabilitative philosophy that accepted evidence-based practices. This cultural shift
was, in some cases, substantial.

In these ways, the service contrast between program- and control-group members, especially
post-release services, was readily apparent.
Despite these similarities, the grantees’ programs varied markedly in a number of ways.
•

Case management providers varied. The type of grantee influenced the type of
individuals who served as SCA case managers: correctional system agencies used
correctional system staff, while service-based agencies used case managers with social
service backgrounds.

•

Grantees used their SCA grant funds differently. The grantees generally used a good
portion of their grant funds to provide case management services, but, beyond that
commonality, each used its funds to provide different types of services both before and
after release.

•

Different types of grantees emphasized different types of partnerships. Developing all
the partnerships that were needed took considerable work. Correctional system
grantees had an easier time implementing changes within the correctional system, but
often had to grow additional partnerships with community organizations, especially to
provide post-release services. By contrast, human services and health-based grantees
had to build partnerships with correctional system agencies to allow for the delivery of
pre-release and transitional services, but they typically had a stronger network of
partnerships for the delivery of post-release services.

•

The adequacy of the continuum of care from pre-release to post-release was better
developed in some sites than others. Some grantees devoted substantial SCA resources
to pre-release services, while others began their SCA programs nearer to release. In the
36
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

latter instances, the grantees relied on the existing programming of the institutions to
provide pre-release services and used the period prior to release to begin SCA service
planning for the post-release period.
Notwithstanding these differences, the implementation study found that the grantees
developed programs that, by and large, met the criteria outlined by BJA in the grant solicitation.
However, full implementation of the service model envisioned by BJA also fell short in some
ways. In particular, grantees lacked the funds to ensure that participants accessed all the
services they needed. Therefore, they all relied heavily on unfunded referrals to provide many
services. Where unfunded referrals were used, coordination with the SCA program was typically
weak and case managers could not always track whether participants received the services to
which they were being referred. Given the reliance on unfunded referrals for many post-release
services, it would be hard to argue that every SCA participant received “all necessary services,”
a topic that will be explored further in a subsequent chapter.

37
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

III. Participants’ Experiences in SCA
Based on MIS data that the grantees provided, this chapter describes the SCA services that
program-group members received and how long they participated in SCA. These data are only
available for study participants who were randomly assigned to the SCA program group,
because control group members were not eligible to receive SCA services. Furthermore, this
chapter only describes SCA services that the grantees knew about and entered into their data
systems.24

Findings in Brief
•

Overall, 36 percent of those in the program group received both pre-release and post-release
SCA services, according to the grantees’ MIS data. As would be expected, those enrolled in SCA
well before release were much more likely to receive both pre-release and post-release
services.

•

Nearly one half of the program group received employment assistance and cognitive behavioral
therapy as part of SCA while they were still incarcerated, and more than one-third received
substance abuse treatment.

•

More than one half of the program group received SCA employment assistance after release,
and nearly one half received substance abuse treatment. Another common post-release service
was cognitive behavioral therapy.

•

For some grantees, participation in SCA was expected to last for up to six months, while for
others it was expected to last for a year or more. Overall, approximately 25 percent of
participants participated in SCA for more than one year and another 37 percent participated for
more than six months.

Services Received
Exhibit III-1 shows the percent of program-group members who received only SCA pre-release
services, only post-release services, or both pre-release and post-release services.25 We noted in
the previous chapter that some participants were enrolled in SCA well before release, and

The next chapter overcomes these limitations by using the survey of study participants—both program- and
control-group members—to describe all the re-entry services received, regardless of source.

As was noted elsewhere, individuals classified as not having received an SCA pre-release service could have
received pre-release services from the institutions’ own programs, and not as a consequence of SCA
enrollment.
39
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

others nearer to or after release. To reflect this, the exhibit tabulates service usage by that
distinction.
Overall, 36 percent of participants received both pre-release and post-release services as part
of SCA, and 40 percent received post-release services only. However, as would be expected,
participants who were randomly assigned to the SCA program group more than 30 days before
release were much more likely to receive both pre-release and post-release services (52
percent), though an appreciable number (35 percent) received pre-release services only. By
contrast, most participants (92 percent) randomly assigned nearer to release or after release
received post-release services only.
Exhibit III-1: Percent of Participants Receiving Pre-release Services,
Post-release Services, or Both, Overall and by Timing of Program Entry

Note: Participants are tabulated according to whether their date of random assignment to the SCA
program group was up to or more than 30 days from their date of release from incarceration. In this
exhibit, the date of random assignment serves as a proxy for when program group members began
receiving SCA services.
Source: MIS data provided by the grantees

Exhibit III-2 shows, overall, the percent of SCA participants who received various types of prerelease and post-release services. Panel A shows this for SCA pre-release services and indicates
that, according to the grantees’ MIS data, nearly half of the SCA participants received prerelease employment assistance and cognitive behavioral therapy, and more than one-third
received substance abuse treatment. Panel B shows the incidence of post-release services.
After release, the same three services—employment assistance, substance abuse treatment,

40
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

and cognitive behavioral therapy—were the most common, with employment assistance the
most common of the three.
Exhibit III-2: Incidence of Pre-Release and Post-Releases Services
among SCA Participants
A. Pre-release SCA Services
Cognitive behavioral therapy
Mental health services
Substance abuse treatment
Pro-social services
Employment assistance
Education and training
Housing assistance
0%

10%

20%

30%

40%

50%

60%

40%

50%

60%

B. Post-release SCA Services
Cognitive behavioral therapy
Mental health services
Substance abuse treatment
Pro-social services
Employment assistance
Education and training
Housing assistance
0%

10%

20%

30%

Notes: Numbers in the exhibit represent the percent of the program group who received the service and
reflect only SCA services that the grantees provided directly or otherwise knew about and captured in their
data systems. Percentages are calculated after excluding the 11 percent of the program group who had missing
data on these variables or might not have received any SCA services at all.
Source: MIS data provided by the grantees

41
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

The Duration of Participation
As shown in Exhibit III-3, the grantees had different guidelines regarding how long participants
were expected to receive SCA program services. As revealed by the implementation study,
Allegheny County, Oklahoma, and South Dakota planned to enroll participants with at least four
months remaining before release, while enrollment was expected to occur nearer to release in
the remaining sites. Participation was to continue up to one year after release for participants
in Allegheny County, Oklahoma, and San Mateo County; in the remaining sites, post-release
participation was generally expected to be shorter.
Exhibit III-3:
Expected Duration of SCA Participation

Allegheny County
Kentucky

• Enrollment occurs at least 5 months prior to release
• Participation continues for 12 months after release
• Enrollment occurs within one month of release
• Participation continues for 6 months after release

Marion County

• Enrollment occurs just prior to release
• Participation continues for the duration of SOAR and
aftercare classes, which occur during the 6 months
after release

Oklahoma

• Enrollment occurs 4-6 months prior to release
• Participation continues after release for a total of 18
months

San Francisco

• Enrollment occurs just prior to release
• Participation continues until completion of parole or
participant no longer seeks services

San Mateo County

• Enrollment occurs at least 2 months prior to release
• Participation continues for 1 year after release

South Dakota

• Enrollment occurs up to 9 months prior to release
• Participation continues until completion of parole

Notes: This exhibit describes the expected duration of participation were a participant to enroll in and
complete SCA services according to the grantee’s planned service model. Actual spells of participation could
be longer or shorter given an individual participants’ needs, and some participants can have much shorter
spells if they fail to report for scheduled services. Kentucky and Marion County also used some portion of
their SCA funds to enhance pre-release classes available in institutions, but SCA participants had no preferred
access to these services; therefore, this time is not included in this exhibit.
Source: Data collected from the evaluation’s implementation study (D’Amico et al. 2013).

42
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

In keeping with these guidelines, the average duration of participation—measured as time from
SCA enrollment to the date of the last SCA service—varied across the grantees, as shown in
Exhibit III-4.26 Allegheny County had two-thirds of its participants receive services for more than
one year, and, in Oklahoma and San Mateo County, more than one-third did. Spells were
shortest in Marion County, but the calculation of duration in this site excludes pre-release
services partly funded by SCA that participants might have received. Across all the grantees, the
average duration of participation was approximately 8.5 months; approximately 25 percent of
the sample participated for more than one year, and more than 60 percent participated for at
least six months.
Exhibit III-4: Duration of Participation in SCA, Overall and by Grantee

Notes: The duration of participation is calculated as the time elapsed from the SCA enrollment date to the
date of the last SCA service. Calculations exclude the 29.5 percent of participants missing their enrollment
date and/or their date of last service. Date of last service was much more likely to be missing than the
enrollment date and can be missing because of participant attrition or because the participant was still
receiving services at the time the MIS data were extracted for the research team.
Source: MIS data provided by the grantees

The calculations are restricted to those with both an enrollment date and a date of last service. We collected
MIS data covering at least one year after RA for every participant. However, those with exceptionally long
spells of participation will disproportionately be missing their date of last service; therefore, the figures could
somewhat understate the true length of participation were complete data available for everyone.
43
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

IV. Impacts on Services
The logic underlying the grantees’ program models is that SCA funding will provide those
returning from incarceration with more comprehensive and coordinated re-entry services than
they would have received in the absence of SCA and that these services will improve desistance
and lead to other desirable outcomes. A first step in the evaluation of the grantees, therefore,
is to assess whether those allowed access to SCA received more services than those assigned to
the control group who could have received other re-entry services but not SCA; if there are no
differences in services between the groups, there is no reason to expect differences in
desistance or other outcomes.
To examine service receipt, we analyze data collected from the follow-up survey for both the
SCA program group and the control group. As noted in Chapter I, the survey was administered
18 months after random assignment and asked about a range of services.
Findings in Brief
•

In comparison to those in the control group, those in the program group were more likely to:







Have someone they could turn to for advice and who went out of the way to help
Take classes to change how they think, feel, or act
Get assistance with finding a job
Receive substance-abuse treatment
Get housing assistance

On these items, the program group was between 10 and 25 percentage points more likely to
receive the service than the control group.
•

Eighteen months after RA, those in the program group were just as likely as those in the control
group to report having unmet needs, with more than half of those in both groups reporting the
need for additional housing assistance, job placement assistance, job training, health services,
and educational services.

Impacts on Services Received
As discussed in Chapter II, a central component of SCA in most study sites was to provide a
social services case manager or enhanced PO who could help broker services. In keeping with
this, Exhibit IV-1 shows that program-group members were much more likely to have received
help with re-entry and have an individual case plan. They were much more likely to say that

45
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

they had a case manager or PO who was able to answer questions about available services,
whom they were comfortable talking to, who went out of the way to help, and whom they
could turn to for advice about personal issues. SCA participants were much more likely than
control-group members to report that their case managers or POs were somewhat or very
helpful for avoiding crime. All these differences are statistically significant and sizable in
magnitude.27
Exhibit IV-1: Impacts of SCA on Case Management Services
Program

Control

Difference

Got help with re-entry

77.5

59.0

18.5***

From whom did you get this helpa
Case manager
Probation/parole
Both

25.3
36.7
38.0

24.8
46.8
28.5

Had an individual case plan

56.8

35.2

21.6***

Case manager or PO was able to
answer questions about services

70.3

48.5

21.8***

Was comfortable talking to a case
manager or PO

68.7

44.2

24.5***

Had a person who went out of the way
to help

62.8

42.0

20.8***

Had a person to turn to for advice about
personal or family issues

59.8

42.8

17.0***

Case manager or PO somewhat or very
helpful for avoiding crime

64.7

43.5

21.2***

Sample size

495

294

0.5
-10.1**
9.5**

Notes: Numbers in the first two columns represent the percent receiving the service anytime since the date of
random assignment; the third column represents the difference between the first two columns.
Source: 18-month survey
a

This is a conditional outcome, with the results restricted to those who reported getting help with re-entry.
Therefore, the random assignment design does not ensure equivalence in baseline characteristics between
the program and control groups, and differences in outcomes between the groups are only suggestive of true
estimates of impacts.
*/**/*** Statistically significant at the .1/.05/.01 level.

Estimates reported in this chapter and the ones to follow were weighted to equalize the odds of selection into
the study groups and to adjust for survey response rates. Tests of significance were calculated using
multivariate models that control for pre-RA characteristics, which improves the precision of the estimates. See
Chapter I for more information about statistical methods. Except for the conditional outcomes noted, results in
this chapter are being reported for the 789 study participants who completed the survey, excluding the very
small number with missing data on given items.
46
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Given that the program group reported receiving greater help with re-entry, they should be
more likely to receive the services they needed. Exhibit IV-2 shows that, in fact, the program
group received more of a greater variety of services than those in the control group. For
example, they were more likely to attend classes to change how they thought or acted, which,
based on the literature, can be critical for reducing recidivism (Lowenkamp et al. 2009).

Exhibit IV-2: Impacts of SCA on Other Services
Program

Control

Difference

61.4

41.5

19.8***

31.9

28.1

60.8

39.6

21.2***

55.9

31.3

24.6***

48.8

30.6

18.2***

30.3
12.3

19.9
12.3

10.4***
-0.0

18.7
12.9
8.9

19.1
11.5
9.1

-0.3
1.3
-0.2

66.3

55.1

11.1***

27.5
26.1
21.5
20.8
13.5
7.3
495

22.7
20.2
14.3
6.3
10.4
5.1
294

4.8
5.9*
7.2**
14.5***
3.1
2.2

Cognitive Change & Mental Health Services
Attended classes to change how I think,
act, or feel
Received other services to change how I
feel, such as mental health services
Employment-Related Services
Got help with job-finding skills (how to look
for a job, prepare a resume, interview)
Received advice on answering employers’
questions about criminal history
Case manager or PO somewhat or very
helpful in finding or keeping a job
Got help finding a job
Got vocational training
Education Services
Took ABE or GED classes
Took college courses for credit
Earned a diploma or degree
Other Services
Received inpatient or outpatient substance
abuse treatment
Help getting Food Stamps
Participated in sponsored social activities
Participated in formal mentoring program
Received housing assistance
Help getting public assistance
Help with child-support system a
Sample Size

3.8

Notes: Numbers in the first two columns represent the percent receiving the service since the date of random
assignment; the third column represents the difference between the first two columns. The sample size shown
does not apply to subsetted variables.
Source: 18-month survey
a

This is a conditional outcome, with the results restricted to those who have children. Therefore, the random
assignment design does not ensure equivalence in baseline characteristics between the program and control
groups, and differences in outcomes between the groups are only suggestive of true estimates of impacts.
*/**/*** Statistically significant at the .1/.05/.01 level.

47
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

The program group was also more likely to receive a range of employment-related services. For
example, they were substantially more likely to get advice on how to look for a job, such as help
with preparing a resume and coaching on how to interview for a job. They also were more likely
to get advice on how to answer potential employers’ questions about their criminal history,
which might help address a significant barrier that the formerly incarcerated face when they
apply for jobs (Pager 2003). Similarly, they felt that their case managers or POs were somewhat
or very helpful in finding a job. However, program-group members were not more likely to
receive vocational training or educational services (including adult basic education or general
educational development classes, or college courses). Furthermore, they were not more likely
to earn a diploma or certificate. In fact, the incidence of vocational training and post-secondary
education was low for both groups.
Finally, the survey asked about a range of other services to help address other barriers that
impair the successful re-entry of those returning from incarceration (Petersilia 2003). Those
assigned to the program group were significantly more likely to receive either inpatient or
outpatient substance abuse treatment and receive housing assistance, and were also more
likely to participate in mentoring programs and sponsored social activities. However, there
were no statistically significant differences between the groups in getting help with navigating
the child-support system or getting food stamp benefits or public assistance.

Unmet Needs
The above tables show that substantial proportions of SCA participants received an array of
services designed to help them with re-entry. But what needs remained unmet? To address this
question, the survey asked study participants what additional services they would have liked to
have received. Despite the fact that the incidence of service receipt was quite high in some
cases (see the preceding tables in this chapter), Exhibit IV-3 shows that 18-months after RA
both those in the program group and those in the control group had substantial unmet needs.
For example, more than half of both groups reported wanting additional housing assistance, job
placement assistance, health services, educational services, and job training. More than onethird wanted family reunification services, substance abuse treatment, and mental health
services. Although SCA participants were much more likely than those in the control group to
have received many of these services, they were no less likely to want more help.

48
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit IV-3: Additional Services Would Have Likeda
Program

Control

Difference

Housing support

68.5

69.2

-0.7

Job placement

60.2

61.5

-1.3

Health services

57.6

55.5

2.2

Educational services

54.2

56.1

-1.9

Job training

52.3

54.0

-1.7

Advice on getting a job

52.0

49.9

2.2

Family reunification

38.8

38.0

0.8

Substance abuse treatment

38.7

41.6

-2.9

Mental health services

38.3

0.1

Child-support issues

27.9

29.0

-1.1

Sample size

450

260

Notes: Numbers in the first two columns represent the percent reporting that they desired additional services
in the category; the third column represents the difference between the first two columns.
Source: 18-month survey
a

These questions were asked only of those not continuously incarcerated since random assignment (n=711).
Because release after random assignment may be determined by participation in SCA, differences in
outcomes between the groups are only suggestive of true estimates of impacts.
*/**/*** Statistically significant at the .1/.05/.01 level.

Impacts for Subgroups
Grantees were encouraged by the grant solicitation to use validated assessment tools at intake
to determine the risks and needs of participants and provide them with appropriate services. If
they followed this guidance, subsets of SCA participants may have been more likely to receive
certain services than others, and the gap between them and their counterparts in the control
group might be wider. To investigate these possible differences in impacts, this section
describes exploratory analyses that estimate impacts on service receipt for subgroups.
Exhibit IV-4 presents results for three of the five subgroups introduced in Chapter I. Numbers in
the exhibit represent impact estimates—that is, the difference between the average response
of the program group compared to the control group for each subcategory listed.28 We also
formally tested whether the difference in impacts between the two subgroups within a
category was statistically significant—for example, the impacts of being assigned to the
program group for males versus females. There are virtually no statistically significant

To be concise, only the impact estimates (that is, the difference in means between the program and control
groups) are reported in this exhibit and others in this report presenting subgroup findings.
49
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

differences between the subgroups in the impact of program-group assignment. For example,
being in the program group increased the likelihood that females would receive the range of reentry services to about the same degree as it increased the likelihood for males. Also, being in
the program group had no different impact on the need for additional services for males and
females. There are also no significant differences in impacts across the age subgroups. Thus,
with regard to services, being in the program group improved service access for males and
females, and the younger and older, about equally.
Similarly, there are no consistent differences in the impact of being in the program group for
those in different risk classifications.29 Thus, both lower-risk and higher-risk individuals
benefited from SCA about equally; that is, program-group members in each subgroup were
significantly more likely than those in the control group to receive case management, cognitive
change therapy, and employment-related assistance, and program impacts are about
equivalent regardless of risk level. In short, SCA appears to boost service receipt by an equal
extent for these two risk subgroups.
Exhibit IV-5 shows the results for the two remaining subgroups introduced in Chapter I. Being in
the program group had a more positive effect for participants who were randomly assigned
well before expected release from incarceration on attending classes focused on cognitive
change and on getting help with finding a job. There are also some differences in impacts for
different types of grantees. For example, criminal justice agencies were more likely to provide
help with looking for a job, but social service agencies were more likely to provide help getting
a job. Other than these differences, being in the program group had similar positive impacts for
all these subgroups.

Summary
SCA significantly increased access to a wide range of re-entry services. Those assigned to the
program group were more likely to receive help with re-entry from a case manager or PO and
were more likely to have an individual case plan. They were more likely to report having
someone who was able to answer their questions, whom they were comfortable talking to, and
who went out of the way to help. They were more likely to report having someone they could
turn to for advice about personal matters, and felt that this person helped them in avoiding
crime. In all these ways, the difference between the program and control groups was
approximately 20 percentage points.

As noted in Chapter I, however, only those determined to be at medium or high risk of recidivism were eligible
for SCA, so the subgroups represent degrees of risk within a narrow range. In other words, the lower-risk
subgroup does not represent low-risk individuals in an absolute sense, but only relative to others determined
eligible for SCA.
50
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Changing attitudes and values is believed to be effective in reducing recidivism (Lowenkamp et
al. 2009) and many of the SCA grantees emphasized providing cognitive behavioral therapy.
According to the survey, approximately 61 percent of program-group members received
cognitive behavioral therapy, in comparison to just 42 percent of those in the control group.
The grantees also emphasized providing employment services. Those in the program group
were more likely to have received help on how to look for a job, and they received advice on
answering employers’ questions about their criminal history. They felt that their case managers
or enhanced POs helped them find or keep a job. In these ways as well, differences between
program and control-group members were notable.
Finally, those in the program group were more likely to receive substance abuse treatment,
receive housing assistance, and participate in a mentoring program and sponsored social
activities.
However, despite SCA’s impacts on receipt of a wide range of services, the difference in service
receipt between the program and control groups is generally modest, never exceeding 25
percentage points. In other words, although SCA participants received more services, many
control-group members also accessed these services. Furthermore, SCA participants, just as
control-group members, reported having many unmet service needs 18 months after RA. Twothirds of both groups wanted additional housing assistance, and more than half wanted
additional job placement assistance, job training, health services, and educational services. In
short, despite the fact that SCA did provide meaningful services well beyond what the typical
individual being released from incarceration would have received, many additional services
were apparently needed.

51
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit IV-4:
Impacts of SCA on Re-entry Services, by Gender, Age, and Risk Subgroups
Gender
Females
Males
Case Management Services
Got help with re-entry
Had an individual case plan
Case manager/PO able to answer questions
Was comfortable talking to case manager/PO
Had person who went out of the way to help
Had a person to turn to for advice
Helpful for avoiding crime
Cognitive Change & Mental Health
Attended classes to change how I act or feel
Received other mental health services
Employment-Related Services
Got help with job-finding skills
Advice on answering employers’ questions
Case manager/PO helpful in finding job
Got help finding a job
Got vocational training
Education Services
Took ABE or GED classes (D21a)
Took college courses for credit (D22)
Earned a diploma or degree
Other Services
Received substance abuse treatment
Help getting Food Stamps
Participated in sponsored social activities
Participated in formal mentoring
Received housing assistance
Help getting public assistance
Help with child-support systema

22.0***
17.7**
17.0**
21.9***
12.0
15.7**
19.3**

Age
Less than 30
30 or Older

Level of Riskc
Lower
Higher

17.4***
22.3***
22.8***
25.0***
22.8***
17.1***
21.5***

19.2***
26.2***
22.9***
28.3***
20.8***
19.4***
22.5***

18.0***
17.8***
20.8***
20.9***
20.6***
13.5***
19.6***

21.8***
24.4***
22.3***
24.8***
21.9***
24.2***
21.8***

15.3***
22.3***
21.6***
24.5***
20.0***
10.4**
20.0***

23.0***
1.8

26.5***
1.4

15.4***
6.0

15.2***
3.6

21.1***
5.2

10.4
22.6***
7.3
8.9
-7.9

23.9***
25.0***
20.8***
10.6***
2.0

19.7***
18.7***
20.6***
11.1**
-0.4

23.3***
29.4***
15.6***
10.5**
-0.2

20.1***
24.2***
17.0***
15.5***
3.6

21.8***
22.0***
18.7***
8.3*
-2.3

0.9
5.3
-4.2

-0.8
0.1
0.8

-3.3
3.3
-1.1

3.4
-0.5
1.1

-0.6
-1.4
-2.9

-0.1
3.5
1.3

6.2
9.0
2.8
3.9
18.1***
11.4*
6.0

12.2***
3.5
6.6*
7.9**
13.3***
0.7
1.1

12.2**
7.1
7.7*
9.5**
17.5***
3.0
2.9

9.5*
2.3
3.2
4.6
11.9***
2.9
2.1

18.9***
6.0
6.7
3.1
16.1***
3.9
-1.6

4.4
4.2
3.2
7.5*
13.1***
1.9
6.0*

7.7
9.9

52
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Gender
Females
Males

Age
Less than 30
30 or Older

Level of Riskc
Lower
Higher

Additional Services Would Have Likedb
Housing support
Job placement
Health services
Educational services
Job training
Advice on getting a job
Family reunification
Substance abuse treatment
Mental health services
Child-support issues
Sample Size

-10.4
6.1
-2.6
-1.2
-7.0
1.3
5.4
2.3
8.1
-9.6
171

1.8
-3.2
3.5
-2.1
-0.3
2.4
-0.4
-4.3
-2.3
1.2
617

-2.9
-4.3
-1.0
-4.5
3.8
6.8
10.0* †
-1.3
-2.9
-0.1
365

1.3
1.4
3.9
0.3
-6.3
-1.9
-7.8
-4.7
1.7
-2.6
423

-2.5
-2.5
0.1
-4.5
-6.3
-1.3
0.5
-1.3
1.1
-2.0
368

-0.8
-0.4
1.0
-0.8
2.9
4.8
-0.2
-2.4
-1.4
-1.5
388

Notes: Numbers in the exhibit represent the impact estimates—that is, the difference within each subgroup between the program and control groups in
the percent reporting affirmatively. A positive number means that more of those in the program group responded affirmatively compared to their
counterparts in the control group; a negative number means that more of those in the control group responded affirmatively. Sample size represents the
unweighted number who completed the survey in each group; sample sizes are slightly lower on some items due to missing data, and are lower for
conditional outcomes. Subgroups are described in Chapter I.
Source: 18-month survey
a

This is a conditional outcome, with the results restricted to those who have children. Therefore, the random assignment design does not ensure
equivalence in baseline characteristics between the program and control groups, and differences in outcomes between the groups are only suggestive of
true estimates of impacts.
b

These questions were asked only of those not continuously incarcerated since random assignment. Because release after random assignment may be
determined by participation in SCA, differences in outcomes between the groups are only suggestive of true estimates of impacts.
*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01 level

† The difference in the impact of SCA between the subgroups in the pair is statistically significant at the .05 level (the symbol is placed by the impact
estimate of the first group of the subgroup pair).

53
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit IV-5:
Impacts of SCA on Re-entry Services, by Timing of Entry
and Grantee Type
Timing of Entry
Well Before
Nearer to
Release
Release
Case Management Services
Got help with re-entry
Had an individual case plan
Case manager/PO able to answer questions
Was comfortable talking
Had person who went out of the way to help
Had a person to turn to for advice
Helpful for avoiding crime
Cognitive Change & Mental Health
Attended classes to change how I act or feel
Received other mental health services
Employment-Related Services
Got help with job-finding skills
Advice on answering employers’ questions
Case manager/PO helpful in finding a job
Got help finding a job
Got vocational training
Education Services
Took ABE or GED classes
Took college courses for credit
Earned a diploma or degree
Other Services
Received substance abuse treatment
Help getting Food Stamps
Participated in sponsored social activities
Participated in formal mentoring program
Received housing assistance
Help getting public assistance
Help with child-support system a
Additional Services Would Have Likedb
Housing support
Job placement
Health services
Educational services
Job training
Advice on getting a job
Family reunification
Substance abuse treatment
Mental health services
Child-support issues
Sample Size

Grantee Type
Criminal
Social
Justice
Service

21.8***
26.4***
26.3***
29.1***
25.1***
21.8***
26.0***

10.7*
11.3*
12.0*
15.2**
12.3*
9.3
11.1**

18.0***
18.8***
20.5***
24.0***
17.2***
15.4***
20.2***

19.5***
27.3***
24.4***
25.6***
28.4***
20.2***
23.4***

28.8*** †
3.6

-3.0
4.3

23.4***
3.3

12.4*
4.7
7.6
22.1***
17.1**
24.3***
5.4

21.3***
25.9***
23.4*** †
11.5***
0.1

20.0***
23.1***
7.0
8.3
-0.0

27.7*** †
25.8***
18.7***
3.8 †
-2.6

2.0
2.4
-0.1

-6.2
0.5
-2.1

-1.5
-0.7
-0.6

13.1***
7.2*
6.7*
5.3
17.3***
2.7
0.3

9.8
0.8
5.8
11.0**
8.2**
4.7
5.8*

12.3***
0.9
3.3
6.6**
15.4***
1.5
-0.1

8.7
12.9**
11.2**
8.3
12.5***
6.5
7.0

-3.7
-0.9
2.6
-0.7
2.9
5.7
4.7
0.1
2.8
0.4
546

6.7
-0.1
3.3
-2.7
-8.6
-2.8
-4.8
-8.0
-1.0
-2.7
243

-1.2
-2.2
-2.1
-6.1
-4.2
1.3
-7.6 †
-4.9
-4.7
-2.9
534

0.2
0.4
10.1
5.9
3.0
3.6
16.7***
0.6
9.1
2.4
255

54
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

2.1
5.7
0.5

Notes: Numbers in the exhibit represent the impact estimates—that is, the difference within each subgroup
between the treatment and control groups in the percent reporting affirmatively. A positive number means
that more of those in the treatment group responded affirmatively compared to their counterparts in the
control group; a negative number means that more of those in the control group responded affirmatively.
Sample size represents the unweighted number who completed the survey in each group; sample sizes are
slightly lower on some items due to missing data, and are lower for conditional outcomes. Subgroups are
described in Chapter I.
Source: 18-month survey
a

This is a conditional outcome, with the results restricted to those who have children at the time of the
survey. Therefore, the random assignment design does not ensure equivalence in baseline characteristics
between the program and control groups, and differences in outcomes between the groups are only
suggestive of true estimates of impacts.
b

These questions were asked only of those not continuously incarcerated since random assignment.
Because release after random assignment may be determined by participation in SCA, differences in
outcomes between the groups are only suggestive of true estimates of impacts.
*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01

level

† The difference in the impact of SCA between the subgroups in the pair is statistically significant at the .05
level (the symbol is placed by the impact estimate of the first group of the subgroup pair).

55
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Impacts on Recidivism

Chapter IV showed that being in the program group led to significant increases in the receipt of
re-entry services for program participants. Did these additional services lead to reductions in
recidivism? We measure recidivism as involvement with the criminal justice system in the 18
months after random assignment that led to arrest, conviction, or incarceration. Both
administrative data and survey data are used to measure recidivism for the full sample and the
key subgroups. The impact of the program on re-incarceration measured for the full sample
using administrative data is the confirmatory analysis for this study; other analyses are
considered exploratory.
Findings in Brief
•

Those in the program group were no less likely than those in the control group to be rearrested, reconvicted or re-incarcerated. This conclusion does not change regardless of
whether recidivism is measured using administrative data or survey data.

•

There is some evidence that those in the program group were more likely to have
probation/parole revoked and to be convicted of new crimes, possibly because of the increased
supervision they experienced.

•

Based on administrative data, within 18 months after RA:




•

More than 40 percent of those in the program and control groups were arrested.
Between 25 percent and 31 were convicted.
Just under half were re-incarcerated.

There were few differences in the impact of assignment to the program group across the
various subgroups.

Impacts Overall
As Chapter II discussed, substantial numbers of study participants were incarcerated at the time
of RA. We first explore whether assignment to the program group had an impact on the timing
of release and, consequently, the duration at risk for recidivism following release. We next
examine impacts on recidivism.

Impacts on Time to Release and Time at Risk
For those incarcerated at the time of RA, being in the program group could have had an impact
on when an individual was released from incarceration. Exhibit V-1 examines this possibility by

57
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

showing, for those incarcerated at the time of RA, the time elapsed from RA to the release date
for both the program and control groups. Among those released, the average time to release
was approximately six months for both groups. There are no statistically significant differences
between the groups.
Exhibit V-1: Time Elapsed from RA to Release,
by Program Group

Notes: The plotted lines represent the cumulative percent of participants by the months elapsed from the
RA date to the date of release. The sample is restricted to those study participants who were incarcerated
at the time of RA.
Source: Administrative data from state and local agencies
*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01

The flip side of time to release is the duration at risk for recidivism. Because enrollment in SCA
hypothetically could have had an impact on the release date, the study was designed such that
all outcomes are measured for the 18 months beginning with the date of RA.30 However, for
individuals incarcerated at the time of RA, the time at risk of recidivism will be less than 18
months—and, based on the above exhibit, is indeed much less for some individuals. We
therefore measured the time at risk for the program and control groups. For those who were
randomly assigned after release, their time at risk is the full 18 months; for those who were
randomly assigned while incarcerated, their time at risk is 18 months minus the time from RA to
release.

In a study using random assignment (or, indeed, any impact analysis) an event that occurs after treatment
services begin is viewed as endogenous. Selecting on an endogenous variable can give rise to selectivity bias.
For this reason, outcomes are measured from the date of RA.
58
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit V-2 shows the cumulative percent of participants by time at risk for the program and
control groups. Reading off the graph, the value at the intercept shows that approximately 20
percent of both groups have the entire 18 months at risk, and just under 60 percent have at
least one year at risk. The lines for the two groups are largely in parallel, but slightly more
control-group members have less than one month of time at risk.
Exhibit V-2: Months at Risk of Recidivism,
by Program Group

Notes: The plotted lines represent the cumulative percent of participants, by the number of months
outcomes were observed following release from incarceration, which represents the time at risk of
recidivism.
Source: Administrative data from state and local agencies
*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01

Impacts on Recidivism for the Full Sample
The top panel of Exhibit V-3 reports various measures of recidivism for the program and control
groups when outcomes were measured using administrative data for the period covering 18
months after RA. Based on these results, being in the program group does not appear to reduce
rates of arrest, conviction, or incarceration, and, by some measures, those in the program
group are more likely to have subsequent involvement with the criminal justice system. Results
based on administrative data are summarized below.
•

More than 40 percent of both the program and control groups were arrested sometime
in the follow-up period, with no difference between the groups.

•

There was an average of just over one arrest per person, with the average slightly higher
for the program group than the control group.
59
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

•

Approximately one-third of those in both groups were arrested with a public order
offense. Those in the program group appeared somewhat more likely to have been
arrested for a property crime. Violent crimes were uncommon.

•

Those in the program group were more likely to be convicted of a new crime (31
percent for the program group versus 25 percent for the control group), and they have a
slightly higher average number of convictions.31

•

There is no difference between the groups in the rates of re-incarceration. A little less
than half of both groups had been re-incarcerated in either a prison or jail.

•

Those in both the program and control groups spent approximately 250 days in either
prison or jail during the follow-up period (this includes time from the date of RA to initial
release for those incarcerated at the time of RA); there was no difference between the
groups.

Based on these results, there is no evidence that assignment to the program group in these
seven sites decreased recidivism and, by some measures, it may even have increased it. That
participation in SCA could increase recidivism is unexpected, but may reflect the finding from
prior research that increased supervision can increase the likelihood of catching new offenses
and violations of the terms of parole/probation when they occur (Taxman 2002, Jalbert et al.
2011).
The lower panel of Exhibit V-3 presents impacts using survey data and also provides no
evidence that the program group had lower rates of recidivism. Generally, there are no
significant differences between the groups across any of the measures of recidivism tabulated
from the survey, with the exception that the number of revocations might be somewhat more
likely for those in the program group. Noteworthy is that all the measures of recidivism
calculated from survey data show a lower incidence than counterpart measures tabulated
based on administrative data. This is consistent with the finding that the self-reported incidence
of crime can have significant under-reporting (Thornberry and Krohn 2003, Wiegand et al.
2015).
We also calculated the time elapsed from the date of RA to the first arrest or re-incarceration.
The cumulative frequency distributions of the first occurrence plotted by time elapsed are
shown in Exhibit V-4 for both the program and control groups. The trajectories run almost
completely in parallel and the gap between the groups is very small. There are no significant
differences between the groups in the cumulative percentages with an occurrence at any time
during the 18 months.

Given that most individuals were incarcerated at the time of RA, we assume that convictions that occurred
after RA were generally for new crimes. However, this assumption may not hold in all cases.
60
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit V-3: Impacts on Recidivism for the Full Sample
Program

Control

Difference

44.7
1.3

41.6
1.0

3.1
0.3**

6.6
21.4
18.5
34.0

9.1
13.3
16.7
32.6

-2.4
8.1**
1.8
1.3

31.3
0.4

24.8
0.3

48.4
40.6
22.3
246.5
606

43.8
37.6
20.0
245.2
360

4.6
3.0
2.3
1.3

34.0
0.8

28.7
0.8

5.3
0.0

21.3
15.2
0.2

17.8
12.6
0.1

3.5
2.6
0.1

32.0
19.5

30.5
14.9

1.5
4.7*

39.9
0.9
32.2
494

36.6
0.8
34.7
294

A. Outcomes Measured from Administrative Data
Arrests
Arrested (%)
Average number of arrests
Arrests, by offense type (%)a
Violent crime
Property crime
Drug crime
Public order crime
Convictions
Convicted of a crime (%)
Average number of convictions
Incarcerations (prison or jail)
Was re-incarcerated in prison or jail (%)
Experienced a new jail incarceration (%)
Experienced a new prison incarceration (%)
Total days incarceratedb
Sample Size

6.4*
0.1***

B. Outcomes Measured from the Survey
Arrests
Arrested (%)
Average number of arrests
New charges and convictions
Formally charged with a new crime (%)
Convicted of a new crime (%)
Average number of new convictions
Parole/probation violations
Charged with a violation (%)
Probation/parole revoked (%)
Incarcerations (prison or jail)
Was re-incarcerated in prison or jail (%)
Average number of re-incarcerations
Currently incarcerated (%)
Sample Size

3.2
0.1
-2.5

Notes: Numbers in the first two columns represent outcomes measured in the 18 months following the date of
random assignment for the program and control groups; the third column represents the difference between
the first two columns.
Source: Administrative data from state and local agencies and 18-month survey data
a

The sum across categories exceeds the percent ever arrested because individuals can be arrested more than
once and with different arrest charges in the 18-month follow-up period
b

For those incarcerated at the time of RA, total days includes days incarcerated following RA but before
release.
*/**/*** Statistically significant at the .1/.05/.01 level.

61
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit V-4: Risk Curves for Arrest and Re-incarceration,
by Program Group

Notes: The plotted lines represent the cumulative percent of participants, by the first occurrence of the
event following the RA date.
Source: Administrative data from state and local agencies
*/**/*** Statistically significant at the .1/.05/.01 level.

62
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Impacts on Recidivism by Time at Risk
The recidivism rates just reported are measured for the 18 months following RA for the full
sample. However, we know from the discussion earlier in this chapter that the time at risk for
recidivism among participants who were incarcerated at the time of RA can be much less than
18 months. In this section, we calculate recidivism rates for the subset of the sample whose
outcomes are observed for at least 12 months (but no more than 18 months) following release
from incarceration. These individuals were either not incarcerated at the time of RA, or were
incarcerated but had less than six months elapse from the RA date to the date of release.32
Furthermore, when calculating total days incarcerated, we excluded the days incarcerated
between RA and initial release.
In comparison to Exhibit V-3, Exhibit V-5 shows that, as one would expect given that this
analysis is restricted to a subset with a longer average time at risk, rates of re-arrest,
reconviction, and re-incarceration are somewhat higher than for the full sample. Also, as would
be expected, total days incarcerated are appreciably less because days of incarceration
between RA and release are excluded from the tally. However, substantive conclusions are
unchanged—assignment to the program group did not reduce recidivism and may even have
increased recidivism on some dimensions.

Impacts for Subgroups
Exhibit V-6 presents impacts for three of the key subgroups. The impacts of assignment to the
program group do not differ significantly by gender and they do not differ between the two risk
groups—that is, being in the program group did not significantly reduce recidivism for any of
these groups, regardless of whether recidivism is measured by arrests, convictions, or
incarcerations, and no matter whether administrative data or survey data are used.
However, there are some significant differences in impacts for subgroups defined by age. Both
administrative data and survey data present a consistent picture that assignment to the
program group increased arrests and convictions for younger cohorts, and, based on survey
data, also increased the incidence of parole/probation violations and re-incarcerations.
However, these significant negative impacts did not appear for those who were older.

As we discussed, the date of release could be influenced by participation in SCA and, for this reason,
differences between the program and control groups for this sample subset may not yield unbiased estimates
of impacts. However, given that differences between the program and control groups in the time elapsed from
RA to release are minimal (see Exhibit IV-1), significant bias is unlikely.
63
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit V-5: Impacts on Recidivism,
for Those with at Least One Year at Risk
Program

Control

Difference

60.9
2.0

55.9
1.5

5.0
0.5***

9.9
31.1
24.8
50.3

9.4
19.3
23.3
45.4

0.5
11.8***
1.5
4.9

41.6
0.6

32.9
0.4

8.7**
0.2***

57.2
50.2
25.2
86.0
315

52.6
45.6
24.6
83.9
203

4.6
4.6
0.6
2.1

40.4
1.0

34.4
0.8

6.0
0.2

25.9
18.6
0.3

21.3
14.1
0.2

4.6
4.6
0.1*

35.9
22.6

37.2
17.7

-1.3
4.9

44.9
1.1
24.7
606

42.0
0.8
29.3
360

2.9
0.3
-4.6

A. Outcomes Measured from Administrative Data
Arrests
Arrested (%)
Average number of arrests
Arrests by offense type (%) a
Violent crime
Property crime
Drug crime
Public order crime
Convictions
Convicted of a crime (%)
Average number of convictions
Incarcerations (prison or jail)
Was re-incarcerated in prison or jail (%)
Experienced a new jail incarceration (%)
Experienced a new prison incarceration (%)
Total days incarcerated
Sample Size
B. Outcomes Measured from the Survey
Arrests
Arrested (%)
Average number of arrests
New charges and convictions
Formally charged with a new crime (%)
Convicted of a new crime (%)
Average number of new convictions
Parole/probation violations
Charged with a violation (%)
Probation/parole revoked (%)
Incarcerations (prison or jail)
Was re-incarcerated in prison or jail (%)
Average number of re-incarcerations
Currently incarcerated (%)
Sample Size

Notes: Numbers in the first two columns represent outcomes measured in the period following release from
incarceration for the program and control groups; the third column represents the difference between the first
two columns. The sample is restricted to those whose outcomes are observed for at least 12 months and no
more than 18 months following release.
Source: Administrative data from state and local agencies and 18-month survey data
a

The sum across categories exceeds the percent ever arrested because individuals can be arrested more than
once and with different arrest charges in the 18-month follow-up period
*/**/*** Statistically significant at the .1/.05/.01 level.

64
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit V-7 presents results for the remaining two subgroups. The impact of being in the
program group is not significantly different between those randomly assigned well before
release from incarceration and those randomly assigned near release or after, with the
exception that being in the program group was more likely to increase the incidence of being
charged with a parole or probation violation for those randomly assigned well before release
compared to those randomly assigned later. Grantees that are associated with the state or local
criminal justice system appear to be about as effective as local agencies; the differences in
impacts between the two groups of grantees are not significant.

Summary
As of 18 months after random assignment, being assigned to the program group did not lead to
desistance. Whether recidivism was measured by survey or administrative data, those in the
program group were not less likely than those in the control group to be re-arrested,
reconvicted, or re-incarcerated (in either prison or jail), and the time to their first event—an
arrest or incarceration—was no shorter. This picture does not change appreciably for the
various subgroups considered, except that SCA increased subsequent involvement with the
criminal justice system among those who are younger but not among those who are older.

65
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit V-6:
Impacts of SCA on Recidivism, by Gender, Age, and Risk Subgroups
Gender
Females
Males

Age
Less than 30
30 or Older

Level of Risk
Lower
Higher

10.0
0.3
-5.0
15.4***
3.9
2.2

1.4
0.3**
-1.6
6.2**
1.2
1.2

2.4
-0.0

7.5**
0.2***

13.2*
10.5
10.2*
-3.8
203

2.6
1.1
0.3
5.3
763

8.4
0.2

4.5
-0.0

7.6
4.4
0.1
0.3
7.6

9.6**
0.6*** †

-3.1
0.0

8.9*
0.3

1.3
0.4*

-0.4
10.2***
6.7*
7.9*

-3.9*
5.7
-2.6
-4.8

-2.9
6.3*
4.6
3.5

-0.5
8.9**
3.0
1.9

14.4*** †
0.3*** †

-0.7
-0.0

10.5**
9.2*
7.4*
-18.1
440

7.8*
0.1**

7.5*
0.2**

0.5
-2.0
-1.2
24.8
526

7.5
7.8*
4.6
17.0
452

4.4
1.1
2.3
-9.9
474

15.4*** †
0.4*

-2.8
-0.2

3.5
-0.1

10.1*
0.2

2.4
2.2
0.1

10.2**
9.5** †
0.1** †

-2.2
-3.7
-0.0

3.1
0.0
0.1

7.2
7.9**
0.1***

1.9
3.9

10.1* †
7.4*

-5.3
2.9

3.7
5.1

2.7
6.2

66
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Gender
Females
Males

Age
Less than 30
30 or Older

Level of Risk
Lower
Higher

B. Outcomes Measured from the Survey (continued)
Incarcerated (prison or jail)
Was re-incarcerated in prison or jail (%)
Average number of re-incarcerations
Currently incarcerated (%)
Sample Size

10.1**
0.3**
6.3*
172

1.4
0.1
-4.6
617

15.0*** †
0.6*** †
4.1
365

-5.6
-0.3
-7.4
424

5.9
-0.1
-6.4
368

5.8
0.3*
3.6
388

Notes: Numbers in the exhibit represent the impact estimates—that is, the difference within each subgroup between the incidence or mean value for the
treatment group versus the control group. A positive number denotes that the incidence or mean value is higher for the treatment group than for the
control group; a negative number means that the incidence or mean value is higher for the control group. Subgroups are defined in Chapter I.
Source: Administrative data from state and local criminal justice agencies and the 18-month survey
a

The sum across categories exceeds the percent ever arrested because individuals can be arrested more than once and with different arrest charges in
the 18-month follow-up period
b

For those incarcerated at the time of RA, total days includes days incarcerated following RA but before release.

*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01 level

67
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit V-7:
Impacts of SCA on Recidivism, by Timing of Entry
and Grantee Type
Timing of Entry
Well Before
Nearer to
Release
Release
A. Outcomes Measured from Administrative Data
Arrests
Arrested (%)
7.1*
Average number of arrests
0.3***
Arrests by offense type (%) a
Violent crime
-2.7
Property crime
8.2***
Drug crime
5.1*
Public order crime
4.2
Convictions
Convicted of a crime (%)
10.0***
Average number of convictions
0.1***
Incarcerations (prison or jail)
Was re-incarcerated in prison or jail (%)
8.4**
Experienced a new jail incarceration (%)
7.3*
Experienced new prison incarceration (%)
5.0
Total days incarceratedb
-16.2
Sample Sizes
594
B. Outcomes Measured from the Survey
Arrests
Arrested (%)
Average number of arrests
New charges and convictions
Formally charged with a new crime (%)
Convicted of a new crime (%)
Average number of new convictions
Parole/probation violations
Charged with a violation (%)
Probation/parole revoked (%)
Incarcerated (prison or jail)
Was re-incarcerated in prison or jail (%)
Average number of re-incarcerations
Currently incarcerated (%)
Sample Size

1.3
0.4*

Grantee Type
Criminal
Social
Justice
Service

7.9*
0.4**

-7.0
0.1

-1.7
9.6
-1.4
1.7

-1.3
8.5***
4.7*
4.6

-4.9
7.3
-4.3
-5.5

3.3
0.1

7.1*
0.2***

5.0
0.1

0.8
-1.4
-1.0
-12.2
372

8.6**
6.5
5.5
-2.4
642

-3.6
-4.1
-4.0
8.6
324

8.8**
0.2

-0.6
-0.3

7.2*
0.1

1.3
0.0

5.9
4.9
0.1**

-1.2
-2.0
0.0

3.2
0.9
0.1

4.1
6.2
0.1

7.9* †
5.8*

-9.6
3.4

1.5
3.4

1.5
7.3*

-5.0
-0.1
-6.4
243

5.0
0.1
-2.8
534

-0.3
0.1
-2.0
255

7.7*
0.2
-1.6
546

Notes: Numbers in the exhibit represent the impact estimates—that is, the difference within each subgroup
between the incidence or mean value for the treatment group versus their control group. A positive number
denotes that the incidence or mean value is higher for the treatment group than for the control group; a
negative number means that the incidence or mean value is higher for the control group. Subgroups are
defined in Chapter I.
Source: Administrative data from state and local criminal justice agencies and the 18-month survey

68
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

The sum across categories exceeds the percent ever arrested because individuals can be arrested more
than once and with different arrest charges in the 18-month follow-up period.
b

For those incarcerated at the time of RA, total days includes days incarcerated following RA but before
release.
*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01

level

† The difference in the impact of SCA between the subgroups in the pair is statistically significant at the .05
level (the symbol is placed by the impact estimate of the first group of the subgroup pair).

69
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

VI. Impacts on Employment and Earnings
Although the nature of the relationship between employment and recidivism is not clear
(Uggen 2000, Tripoldi et al. 2009, Apel and Horney 2017), there are strong theoretical and
practical reasons for believing that helping the formerly incarcerated obtain employment can
improve desistance (Uggen and Staff 2001, Duran et al. 2013). Accordingly, all the grantees
provided employment assistance as part of their programs, through direct service or formal or
informal partnerships (see Chapter II). This chapter examines whether these efforts led to
improved employment and earnings.
Two sources of data are used for measuring employment-related outcomes: survey data and
NDNH administrative data. The survey, administered 18 months after RA, asked study
participants whether they were ever employed after RA, whether they were employed at the
time of the survey (labelled “currently employed” in this chapter) and, if currently employed,
their hourly wage and whether they worked full time or had a part-time, temporary or “off-thebooks” job. NDNH was used to measure whether study participants were employed anytime in
the fifth and sixth calendar quarters after the quarter of RA, and, if employed, their earnings.33

Findings in Brief
•

For the full study sample, being assigned to the program group did not improve the probability
of being employed, hourly wages, or earnings.

•

Assignment to the program group did not improve employment-related outcomes consistently
for any subgroup.

Impacts Overall
The top panel of Exhibit VI-1 reports employment outcomes measured using survey data for the
period covering 18 months after RA.

NDNH measures employment and earnings in calendar quarters. The study team used these data to measure
outcomes in the fifth and sixth calendar quarters after the quarter of RA—thus, for two quarters (six months)
that fell sometime within the interval of 13 to 21 months after RA, depending on each participant’s date of RA
within the calendar quarter of RA. Due to limitations on data access, we could not measure employment and
earnings using NDNH for the entire period since RA for each study participant.
71
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit VI-1: Impacts on Employment and Earnings for the Full Sample
Program

Control

Difference

72.8
33.0

72.5
34.9

0.4
-1.9

68.6

64.6

4.0

31.4

35.4

-4.0

12.39
494

11.43
294

0.96

35.7
35.5

33.5
33.0

2.2
2.5

43.3

40.4

2.9

1,609
1,612
3,207
602

1,529
1,446
2,975
355

80
166
232

A. Outcomes Measured from the Survey
Ever employed since RA (%)
Currently employed (%)
Of those currently employeda
Employed full time (%)
Employed part time or in temporary or
seasonal jobs, or off-the-books (%)
Hourly rate of pay ($)
Sample Size
B. Outcomes Measured from NDNH
Employment Status (%)
Employed anytime in the fifth quarter
Employed anytime in the sixth quarter
Employed anytime in either the fifth or sixth
quarters
Earningsb ($)
Earnings in the fifth quarter
Earnings in the sixth quarter
Earnings in both the fifth and sixth quarters
Sample Size

Notes: Numbers in the first two columns represent the percent with the outcome, measured in the 18 months
following the date of random assignment; the third column represents the difference between the first two
columns. Sample sizes for analysis using survey data are reported regardless of whether the sample member is
employed. One individual in the SCA program group was excluded from the calculation of NDNH earnings,
because this individual’s earnings were an extreme outlier.
Source: 18-month survey and the National Directory of New Hires
a

These are conditional outcomes, with the results restricted to those who were employed at the time of the
survey. Therefore, the random assignment design does not ensure equivalence in baseline characteristics
between the program and control groups, and differences in outcomes between the groups are only
suggestive of true estimates of impacts.
b

Those not employed in the quarter are treated as having zero earnings.

*/**/***

Statistically significant at the .1/.05/.01 level.

Results based on survey data can be summarized as follows.
•

Just over 72 percent of those in both the program and control groups reported that they
worked sometime in the 18 months since random assignment.

•

Approximately one-third of those in both groups were working at the time of the survey.

•

Of those who were working at the time of the survey:



Approximately two-thirds of those in both groups were working full time in regular
jobs.

72
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice



The average hourly rate of pay was $12.39 for the program group and $11.43 for
the control group.

None of the differences between the program and control groups is statistically significant.
The lower panel of Exhibit VI-1 presents results using data from NDNH. Consistent with the
findings discussed above, there are no statistically significant differences between the program
group and the control group. Approximately one-third of each group was employed sometime
in each of the fifth and sixth quarters after RA. Including zero earners, the average combined
earnings in both the fifth and sixth quarters was $3,207 for the program group and $2,975 for
those in the control group, but the difference is not statistically significant.

Impacts for Subgroups
Exhibits VI-2 and VI-3 present impact estimates for the various subgroups that were introduced
in Chapter I; the numbers shown in the table are impact estimates (that is, the differences in
outcomes between the program and control groups within each subgroup). Using survey data,
results show that being assigned to the program group has no statistically significant impact on
employment-related outcomes for any of the ten subgroups, and there are no differences in
impacts within any subgroup pair. In other words, based on survey data, being in the program
group did not improve employment outcomes for either males or females, for younger or older
study participants, for those at various risks of recidivism, for those randomly assigned well
before release from incarceration as opposed to nearer to or after release, or for those served
by criminal justice or social service agencies.
Results based on NDNH tell much the same story. However, there is some evidence that males
are more likely to benefit from SCA participation than females in terms of being employed in
the fifth and sixth quarters after RA.

Summary
In the seven grantee sites participating in this study, assignment to the program group did not
improve employment-related outcomes overall or consistently for any subgroup.

73
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit VI-2:
Impacts of SCA on Employment and Earnings, by Gender, Age, and Risk Subgroups
Gender
Females
Males

Age
Less than 30
30 or Older

Level of Risk
Lower
Higher

-6.3
1.0

2.2
-2.5

6.1
-6.8

-5.0
2.3

0.7
0.9

0.9
-5.6

-2.3

6.1

-0.0

6.8

5.2

2.1

2.3

-6.1

0.0

-6.8

-5.2

-2.1

-0.44

1.34

-0.50

1.98

Sample Sizes

171

617

365

423

-8.9†
-9.4†

5.0
5.5

0.2
-0.6

4.5
6.1

-------

-9.7†

6.2

-0.4

6.8

---

-304
115
-189

190
189
363

-269
-288*
-585*

419
621*
1,040

-------

200

757

482

475

2.39*
368

-0.20
388

B. Outcomes Measured from NDNH
Employment Status (%)
Employed anytime in the fifth quarter
Employed anytime in the sixth quarter
Employed anytime in either the fifth or sixth
quarters
Earningsb ($)
Earnings in the fifth quarter
Earnings in the sixth quarter
Earnings in both the fifth and sixth quarters
Sample Sizes

Notes: Numbers in the exhibit represent the impact estimates—that is, the difference within each subgroup between the incidence or mean value for the
program group versus the control group. A positive number denotes that the incidence or mean value is higher for the program group than for the control
group; a negative number means that the incidence or mean value is higher for the control group. Subgroups are defined in Chapter I. Sample sizes for
analysis using survey data are reported regardless of whether the sample member is employed. One individual in the SCA program group was excluded
from the calculation of NDNH earnings, because this individual’s earnings were an outlier. Due to limitations on data access, subgroup estimates of impacts
could not be estimated using NDNH for those at different risks of recidivism.
Source: The 18-month survey and NDNH.

74
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

These are conditional outcomes, with the results restricted to those who were employed at the time of the survey. Therefore, the random assignment
design does not ensure equivalence in baseline characteristics between the program and control groups, and differences in outcomes between the
groups are only suggestive of true estimates of impacts.
b

Those not employed in a quarter are treated as having zero earnings.

*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01 level.

† The difference in the impact of SCA between the subgroups in the pair is statistically significant at the .05 level (the symbol is placed by the impact estimate
of the first group of the subgroup pair).

75
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit VI-3:
Impacts of SCA on Employment and Earnings, by Timing of Entry
and Grantee Type
Timing of Entry
Well Before
Nearer to
Release
Release
A. Outcomes Measured from the Survey
Ever employed since RA (%)
Currently employed (%)
Of those currently employeda
Employed full time (%)
Employed part time or in temporary or
seasonal jobs, or off-the-books (%)
Hourly rate of pay ($)
Sample Sizes
B. Outcomes Measured from NDNH
Employment Status (%)
Employed anytime in the fifth quarter
Employed anytime in the sixth quarter
Employed anytime in either the fifth or
sixth quarters
Earningsb ($)
Earnings in the fifth quarter
Earnings in the sixth quarter
Earnings in both the fifth and sixth quarters
Sample Sizes

Grantee Type
Criminal
Social
Justice
Service

3.1
-2.9

-2.6
2.0

-0.6
-2.9

2.3
0.4

1.4

7.0

1.8

9.2

-1.4

-7.0

-1.8

-9.2

1.36

0.28

-0.23

3.13*

545

243

534

255

0.9
1.5

5.9
4.4

2.8
5.4

1.2
-3.3

0.3

8.3

4.7

-0.5

131
202
313

23
148
172

205
331
536

-171
-169
-384

653

304

639

318

Notes: Numbers in the exhibit represent the impact estimates—that is, the difference within each subgroup
between the incidence or mean value for the program group versus the control group. A positive number
denotes that the incidence or mean value is higher for the program group than for the control group; a
negative number means that the incidence or mean value is higher for the control group. Subgroups are
defined in Chapter I. Sample sizes for analysis using survey data are reported regardless of whether the sample
member is employed. One individual in the SCA program group was excluded from the calculation of NDNH
earnings, because this individual’s earnings were an extreme outlier.
Source: The 18-month survey and NDNH.
a

Those not employed in a quarter are treated as having zero earnings.

*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01

level.

† The difference in the impact of SCA between the subgroups in the pair is statistically significant at the .05 level
(the symbol is placed by the impact estimate of the first group of the subgroup pair).

76
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

VII. Impacts on Other Outcomes
This chapter examines the impacts of assignment to the program group on a variety of
outcomes not yet discussed: housing, health status, substance abuse, ability to meet childsupport obligations, and total income. It reports impacts for each of these outcomes measured
from the participant survey, first for the full sample, and then for key subgroups.
Findings in Brief
•

Being in the program group had no impacts on housing, health status, substance abuse or
ability to meet child support obligations, as measured by indicators available in the survey.

•

Being in the program group may have improved the likelihood that a participant had enough
income for meeting essential needs.

•

There were no noteworthy differences in impacts across subgroups.

Impacts Overall
Exhibit VII-1 reports results for the full sample, and, with scant exception, there were no
impacts of assignment to the program group on any of the outcomes shown.
•

Housing. Approximately one-fourth of both the treatment and control groups were
living in their own house, apartment, or room at the time of the survey, and, among
those doing so, approximately half contributed money for rent or a mortgage.

•

Health status. Questions about health and health access were asked only of those who
were not incarcerated at the time of the survey. Of this group, approximately threefourths reported that their health was good, very good or excellent. Substantial
numbers needed to see a doctor or dentist but could not afford to do so. Approximately
40 percent made a visit to an emergency room. Approximately 20 percent had a health
condition limiting work or other activities in the month prior to the survey;
approximately 15 percent had a limiting emotional problem. There was no impact of
assignment to the program group on any of these measures.

•

Substance abuse. Questions about substance abuse also were asked only of those who
were not incarcerated at the time of the survey. Relatively small numbers admitted to
using illegal drugs or drinking to excess within the month prior to the survey.

•

Child support. Of those with an order to pay child support, approximately half paid
through the child-support enforcement system; others made informal payments to the
custodial parent. Being in the program group did not have an impact on whether
payments were made.

77
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit VII-1: Impacts on Other Outcomes for the Full Sample
Program

Control

Difference

Housing
Currently living in own house, apartment, or
rooma
Contributes to the cost of rent/mortgageb
Health

25.0

23.8

1.1

57.3

55.2

2.1

Health is good, very good, or excellentc
Needed to see a doctor but could not afford toc
Needed to see a dentist but could not afford toc
Made a visit to an emergency room since RAc
Health condition limited work or other activities
in the past monthc
Emotional problems limited work or other
activities in the past monthc
Substance Abuse

78.6
29.9
38.1
40.3

77.9
30.1
40.2
37.1

0.7
-0.1
-2.1
3.3

20.9

19.4

1.5

14.9

15.4

-0.5

Used illegal drugs last monthc
Had 5 or more drinks in a row sometime last
monthc
Child Support

12.7

13.9

-1.2

12.5

14.6

-2.0

Paid required child support last monthd
Paid other support to a custodial parentd
Income

53.7
39.1

54.9
47.7

-1.1
-8.6

Had enough income to support self last monthc
Total income since RA/release:c
Up to $5,000
$5,001 to $10,000
$10,001 to $20,000
More than $20,000
Sample Size

68.6

60.2

43.9
19.8
22.9
13.5
495

42.9
21.3
20.3
15.5
294

8.4**
1.0
-1.5
2.5
-2.0

Own house, apartment or room does not include those living in transitional housing or a treatment facility.

This is a conditional outcome restricted to those living in their own house, apartment, or room. Therefore,
random assignment does not ensure equivalence in baseline characteristics between the groups, and
differences between the groups are only suggestive of true estimates of impacts.
c

This is a conditional outcome, with the results restricted to those who were not incarcerated at the time of
the survey. Therefore, random assignment does not ensure equivalence in baseline characteristics between
the groups, and differences are only suggestive of true estimates of impacts.
d

This is a conditional outcome, with the results restricted to those who were not incarcerated at the time of the
survey and who had an order to pay child support. Therefore, random assignment does not ensure equivalence in
baseline characteristics between the groups, and differences are only suggestive of true estimates of impacts.
*/**/*** Statistically significant at the .1/.05/.01 level.

78
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

•

Income. In the month prior to the survey, approximately 69 percent of those in the
program group reported having enough income, compared to only 60 percent of those
in the control group. This difference is statistically significant. There were no significant
differences in total income in the period since RA (or release, for those incarcerated at
the time of RA).

Impacts for Subgroups
Exhibits VII-2 and VII-3 present impact estimates for the subgroups. There is no evidence that
the impacts discussed in this chapter were consistently different across any of the subgroups.

Summary
Based on indicators measured from the survey, assignment to the program group did not
improve housing or health status, did not decrease the incidence of substance abuse, and did
not increase the likelihood that a participant paid child support. There is modest evidence that
being in the program group increased the adequacy of one’s income for meeting necessary
expenses, although there were no differences in actual reported income between the program
and control groups.

79
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit VII-2:
Impacts of SCA on Other Outcomes, by Gender, Age, and Risk Subgroups
Gender
Females
Males

Age
Less than 30
30 or Older

Level of Risk
Lower
Higher

Housing
Currently living in own house/apartment/room a
Contributes to the cost of rent/mortgageb

2.3
9.5

0.6
-0.0

-1.4
-4.8

2.5
6.3

-1.8
3.4

2.0
-3.6

Health is good, very good, or excellentc
Needed to see a doctor but could not afford toc
Needed to see a dentist but could not afford
toc
Made a visit to an emergency room since RAc
Health condition limited work or other activities
in the past monthc
Emotional problems limited work or other
activities in the past monthc
Substance Abuse

2.9
7.4

0.2
-2.2

1.5
-0.0

1.4
-0.8

2.1
1.1

0.7
0.1

2.7

-3.6

5.1

3.8

0.5

-3.7

-1.8

7.1

2.5

-7.2
4.7

4.5

0.4

-2.8

4.1

-0.3

1.5

2.7

-1.7

-2.6

1.0

-0.3

-1.3

Used illegal drugs last monthc
Had 5 or more drinks in a row sometime last
monthc
Income

3.0

-2.6

0.5

-2.3

-0.9

-0.5

-7.2

-0.4

3.9

-5.8

-3.2

2.1

9.3

8.1

14.3**

3.7

12.6**

3.7

6.9
-2.0
-2.4
-2.5
172

-0.4
-1.5
3.8
-1.9
617

0.0
1.3
1.0
-2.3
365

2.8
-3.2
3.2
-2.9
424

-0.9
-4.4
5.0
0.2
368

2.5
0.7
-0.5
-2.6
388

Health

Had enough income to support self last monthc
Total income since RA/release:c
Up to $5,000
$5,001 to $10,000
$10,001 to $20,000
More than $20,000
Sample Sizes

-11.3** †

80
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Notes: Numbers in the exhibit represent the impact estimates—that is, the difference within each subgroup between the incidence or mean value for the
treatment group versus the control group. A positive number denotes that the incidence or mean value is higher for the treatment group than for the
control group; a negative number means that the incidence or mean value is higher for the control group. Sample size represents the unweighted number
who completed the survey in each group. Subgroups are defined in Chapter I. Results are not shown for payment of child support because subgroup sample
sizes are too small for estimates to be reliable.
Source: 18-month survey
a

Own house, apartment or room does not include those living in transitional housing or a treatment facility.

This is a conditional outcome restricted to those living in their own house, apartment, or room. Therefore, random assignment does not ensure
equivalence in baseline characteristics between the groups, and differences between the groups are only suggestive of true estimates of impacts.
c

This is a conditional outcome, with the results restricted to those who were not incarcerated at the time of the survey. Therefore, random assignment
does not ensure equivalence in baseline characteristics between the groups, and differences are only suggestive of true estimates of impacts.
*/**/***

The difference between the program and control groups is statistically significant at the .1/.05/.01 level.

81
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit VII-3:
Impacts of SCA on Other Outcomes, by Timing of Entry
and Grantee Type
Timing of Entry
Well Before
Nearer to
Release
Release

Grantee Type
Criminal
Social
Justice
Service

Housing
Currently living in own house, apartment, or
rooma
Contributes to the cost of rent/mortgageb
Health

-0.7

5.1

-0.2

4.0

0.4

6.1

-2.1

8.6

Health is good, very good, or excellentc
Needed to see a doctor but could not afford
toc
Needed to see a dentist but could not afford
toc
Made a visit to an emergency room since RAc
Health condition limited work or other
activities in the past monthc
Emotional problems limited work or other
activities in the past monthc
Substance Abuse

2.5
-2.6

-3.4
4.6

2.7
-1.2

-3.1
2.0

-2.1

-0.8

-5.8

4.7

5.7

2.5

2.8

4.1

0.1

5.3

-2.9

9.9*

2.2

-5.2

-0.1

-1.5

Used illegal drugs last monthc

-0.8

-2.6

-1.5

-0.6

-3.1

-1.1

-5.1

2.7

8.7

9.3

3.0

16.6**

1.7
-2.7
1.1
-0.0
546

-6.3
1.1
6.6
-1.3
243

3.3
-1.1
0.4
-2.7
534

-3.4
-2.2
6.4
-0.7
255

Had 5 or more drinks in a row sometime last
monthc
Income
Had enough income to support self last
monthc
Total income since RA/release:c
Up to $5,000
$5,001 to $10.000
$10,001 to $20,000
More than $20,000
Sample Sizes

Notes: Numbers in the exhibit represent the impact estimates—that is, the difference within each subgroup
between the treatment and control groups in the percent reporting affirmatively. A positive number means
that more of those in the treatment group responded affirmatively compared to their counterparts in the
control group; a negative number means that more of those in the control group responded affirmatively.
Sample size represents the unweighted number who completed the survey in each group. Subgroups are
defined in Chapter I. Results are not shown for payment of child support because subgroup sample sizes are
too small for estimates to be reliable.
Source: 18-month survey
a

Own house, apartment or room does not include those living in transitional housing or a treatment facility.

82
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

The difference between the program and control groups is statistically significant at the .1/.05/.01

level.

† The difference in the impact of SCA between the subgroups in the pair is statistically significant at the .05
level (the symbol is placed by the impact estimate of the first group of the subgroup pair).

83
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

VIII. Summary and Conclusions
This study estimated the 18-month impacts of seven programs that were awarded grants
through the first round of funding under the SCA Adult Demonstration Program. Impacts were
estimated using an RA design with individuals screened and eligible for SCA randomly assigned
to either be allowed entry into the SCA program or receive re-entry services normally available
but not enroll in SCA. The differences in outcomes between the two groups were then
compared.

About the Grantees and Their Programs
According to the SCA grant solicitation, the grantees were expected to serve individuals with a
moderate to high risk of recidivism, develop re-entry plans for them based on validated risk and
needs assessments, and provide supervision and comprehensive services that should include,
as needed, educational, literacy, vocational, and job placement services; substance abuse
treatment; housing assistance; and mental and physical health care.
The SCA awards were an important source of funds and led to system improvements. The
implementation study revealed that the SCA awards helped the grantees expand re-entry
services in the grantees’ communities, improve partnerships with other community agencies,
and strengthen the connection between pre-release and post-release services.
Case management was central. Case management was a central feature of all the grantees’
programs except one. Across grantees, the goal of case management was to help prevent
recidivism by providing individualized support and coordinating access to services based on
identified needs and risk factors.
Beyond providing case management, each grantee used its SCA grant to support a variety of
services. All the grantees devoted significant SCA resources to post-release services, and some
also enhanced pre-release services. Specific services supported through SCA included cognitive
behavioral therapy, employment assistance, substance abuse treatment, and housing
assistance, among others. Services that a grantee could not fund through its grant were made
available through unfunded referrals.
Just over one-third of those assigned to the program group received both pre-release and
post-release SCA services following their program enrollment. According to the grantees’ MIS
data, 36 percent of those in the program group received both pre-release and post-release SCA
services, 40 percent received only post-release service, and just over 20 percent received only
pre-release services.

85
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Employment assistance, cognitive behavioral therapy, and substance abuse treatment were
the most common services provided through SCA, both before and after release. Nearly onehalf of the program group received employment assistance and cognitive behavioral therapy as
part of SCA while they were still incarcerated, and more than one-third received substance
abuse treatment. These three services were also the most common ones provided through SCA
after release.
The length of formal participation in SCA varied greatly. Approximately 25 percent of those
assigned to the program group participated in SCA for more than one year and another 37
percent participated for more than six months. A little less than 40 percent participated for less
than six months.

Summary of Impact Findings
Outcomes were assessed for a period covering 18 months after individuals were randomly
assigned. Because case management was the central focus of most grantees’ programs, the
impact study can be thought of as primarily representing the effects of this services. However,
given the differences across grantees in service designs, this study does not test the efficacy of
a specific program model.
Being in the program group led to a substantial increase in the receipt of services. Those in
the program group were significantly more likely to receive a wide range of re-entry services.
They were more likely to receive help with re-entry and were more likely to have an individual
case plan. They were more likely to report having someone who went out of the way to help.
They were also more likely to receive cognitive behavioral therapy, help with looking for a job,
substance abuse treatment, housing assistance, and mentoring.
Increased access to services by those assigned to the program group did not lead to increased
desistance. As of 18 months after random assignment, those in the program group did not have
less involvement with the criminal justice system. Whether recidivism was measured using
survey or administrative data, those in the program group were no less likely than those in the
control group to be re-arrested, reconvicted, or re-incarcerated; their time to re-arrest or reincarceration was no shorter; and they did not have fewer total days incarcerated (including
time in both prisons and jails). Program-group members may have been somewhat more likely
to have probation or parole revoked and to have new convictions, possibly because of the
increased supervision they experienced.
There were no impacts on employment-related outcomes. In the seven grantee sites
participating in this study, assignment to the program group did not improve the probability of
being employed in the follow-up period. In the last six months of follow-up, those in the

86
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

program group earned an average of about $3,200 and those in the control group about
$3,000, but the difference between the groups is not statistically significant.
The program group may have better income adequacy. Study participants were asked about
their income for the last month of the 18-month follow-up period. Those in the program group
were more likely than those in the control group to report that they had enough income to
support themselves during that month.
The program had no effect on a range of other outcomes. The survey measured the adequacy
of housing, health status, self-reported illegal drug use and excessive alcohol consumption, and
the ability to meet child-support obligations. Being in the program group had no effect on any
of these outcomes.
There are, at best, modest differences across subgroups. We compared the estimated impacts
across different subgroups—males versus females, those younger versus those older, those at
lower versus higher risk of recidivism, those enrolled well before release versus those enrolled
near or after release, and those served by corrections agencies versus social services agencies.
There appear to be only modest differences in program impacts across these groups; that is,
assignment to the program group worked about the same for each subgroup in the subgroup
pair.
The study’s major findings are robust to alternative model specifications and data sources.
We estimated impacts as a simple difference in means between the program and control
groups and using more complex statistical methods. For recidivism, the key outcome of
interest, we estimated impacts using various measures of recidivism and using both
administrative data and survey data, which provide independent estimates of desistance. The
findings summarized above hold up to alternative model specifications and data sources.

Why Were There No Impacts on Recidivism?
SCA represented a substantial infusion of funds for these seven grantees, and this study has
demonstrated that this led to a significant increase in service receipt for the program group.
Why did these additional services not improve desistance? A number of general reasons can be
suggested (although not every reason applies to each grantee).
1.

87
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

services were effective, the gap in service receipt between the groups might not be
large enough to translate into differences in recidivism or other outcomes. There are
several reasons why so many control group members were able to access services.
a. Control group members had access to services available in prisons and jails after RA
but while still incarcerated. Most institutions offered courses and workshops
available to their inmates without regard to SCA eligibility. Depending on the
institution, these services included substance abuse treatment, adult literacy
instruction, employment assistance, cognitive change therapy, and others. RA
generally occurred while individuals were incarcerated, so the control group, just as
the program group, had access to these services. Although SCA case managers who
worked with SCA participants prior to release might have made special efforts to
encourage program-group members to take advantage of these services, the
services were generally available to those in the control group without restriction.
b. A substantial proportion of control-group members got help with re-entry. According
to the participant survey, 59 percent of the control group reported that they got
help with re-entry (compared to 78 percent of the program group). Whether this
help was provided by a traditional PO or someone else, this individual could have
provided many of the same services that SCA case managers did: assessing service
needs, offering advice, providing referrals, etc. From qualitative findings, we know
that SCA case managers and enhanced POs were more involved than traditional POs
were in brokering services, but the difference was one of degree
c. Grantees and their partners had other sources of funding, which were, in many
cases, quite substantial. All the grantees were required to leverage funds from
multiple sources, which could include state and local funds and grants from
philanthropic organizations or other sources. Similarly, the grantees’ partners were
existing organizations with their own funding sources and pre-existing outreach
mechanisms. SCA funds, while much appreciated and valued by all the grantees and
the partners the grantees funded, were often not the largest share of the
organizations’ budgets. These other sources of funds were not specifically
earmarked for SCA participants and could have been used to serve control group
members or others in need.
2.

Given available funding, there were limitations to what SCA could do. Those returning
from incarceration face challenges to re-entry that are many and complex (e.g.,
Petersilia 2003). The grantees’ services could not help participants fully overcome these
challenges.
a. Funds were not adequate to directly fund all participants’ needs. Due to resource
constraints, all the grantees relied heavily on informal referrals to provide many
services. For services that were not SCA funded, program-group members did not
have priority access over anyone else who sought services.
b. At the end of 18 months, SCA participants had many unmet needs. Despite SCA’s
significant impacts on services received, those in the program group still reported

88
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

many unmet service needs 18 months after RA. Two-thirds wanted additional
housing assistance, and more than half wanted additional job placement assistance,
job training, health services, and educational services. In fact, at the end of the 18month follow-up period, those in the program group were just as likely to express
the need for additional services as those in the control group.
3.

There were inherent limitations to the projects that grantees developed. Although the
grantees used evidence on what works in developing their programs, there were
limitations to their program models.
a. Case management, even with reduced caseloads, has not been demonstrated to be
effective. All but one of the grantees emphasized case management as part of their
SCA programs. For several grantees, this case management was provided by
traditional POs who were given reduced caseloads; for others, it was provided by
staff members from a social services agency or community-based organization.
However, in their review of correctional rehabilitation approaches, Cullen and
Gendreau (2000) cite evidence that “casework” has not been demonstrated to be
very successful as a re-entry approach. Others have concluded that giving POs
reduced caseloads does not by itself appear to reduce recidivism, and the increased
supervision can increase revocation rates (Petersilia 1999, Jalbert et al. 2011).
b. It was hard to ensure that participants got the services they needed through
unfunded referrals. Many services were provided through unfunded referrals. Using
unfunded referrals to provide services had some clear advantages: this strategy
conserved limited project resources and enabled grantees to draw on a wide
network of community agencies experienced at addressing the many complex needs
of those returning from incarceration. However, one limitation was that there was
often no way for the grantee to ensure that participants would seek out the services
to which they were referred. Moreover, the quality of services provided by loosely
connected partners can be uncertain.
c. Developing strong programs based on the risk-need-responsivity (RNR) framework is
difficult. Programs that address criminogenic needs have been shown to be effective
in reducing recidivism (e.g., Latessa and Lowenkamp 2006). However, Bonta and
Andrews (2007) argue that taking the RNR framework out of a tightly controlled
setting and trying to widely use its principles in the real world tends to make the
model much less effective. Furthermore, in their systematic review of the literature,
Weisburd et al. (2017) note that, while we generally know what works in reducing
recidivism, the specific guidance that practitioners need to convert principles into
practice is often lacking. In short, implementing evidence-based practices and taking
them to scale is not easy.

Conclusions and Caveats
The SCA grant funds helped the grantees enhance their existing programs and capacity and
strengthen their partnerships. Absence of evidence that these funds reduced recidivism to

89
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

some degree highlights a well-known limitation of impact studies: if there are alternative
sources of funds for services, then each source is important in expanding a community’s
capacity but no one source is singularly impactful when compared against all the others
(Heckman et al. 2000).
At the same time, modifications to the service models that the grantees developed might help
improve outcomes. Even before these impact findings were made available, the Department of
Justice learned from the experiences of the grantees in this study and others that received early
funding through the Adult Demonstration Program. Based on what it learned from the
grantees’ implementation experiences, it tightened requirements for grantees that received
subsequent waves of grant funding under the Adult Demonstration program (now called Smart
Reentry). For example:
•

To ensure adherence to evidence-based practices and the provision of meaningful reentry services, grantees are required to complete a planning process before being
approved for implementation funds. During this time, they are to work with a technical
assistance provider to improve their program models.

•

Grantees are required to establish a memorandum of understanding with providers to
ensure that there is a mechanism for follow-up when referrals are made.

•

Grantees must engage with participants prior to release.

•

Grantees must ensure adequate dosage of cognitive-based interventions.

With these modifications to grant requirements, this next generation of Smart Reentry holds
significant promise for yielding more meaningful benefits.

Next Steps
The findings described in this report cover 18 months after RA. This represents a relatively short
observation period. Many SCA participants in this study were enrolled in SCA while they were
incarcerated and were not released from custody for six or more months after RA. Therefore,
the post-release period covered by this study is much shorter than 18 months for many
individuals. As a consequence, there was limited time during the post-release observation
period for program and control-group members to differentiate themselves. It is possible that
these programs will be shown to be effective with a longer post-RA observation period.
The next step for the evaluation is to estimate impacts measured 30 months after RA, which
may shed additional light on the programs’ effectiveness. A report based on these data is
forthcoming.

90
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

References
Andrews, D. A., James Bonta and J. Stephen Wormith. “The Recent Past and Near Future of Risk
and/or Need Assessment.” Crime & Delinquency 52 (January 2006): 7-27.
Apel, Robert and Julie Horney. “How and Why Does Work Matter? Employment Conditions,
Routine Activities, and Crime Among Adult Male Offenders.” Criminology (May 2017): 307343.
Baird, Christopher. “A Question of Evidence: A Critique of Risk Assessment Models Used in the
Justice System.” Madison, WI: National Council on Crime and Delinquency, 2009.
Baird, Chris, Theresa Healy, Kristin Johnson, Andrea Bogie, Erin Wicke Dankert, and Chris
Scharenbroch. 2013. “A Comparison of Risk Assessment Instruments in Juvenile Justice.”
Report for the Office of Juvenile Justice and Delinquency Prevention, grant number 2010JR-FX-0021. Available at https://www.ncjrs.gov/pdffiles1/ojjdp/grants/244477.pdf.
Berman, Judith. Women Offender Transition and Reentry: Gender Responsive Approaches to
Transitioning Women Offenders from Prison to the Community. Washington, DC: U.S.
Department of Justice, National Institute of Corrections, 2005.
http://static.nicic.gov/Library/021815.pdf (accessed March 24, 2016).
Bloom, Barbara, Barbara Owen, and Stephanie Covington. 2003. Gender-Responsive
Strategies: Research, Practice, and Guiding Principles for Women Offenders. Washington,
DC: U.S. Department of Justice, National Institute of Corrections, June 2003, NIC 018017.
Bonta, James and D.A. Andrews. Risk-Need-Responsivity Model for Offender Assessment and
Rehabilitation. Ottawa, Ontario: Public Safety Canada, 2007.
Chaplin, Duncan. “Hierarchical Linear Models: Strengths and Weaknesses.” Presented at the
annual meeting of the Association for Public Policy Analysis and Management, Washington
D.C., November 2003.
Cullen, Francis and Paul Gendreau. “Assessing Correctional Rehabilitation: Policy, Practice, and
Prospects,” in Criminal Justice 2000: Policies, Process, and Decisions in the Criminal Justice
System, ed. J. Horney, Washington, DC: U.S. Department of Justice, National Institute of
Justice, 2000: 109-175.
D’Amico, Ronald, Christian Geckeler, Jennifer Henderson-Frakes, Deborah Kogan, and Tyler
Moazed. 2013. “Evaluation of the Second Chance Act (SCA) Adult Demonstration 2009
Grantees,” Interim report for the National Institute of Justice, grant number 2010-RY-BX0003. NCJRS 243294, available at https://www.ncjrs.gov/pdffiles1/nij/grants/243294.pdf.
D’Amico, Ronald and Christian Geckeler. “Organizational Type as a Determinant of Reentry
Programming: Findings from Two National Reentry Programs.” Paper presented at the
Annual Meeting of the Association for Public Policy Analysis and Management,
Albuquerque, November 8, 2014.

91
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Drake, Elizabeth, Steve Aos and Marna Miller. “Evidence-Based Public Policy Options to Reduce
Crime and Criminal Justice Costs: Implications in Washington State.” Victims and Offenders
(2009): 170-196.
Duran, Le’Ann, Martha Plotkin, Phoebe Potter, and Henry Rosen. Integrated Reentry and
Employment Strategies: Reducing Recidivism and Promoting Job Readiness. Washington,
DC: Council of State Governments Justice Center, 2013.
Durose, Matthew, Alexia Cooper and Howard Snyder. Recidivism of Prisoners Released in 30
States in 2005: Patterns from 2005 to 2010. Washington, DC: U.S. Department of Justice,
Bureau of Justice Statistics, April 2014, NCJ 244205.
European Medicines Agency. Guideline on Adjustment for Baseline Covariates in Clinical Trials.
London: European Medicines Agency Committee for Medicinal Products for Human Use,
February 2015, EMA/CHMP/295050/2013.
Gendreau, Paul, Tracy Little and Claire Goggin. “A Meta-Analysis of the Predictors of Adult
Offender Recidivism: What Works!” Criminology 34 (1996): 575-608.
Glaze, Lauren. Correctional Populations in the United States, 2009. Washington, DC: U.S.
Department of Justice, Bureau of Justice Statistics, December 2010, NCJ 231681.
Harlow, Caroline Wolf. Education and Correctional Populations. Washington, D.C: U.S.
Department of Justice, Bureau of Justice Statistics, January 2003, NCJ 195670.
Heckman, James, Neil Hohmann, and Jeffrey Smith. “Substitution and Dropout Bias in Social
Experiments: A Study of an Influential Social Experiment.” Quarterly Journal of Economics
115 (May 2000).: 651-694.
Hammett, Theodore M., Cheryl Roberts and Sofia Kennedy. “Health‐Related Issues in Prisoner
Reentry.” Crime and Delinquency 47 (July 2001): 390‐409.
Hernandez, Adrian, Ewout Steyerberg, and J. Dik Habbema. “Covariate Adjustment in
Randomized Controlled Trials with Dichotomous Outcomes Increases Statistical Power and
Reduces Sample Size requirements.” Journal of Clinical Epidemiology (57: 2014): 454-460.
Hirschi, Travis and Michael Gottfredson. “Age and the Explanation of Crime.” American Journal
of Sociology 89 (November 1983): 552-584.
Holzer, Harry, Steven Raphael and Michael Stoll. The Effect of an Applicant’s Criminal History on
Employer Hiring Practices and Screening Decisions: Evidence from Los Angeles. Ann Arbor,
MI: National Poverty Center, University of Michigan, 2004.
Jalbert, Sarah Kuck, William Rhodes, Michael Kane, Elyse Clawson, Bradford Bogue, Chris
Flygare, Ryan Kling, Meaghan Guevara. A Multi-Site Evaluation of Reduced Probation
Caseload Size in an Evidence-Based Practice Setting. Washington, DC: U.S. Department of
Justice, National Institute of Justice, June 2011, NCJRS 234596.
James, Nathan. Risk and Needs Assessment in the Criminal Justice System. Washington, DC:
Congressional Research Service, October 2015, R44087.

92
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

James, Doris and Lauren Glaze. Mental Health Problems of Prison and Jail Inmates. Washington,
DC: U.S. Department of Justice, Bureau of Justice Statistics, September 2006, NCJ 213600.
Kaeble, Danielle, Lauren Glaze, Anastasios Tsoutis, and Todd Minton. Correctional Populations
in the United States, 2014. Washington, DC: U.S. Department of Justice, Bureau of Justice
Statistics, December 2016, NCJ 249513.
Kaeble, Danielle and Thomas Bonczar. Probation and Parole in the United States, 2015.
Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics, December 2016,
NCJ 250230.
Kahan, Brennan, Vipul Jairath, Caroline Dore, and Tim Morris. “The Risks and Rewards of
Covariate Adjustment in Randomized Trials: An Assessment of 12 Outcomes from 8
Studies.” Trials (2016): 1-7.
Kemple, James and Jason Snipes. “A Regression-Based Strategy for Defining Subgroups in a
Social Experiment.” New York: MDRC, 2001.
Kroner, Daryl and Jeremy Mills. “The Accuracy of Five Risk Appraisal Instruments in Predicting
Institutional Misconduct and New Convictions.” Criminal Justice and Behavior (August:
2001): 471-489.
Latessa, Edward and Christopher Lowenkamp. “What Works in Reducing Recidivism?”
University of St. Thomas Law Journal (2006): 521-535.
La Vigne, Nancy and Vera Kachnowski. “A Portrait of Prisoner Reentry in Maryland.”
Washington, DC: Urban Institute, 2003.
Lee, Valerie. “Using Hierarchical Linear Modeling to Study Social Contexts: The Case of School
Effects.” Educational Psychologist 35 (2000), 125-141.
Lipsey, Mark and Francis Cullen. “The Effectiveness of Correctional Rehabilitation: A Review of
Systematic Reviews.” Annual Review of Law and Social Science 3 (December 2007): 297320.
Lowenkamp, Christopher, Dana Hubbard, Matthew Makarios, and Edward Latessa. “A Quasiexperimental Evaluation of Thinking for a Change: A ‘Real-World’ Application.” Criminal
Justice and Behavior 36 (February 2009): 137-146.
McNiel, Dale and Renee Binder. “Effectiveness of a Mental Health Court in Reducing Criminal
Recidivism and Violence.” American Journal of Psychiatry 165 (2007): 1395-1403.
Mumola, Christopher and Jennifer Karberg. Drug Use and Dependence, State and Federal
Prisoners, 2004. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics,
October 2006, NCJ 213530.
Pager, Devah. “The Mark of a Criminal Record.” American Journal of Sociology 108 (March
2003): 937–75
Petersilia, Joan. “Parole and Prisoner Reentry in the United States.” Crime and Justice (1999):
479-529.

93
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Petersilia, Joan. When Prisoners Come Home: Parole and Prisoner Reenty. New York: Oxford
University Press, 2003.
Petersilia, Joan. “What Works in Prisoner Reentry? Reviewing and Questioning the Evidence.”
Federal Probation: A Journal of Correctional Philosophy and Practice (September 2004): 4-8.
Pew Center on the States. State of Recidivism: The Revolving Door of America’s Prisons.
Washington, DC: Pew Charitable Trusts, 2011.
Pew Charitable Trusts. Max Out: The Rise in Prison Inmates Released Without Supervision.
Philadelphia, PA: Pew Charitable Trusts, 2014.
Raphael, Steven. The New Scarlet Letter?: Negotiating the U.S. Labor Market with a Criminal
Record. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, 2014.
Schochet, Peter. Technical Methods Report: Guidelines for Multiple Testing in Impact
Evaluations. Washington, DC: U.S. Department of Education, May 2008, NCEE 2008-4018.
Seiter, Richard and Karen Kadela. “Prisoner Reentry: What Works, What Does Not, and What Is
Promising.” Crime and Delinquency (July 2003): 360-388.
Spruance, Spotswood, Julia Reid, Michael Grace, and Matthew Samore. 2004. “Hazard Ratio in
Clinical Trials.” Antimicrobial Agents and Chemotherapy 48 (August 2004): 2787-2792.
Taxman, Faye. “Supervision—Exploring the Dimensions of Effectiveness.” Federal Probation 66
(September 2002): 14-27.
Thornberry, Terence and Marvin Krohn. “Comparisons of Self-Report and Official Data for
Measuring Crime,” in Measurement Problems in Criminal Justice Research: Workshop
Summary, eds. J. Pepper and C. Petrie, Washington, DC: The National Academies Press,
2003: 43-94
Travis, Jeremy, Amy L. Solomon, and Michelle Waul. “From Prison to Home: The Dimensions
and Consequences of Prisoner Reentry,” Washington, DC: Urban Institute, 2001
Tripoldi, Stephen, Johnny Kim, and Kimberly Bender. “Is Employment Associated With Reduced
Recidivism?: The Complex Relationship Between Employment and Crime.” International
Journal of Offender Therapy and Comparative Criminology 54 (August 2009): 706-720.
Uggen, Christopher. “Work as a Turning Point in the Life Course of Criminals: A Duration Model
of Age, Employment, and Recidivism.” American Sociological Review 67 (August 2000): 529546.
Uggen, Christopher and Jeremy Staff. “Work as a Turning Point for Criminal Offenders.”
Corrections Management Quarterly 5 (2001): 1–16.
U.S. Department of Justice. Second Chance Act Prisoner Reentry Initiative FY 2009 Competitive
Grant Announcement. Washington, DC: U.S. Department of Justice, Office of Justice
Programs, Bureau of Justice Assistance, 2009.
Weisburd, David, David Farrington, and Charlotte Gill. “What Work in Crime Prevention and
Rehabilitation.” Criminology & Public Policy 16 (May 2017): 415–449.

94
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Wiegand, Andrew, Jesse Sussell, Erin Valentine, and Brittany Henderson. Evaluation of the ReIntegration of Ex-Offenders (RExO) Program: Two-Year Impact Report. Oakland, CA: Social
Policy Research Associates, 2015.
Wilson, James and Christine Zozula. “Reconsidering the Project Greenlight Intervention: Why
Thinking About Risk Matters.” National Institute of Justice Journal 268 (October 2011): 1015.
Zweig, Janine, Jennifer Yahner, and Cindy Redcross. Recidivism Effects of the Center for
Employment Opportunities (CEO) Program Vary by Former Prisoners’ Risk of Reoffending.
New York: MDRC, 2010.

95
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Appendix A. Implementing Random Assignment
This appendix describes the way RA was implemented. It first discusses changes to the their
pre-existing eligibility rules that the grantees made to accommodate random assignment. Next,
the appendix discusses the mechanics of random assignment.

Changes to Eligibility to Accommodate Random Assignment
As discussed in Chapter I, each grantee had its own criteria for determining eligibility for SCA
and its own service model. The evaluation endeavored to accommodate these existing
procedures so that it would be evaluating the programs as the grantees meant them to
operate. However, the grantees did make some changes as the evaluation was introduced,
mostly to increase the pool of eligible individuals recruited for the study. These changes were
modest and included the following changes in three sites.
•

Allegheny County. This grantee’s original plan was to recruit individuals into its SCA
program who had at least six months remaining on their sentences. The grantee
changed this to five months remaining when the evaluation was introduced and, to
increase its pool of eligible individuals, conducted outreach to those incarcerated in
alternative housing sites (as well as jails).

•

Kentucky. As the study was getting underway, the state tightened its criteria for granting
discretionary release. This change was not influenced by the study, but, as a
consequence of it, the grantee was falling short of its enrollment targets for the study.
Consequently, the grantee began recruiting from jails as well as prisons.

•

South Dakota. At the outset, persons who met South Dakota’s eligibility criteria for SCA
were required to participate in SCA—that is, they were required to meet with a re-entry
staff member who coordinated pre-release services, and they were assigned to an
“enhanced PO” upon release. According to conditions established by the study’s IRB,
participation in the study must be voluntary. Therefore, as a consequence of
participating in the study, South Dakota made participation in SCA voluntary.

There were no notable changes to eligibility or outreach caused by the study in Marion County,
Oklahoma, San Francisco, or San Mateo County.

The Random Assignment Process
Each grantee randomly assigned persons determined eligible for SCA. The process laid out by
the study team required that, before RA, grantees were to provide a study orientation to
applicants and obtain informed consent, and only then could they conduct RA.

97
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Providing an Orientation and Obtaining Consent
To adhere to the IRB’s requirements for conducting random assignment, every potential study
participant needed to understand the research study and give consent to participate. The
research team helped the grantees to provide this study orientation by developing materials for
them to use. These materials, which were reviewed and approved by the IRB, included:
•

A video. Grantees were provided with a short video, available on DVD, which they could
play at study orientation sessions. The video described the purposes of the study, the
random assignment process, and what data would be collected as part of the study on
each person who was randomly assigned.

•

Scripts. Scripts for explaining the study and a frequently asked questions (FAQ)
document were provided.

•

Notification materials. Some grantees notified individuals about the results of random
assignment by written correspondence. We provided the grantees with draft letters for
them to use if they desired.

•

Informed consent forms. After receiving an orientation to the study, every person being
considered for random assignment needed to give written consent to participate in the
study before random assignment could occur. The consent form was developed by the
study team and approved by the study’s IRB. It covered, among other things, the
purposes of the study, what information would be collected on study participants, how
participants’ data would be kept secure, and the benefits and risks of participation.
Importantly, the form made clear that participation was voluntary, that the decision to
participate would not affect conditions of incarceration or the likelihood of receiving
parole or probation, and that individuals could drop out of the study at any time without
penalty. Those who declined to sign the consent form were not enrolled in the study
and could not participate in SCA. (Grantees told us that no more than a few people
declined to give consent and no one dropped out after being randomly assigned.)

In addition to providing the materials described above, the study team also provided each
grantee with a customized procedures manual and delivered in-person training on how to use
the above materials and carry out the study’s procedures.

The Mechanics of Random Assignment
Once the study orientation was provided and written consent was given, each SCA applicant
completed the BIF. Next, random assignment occurred.
To ensure rigor in conducting random assignment, the study team developed an online random
assignment system which the grantees were required to use. Each grantee staff member
conducting random assignment was given a personal username and password and used these
credentials to log into a secure virtual private network to access the online random assignment
system. Once logged in, the staff person would enter a few pieces of information about the

98
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

person to be randomly assigned, such as name and date of birth. Once these fields were
entered, the applicant would be randomly assigned instantaneously, and the staff member
would be instantaneously notified of the applicant’s group assignment.
During the period of random assignment, the grantee sent the signed consent forms and BIFs to
the study team in approximately monthly batches using a traceable delivery service. The study
team checked the forms to be sure that a signed consent form and BIF were provided for every
person randomly assigned. Those who were randomly assigned but lacking a consent form were
removed from the study (seven individuals were removed for the study because their consent
forms were missing).
An individual from the study team was designated as the primary site liaison for each grantee
during the period that random assignment occurred and was available to provide help. The site
liaison scheduled regular telephone calls with the grant manager at each site, weekly when
random assignment first began and less frequently after a time. The purposes of the calls were
to provide support, answer questions, and troubleshoot problems that arose. Additionally, the
study team monitored sample build-up through weekly reports generated from the random
assignment system and checked periodically that the program and control groups were
balanced on the BIF’s baseline characteristics (as would be expected if random assignment
were being carried out properly). Finally, during the site visits conducted as part of the
implementation study, the liaison assigned to the site provided additional support on RA to
grantee staff and observed at least one study orientation session to be sure that procedures
were being followed correctly.

99
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Appendix B. Survey Methods
A survey was administered to study participants (both program- and control-group members)
approximately 18 months after their dates of RA. The survey asked about background
characteristics, re-entry services received, involvement with the criminal justice system, and
employment, among other topics. Data collection began in June 2013 and ended in early 2015,
with an 82.3 percent response rate.

Locating Respondents
Participant cases were assigned to trained field interviewers 18 months after each participant’s
RA date. These monthly assignments for the follow-up interview began in June 2013 (18
months after the first study participant was randomly assigned), with the last assignment of
participants in September 2014. Assignments ranged from 19 to 88 participants per month.
Initial efforts to locate participants made use of the contact information forms, which
participants filled out at the time of RA. The contact information form asked the participants to
provide information about how they could be reached for the follow-up survey and elicited
contact information for up to three of each participant’s significant others.
Attempting to interview the formerly incarcerated can be challenging under any circumstances,
even with this contact information. Particularly for the SCA study, locating participants at
follow-up was difficult because most were incarcerated at the time of random assignment. This
situation often caused contact information to be incomplete because participants were unsure
of their future housing situation following release. Additionally, contact information for
participants could easily change during the 18-month follow-up period, because housing
situations immediately following release were often temporary.
The survey data collection procedures were designed to overcome these challenges. To begin, a
letter was sent to the participant’s last known address. This letter:
•

Described the survey and reminded participants that they had previously agreed to
participate in the study,

•

Noted that each participant would be provided with a $50 incentive payment for
completing the survey, and

•

Invited the participant to call a toll-free number to complete the interview and offered
an additional $25 if the participant called within two weeks of receiving the letter.

If after two weeks the participant did not call the interviewer, then interviewers began making
outgoing calls to participants. Letters were also sent to all secondary contacts (i.e., friends

101
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

and/or relatives provided on the contact information forms) with a valid address. This ‘locating’
letter contained the participant’s name and asked the recipient to call the interviewer to
provide updated contact information for the participant so that the participant could complete
the interview and receive the incentive payment. If after these steps the participant had not
been reached, the case was referred for more intensive locating using online databases.
Outgoing calls continued using updated contact information when it was obtained.
Sometimes, queries provided evidence that the participant had been re-incarcerated. In these
cases, the interviewer or other evaluation staff checked recent law enforcement or other
criminal records databases to verify the participant’s status. If it was confirmed that a
participant had been recommitted, a formal query was submitted to the institution requesting
permission to conduct the interview. If the facility gave permission, an attempt was made to
interview the participant by telephone. In total, interviews were completed in facilities in six
departments of corrections, 13 local jails, and the Federal Bureau of Prisons.
When all other options to find study participants were exhausted, the final step in locating
participants was a site visit to the participant’s assumed geographical area (based on
information on the contact information form). During the visits, field staff attempted to obtain
updated contact information through neighbors, significant others, friends, and employers. The
field staff also offered to conduct the interview in-person if the participant was located.

Conducting the Interviews
Follow-up interviews were 40 minutes in length on average. Most interviews were completed
over the telephone using computer-assisted-telephone-interviews (CATI), with computerassisted-personal-interviews (CAPI) administered for participants that required in-person
contact.
Interviews were completed with 787 of the 966 randomly assigned participants, resulting in an
82.3 percent response rate at follow-up [completed interviews/ (total sample-deceased
participants)]. Of these, 258 were completed while the participant was incarcerated; among
those not incarcerated, 66 interviews were completed in-person and 463 were completed by
telephone. Interviews were not completed with 169 individuals for multiple reasons, including
participant refusals, inability to locate the participant, lack of access to incarcerated
participants, and participant deportation. The response rate was 82.2 percent for the SCA
program group and 82.6 percent for the control group. The overall response rates across the
seven grantees ranged from 75.4 percent to 87.3 percent (see Exhibit B-1).

102
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit B-1: Response Rate by Site
N of cases

Deceased

Completes

Percent
Complete

Allegheny County (PA)

133

109

82.6

Kentucky DOC

187

146

79.8

Marion County (OR)

119

103

87.3

Oklahoma DOC

134

110

84.0

76.3

San Mateo County (CA)

114

75.4

South Dakota DOC

202

175

86.6

Total

966

787

82.3

San Francisco (CA)

Source: Survey database.

Weighting
The follow-up interview data were weighted to adjust for nonresponse. A simple first step in
the weighting plan was to create a base weight, which, given the nature of this survey, was set
to a value of 1.0 for each respondent. Next, we calculated the nonresponse adjustment using a
propensity score approach. In this method, a logistic model was run to identify variables
associated with the likelihood of responding. Multiple variables were considered for this
adjustment, including treatment or control-group status, site, gender and race/ethnicity. The
propensity model and estimated scores were calculated separately for the treatment and
control groups. The weight is proportional to the inverse of the probability of responding.

103
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Appendix C.

Methods for Data Analysis

This technical appendix describes the statistical methods used to estimate the impacts of SCA in
the seven grantee sites. It first describes the methods generally, including the simple difference
in mean outcomes presented throughout the main body of the report as the estimate of
impacts and alternative methods used to demonstrate the robustness of the report’s major
conclusions to different model specifications and methods. After describing the methods, the
appendix concludes by presenting selected results from those additional analyses.

Statistical Methods Used
The evaluation implemented a randomized controlled trial (RCT), whereby those screened and
eligible for SCA within each of the seven grantee sites were randomly assigned to either the
program group or control group. Random assignment, by design, enabled unbiased estimates of
the impact of being assigned to the program group by generating program and control groups
that should not systematically differ in any way except in their exposure to the program and
things affected by it. Random assignment eliminates any selection biases that might occur in
studies using observational data (where the program and comparison groups may
systematically differ in both observed and unobserved ways), which can bias impact estimates.
To verify that the program and control groups were comparable, means for the two groups
were contrasted on observable background characteristics measured at baseline (see
Chapter I). These characteristics included the participant’s age, racial and ethnic background,
disability status, employment history, criminal record, and educational attainment. Generally,
the program group was not statistically different from the control group on these background
characteristics—with similar equivalence expected for unobserved characteristics as well.
Using an intent-to-treat (ITT) approach, impacts were assessed by comparing the outcomes for
those assigned to the program group to outcomes of the control group. In keeping with ITT,
control groups members could have accessed re-entry services from other sources, but could
not enroll in SCA; conversely, not all those randomly assigned to the SCA program group
necessarily received all the SCA services that they needed. Thus, impacts are properly
interpreted as the effect of being allowed to access SCA relative to receiving whatever re-entry
services were normally available from other sources. The experiences of the control group
provide measures of what would have happened to the program group had enrollment in SCA
not been available.

105
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Difference in Means
Given the RCT design and the resulting baseline equivalence, the difference in means on
outcomes produces an unbiased estimate of the treatment effect. These mean differences
were predominantly used as the impact estimates throughout the report.

Regression Analysis
Regression analysis was used both to assess levels of statistical significance and as a sensitivity
test in estimating impacts. Additionally, because whether an individual was incarcerated at the
time of RA is included as a covariate, the regression models serve to partially control for the
period at risk of recidivism following RA.
The regression analysis adds covariates to a model estimating the treatment effect. Including
covariates is beneficial to the extent the covariates are correlated with the outcome. If they
are, regression adjustment increases the overall variation explained and reduces unexplained
error, which can improve the precision of the estimate of the treatment effect (for continuous
variables) and increases the power of statistical tests (Kahan et al. 2014; Hernandez et al. 2014).
Two types of regression models were used for this study: ordinary least squares (OLS) for
outcomes that are continuous, and logistic regressions for outcomes that are dichotomous.
While OLS regressions are appropriate for outcomes that are continuous variables, logistic
regressions are needed for assessing binary outcomes, because the OLS analysis of them
violates OLS’s assumptions regarding the distribution of errors.
The regression models included a vector of individual and grantee-level characteristics, as
represented in Equation 1:
Yn= β0 + β1Group Assignmentn +  βpXpn+ εn

(1)

In this equation, Group Assignment is coded 1 for those assigned to the program group and 0
otherwise; β1 provides the estimated treatment effect of SCA on outcome Y; Xp represents each
of the covariates p, with βp providing the corresponding coefficients for these covariates; the
error term (ε) represents the difference between the observed and predicted outcome for
person n. Because regression adjustment improves statistical power, the simple differences in
means reported in the main body of the report were assessed for statistical significance after
using regression adjustments.
Following guidance in the literature for deciding which covariates to include (e.g., European
Medicines Agency 2015), we focused on factors felt to be moderate or strong predictors of
recidivism, the main outcome of interest in this study. Based on literature identifying static
predictors of recidivism (see, for example, Gendreau et al. 1996), the variables we included
were gender, age, and indicators of criminal history, among others. Exhibit C-1 details the

106
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

individual and grantee-level characteristics included as covariates in the regression analysis and
presents their summary statistics. Note that not all baseline characteristics reported in
Chapter I were included in these regression models. Some of these characteristics were not
known to be strong predictors of recidivism, were collinear with variables already included, or
had modest amounts of missing data. The inclusion of these variables would not increase the
explanation of variance and, in some cases, could introduce bias in the estimation of the
treatment effect (to the extent that sample cases needed to be dropped due to missing data).
Therefore, these variables were not included in the regression models.
Exhibit C-1: Descriptive Statistics of Background Characteristics
Included in Regression Models
N

Mean

Standard
Deviation

Male (1=yes, 0=no)

966

78.9

40.8

Age (in years)

966

33.3

10.4

Hispanic (1=yes, 0=no)

965

9.7

29.7

African-American (1=yes, 0=no)

965

31.6

46.5

Other non-white non-Hispanic (1=yes, 0=no)

965

12.6

33.2

Has at least a H.S. diploma or GED equivalent (1=yes, 0=no)

956

76.1

42.6

Incarcerated at time of random assignment (1=yes, 0=no)

966

80.5

39.6

Total years incarcerated in prior 10 years

966

2.8

2.5

Number of arrests in prior 10 years

939

10.7

10.1

Variable

Source: Baseline information forms and administrative data
Notes: In addition to the variables shown, a treatment dummy variable was also included, representing
whether or not the individual was randomly assigned to the SCA program group. Dichotomous variables in
the table above were multiplied by 100, for ease of presentation. Estimates are unweighted.

Hierarchical Linear Modeling
HLM is useful in analyzing data when sample members are drawn within discrete units. HLM
takes into account this hierarchical structure in its estimation, correcting for the correlation of
errors within the clusters and eliminating potential bias (typically downward) in the estimation
of standard errors (Chaplin 2003). In this evaluation, HLM is used to account for the nested
structure of participants within grantee sites.
The intraclass correlation coefficient (ICC) examines how much of the total variance in the
outcome measure can be attributed to group identification and is calculated by dividing the
group-level variance over the total variance (see Equation 2). A multilevel model is generally
only required when the ICC is non-trivial (Lee 2000).

107
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

𝐼𝐶𝐶 =

𝑣𝑎𝑟(𝑈0𝑗 )
𝑣𝑎𝑟( 𝜀𝑖𝑗)+𝑣𝑎𝑟(𝑈0𝑗 )

(2)

The multilevel model used in this study is represented through the following multilevel
equation:
Yn= β0j + β1jGroup Assignmentn +  βpjXpn+ εn

(3)

βpj=γp Zpj + Uj

Equation 3 is identical to Equation 1 but with the addition of a level-2 equation, which allows
estimation to vary by site j. The level-2 equation estimates site-level intercepts and slopes (β)
using site-level covariates (Zpj) and corresponding coefficients (γp).
For the purposes of a sensitivity analysis, the impacts of program group assignment using HLM
are presented in this appendix, but site-specific effects were not estimated.

Survival Analysis
Differences between group means on key recidivism outcomes included in Chapter V served as
indicators of the program’s impact on recidivism. However, indicator variables, while providing
simple and easy-to-understand metrics, potentially lose nuances in the information on times to
an event. For example, one individual might have been re-incarcerated 1 month after random
assignment, and a second individual might have been re-incarcerated 17.5 months after
random assignment. At the time of the 18-month follow-up period, both individuals are
identified as having been re-incarcerated, even though there is a qualitative difference between
these individuals in their time to re-incarceration. To supplement the key recidivism outcome
measures reported in Chapter IV, survival analysis was conducted to examine the impact of the
grantees’ programs on the time until recidivism.
One approach to conducting survival analysis is using the Cox proportional hazards model
(McNiel and Binder 2007). While random assignment of individuals to the program group
should account for confounding variables, the hazard model includes covariates to account for
baseline characteristics to improve estimation precision—similar to adding covariates in the
regression analysis. The hazard model estimates a hazard ratio, which is the probability of an
event occurring at a specific time, given that the event has not already occurred. The survival
analysis assessed the impact of assignment to the program group on the time to first arrest,
conviction, and incarceration during the 18 months following random assignment. A hazard
ratio of 1 indicates that those in the program and control groups have a comparable probability
of recidivism; a hazard ratio less than 1 indicates that those in the program group who have not
yet recidivated have a lower probability of recidivism in the next period compared to the
control group; and a hazard ratio greater than 1 indicates that those in the program group who
have not yet recidivated have a higher probability of recidivism compared to the control group.
Using a hazard ratio of 0.75 as an example, a more precise interpretation is that an individual

108
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

from the program group that has not already recidivated by a specified time has 0.75 times the
chance of recidivism by the next specified time compared to an individual from the control
group. The hazard ratio can be converted to probabilities (shown in Equation 4), which provides
a more intuitive interpretation of the results (Spruance et al. 2004).
Hazard Ratio (HR) = odds = P / (1 – P)

(4)

P = HR / (1 + HR)
Therefore, a hazard ratio of 0.75 means that an individual in the program group who has not
already recidivated has a 43 percent chance of recidivating before an individual in the control
group. The results of the survival analysis served as a robustness test for the related indicator
measures pertaining to arrest, conviction, and incarceration reported in Chapter V.

Results of the Sensitivity Analysis
The exhibits in Chapters IV through VII reported the mean outcomes for the program and
control groups. Whether the difference between the groups was statistically significant was
assessed using the results of the regression analysis described earlier in this appendix.
Exhibit C-2 compares this approach with two others for estimating the impacts of assignment to
the SCA program group on the recidivism outcomes discussed in Chapter V. The three methods
are:
•

The simple difference in means (the standard model). These estimates are identical to
the ones reported in Chapter V.

•

Regression analysis with inclusion of the control variables listed in Exhibit C-1 and a
treatment dummy, and

•

HLM with inclusion of the same set of control variables.

Consistent with Chapter V, the sensitivity analysis relied on weight-adjusted data.
As revealed in Exhibit C-2, neither the inclusion of covariates nor inclusion of a multilevel
framework notably altered the conclusions. Thus, the impact estimates on recidivism are robust
to model specification. The ICCs were assessed to determine the need for using a multilevel
model. Generally, the ICCs were considered trivial.
The sensitivity analysis was additionally carried out for other outcome measures beyond
recidivism, such as services received. The results were consistent with the findings observed for
recidivism and were not included in the technical appendix for concision.

109
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit C-2: Impacts on Recidivism for the Full Sample,
Using Alternative Model Specifications
Difference
in Means

Regression
with Covariates

HLM with
Covariates

A. Outcomes Measured from Administrative Data
Arrests
Arrested
Average number of arrests
Arrests by offense type a
Violent crime
Property crime
Drug crime
Public order crime
Convictionsa
Convicted of a crime
Average number of convictions
Incarcerations (prison or jail)
Was re-incarcerated in prison or jail
Experienced a new jail incarceration
Experienced a new prison incarceration
Total days incarceratedb
B. Outcomes Measured from the Survey
Arrests
Arrested
Average number of arrests
New charges and convictions
Formally charged with a new crime
Convicted of a new crime
Average number of new convictions
Parole/probation violations
Charged with a violation
Probation/parole revoked
Re-incarcerations (prison or jail)
Was ever re-incarcerated in prison or jail
Average number of re-incarcerations
Currently incarcerated

3.07
0.30**

3.40
0.23*

4.40
0.25**

-2.44
8.06***
1.78
1.35

-1.64
6.55***
1.29
1.69

-1.80
7.48***
1.55
2.19

6.43**
0.13***

5.90*
0.12***

6.07**
0.12***

4.61
2.97
2.33
1.31

4.65
2.39
2.70
-0.12

5.14
2.71
0.71
-1.38

5.29
0.04

5.75
-0.01

6.79*
-0.00

3.46
2.60
0.06*

3.31
1.96
0.06

3.84
1.92
0.06*

1.51
4.66*

2.03
4.37*

2.60
3.74*

3.23
0.11
-2.55

4.41
0.07
-1.11

4.79
0.07
-1.61

Notes: Numbers in the table show the estimated impacts of assignment to the program group. Estimates
shown in the first column represent a simple difference in-mean outcomes for the treatment and control
groups; these numbers correspond to the impact estimates shown in the main body of the report; significance
levels are calculated based on a t-test for differences in group means, assuming unequal variances. The second
column shows regression-adjusted estimates of the treatment effect, calculated by including covariates. For
outcomes that are dichotomous, logit models were used, and the treatment effect was calculated as the
difference in the predicted outcome calculated separately for the treatment and control groups, estimated at
the mean value of all covariates and with the predicted outcomes converted to a predicted probabilities. The
third column proceeds similarly but was estimated using HLM.
Source: Administrative data from state and local agencies and 18-month survey data
a

The sum across categories exceeds the percent ever arrested because individuals can be arrested more than
once and with different arrest charges in the 18-month follow-up period
b

For those incarcerated at the time of RA, total days includes days incarcerated following RA but before release.

*/**/*** Statistically significant at the .1/.05/.01 level.

110
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit C-3 reports results from the hazard models on time to first instance of re-arrest, reconviction, and re-incarceration. There was no statistically significant difference between the
program and control groups regarding time to first arrest or first incarceration. However, those
in the program group had a somewhat greater chance of being convicted before the control
group. These results are broadly consistent with the findings reported in Exhibit V-3, which
examined the percentage of program and control-group participants who experienced rearrest, reconviction, and re-incarceration, further evidencing the robustness of the results to
model specification.

Exhibit C-3: Hazard Ratios for Time to Recidivism for the Full Sample
Hazard Ratio
1.12
1.29**
1.11
1.13

Time to first arrest
Time to first conviction
Time to first jail admission
Time to first prison admission

Hazard Ratio
with Covariates
1.15
1.28**
1.14
1.12

Notes: Numbers in the first column represent the results of the hazard model with only program participation
as a predictor; the second column is a replication of the hazard model, including covariate.
Source: Administrative data from state and local agencies
*/**/*** Statistically significant at the .1/.05/.01 level.

111
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Appendix D. Defining Risk Subgroups
Some research has suggested that re-entry services are most effective for those at higher risk of
recidivism and, in fact, in some cases can increase recidivism if targeted to low-risk individuals
(Bonta and Andrews 2007, Cullen and Gendreau 2000, Latessa and Lowenkamp 2006, Lipsey
and Cullen 2007). For this reason, we estimated impacts for subgroups defined based on the
relative risk of recidivism.
The first step was to define the subgroups. One approach would have been to draw on the
validated assessment instruments used by the grantees in determining access to SCA and
developing service plans. However, these scores were not available to us.34 As an alternative,
we used a regression-based approach described by Kemple and Snipes (2001). This approach
takes advantage of the fact that, because of random assignment, the control group constitutes
a pool for whom the determinants of recidivism in the absence of SCA can be identified.
Steps were as described below.
1.

Identify the key outcome of interest. In our study, recidivism was measured in several
different ways (e.g., arrests, convictions, and incarcerations; severity of charge; number
of instances), using both administrative and survey data. For purposes of defining the
risk subgroups, we used as the key outcome whether the individual was ever reincarcerated in the 18 months after RA, measured using administrative data. We chose
this variable, because it corresponds to the outcome measure used for the study’s
confirmatory analysis (see Chapter I).

Identify determinants of re-incarceration. In the absence of having data on dynamic risk
factors, we used static risk factors associated with the “second generation” of risk
assessments (Andrews et al. 2006), which have been found to be quite good as
predictors of recidivism (Gendreau et al. 1996). Explanatory variables we used included:
a. Demographic characteristics, specifically age and gender, and
b. Criminal history (measured prior to RA), including total number of times
incarcerated (one, two to four, or five or more times), whether incarcerated at the
time of RA, and total days incarcerated in the ten years prior to RA (divided by 365,
to convert to fractional parts of years).

Even if those scores had been available, a problem with using them is that the seven grantees used different
assessment instruments. Although many instruments in general use have been shown to be comparable as
predictors of recidivism (e.g., Gendreau et al. 1996, Kroner and Mills 2001, James 2015), different instruments
need not yield the same measure of risk for a given individual (Baird 2009, Baird et al. 2013). Therefore, when
used in a pooled sample, scores from different assessment instruments used by different grantees might not
yield comparable evidence of risk.
113
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Model the relationship between the outcome and the predictors. We used logit analysis
to estimate the relationship between the predictor variables and the probability of reincarceration. As noted, this relationship was modeled based on the control-group
sample only, because participation in SCA could temper the underlying risk of
recidivism. Coefficients from the estimation are shown in Exhibit D-1.
Exhibit D-1: Coefficients from a Logit Model Predicting the Probability
of Re-incarceration within 18 Months after RA, for the Control Group
Coefficient

Standard Error

Intercept

.6158

.6640

Male

.6135

.2948**

Age

-.0452

.0136***

Incarcerated 2-4 times

.5810

Incarcerated 5 or more times
Incarcerated at RA

.4148

1.2009

.3938***

-1.1024

.3276***

Time incarcerated in the prior
ten years

.0357

.0482

Source: Administrative data and baseline information.
*/**/*** Statistically significant at the .1/.05/.01 level.

Apply the coefficient weights to create a risk score for each individual. Coefficients from
the model were used to estimate a risk score for each person in both the program and
control groups.

Divide the sample into risk subgroups. The sample was divided into two roughly equal
groups, a lower-risk group and a higher-risk group. The predictive utility of this
classification is demonstrated in Exhibit D-2, which shows the re-incarceration rate for
the two risk groups, measured for the control group 18 months after random
assignment.

Zweig et al. (2010) note that the above procedure tends to over-predict the probability of reincarceration for the control group. To correct this problem, they recommend dividing the
control-group sample into two equal halves, and estimating the logit model with one-half of the
sample and defining the control-group risk subgroup using the second half. However, small
sample sizes made this refinement infeasible.
Note that, to be eligible for SCA, individuals needed to be determined to be at medium or high
risk of recidivism based on whatever assessment instruments the grantee used and given its
target population (e.g., females versus males, those incarcerated in prison versus jail).
Therefore, the risk groups we defined represent those at relative risk of recidivism within this
constrained set.

114
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Exhibit D-2: Percent Re-incarcerated for the Control Group,
by Risk Category
60%
53.2%
50%

40%
31.7%
30%

20%

10%

0%
Lower risk

Higher risk

Notes: The bar chart shows the percent of the control group, divided into lower-risk and higher-risk
subsets, who were re-incarcerated within 18 months after RA.
Source: Administrative data and baseline information.

115
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice