The author(s) shown below used Federal funds provided by the U.S.
Department of Justice and prepared the following final report:


Document Title:
Evaluation of the Breaking the Cycle
Demonstration in Birmingham, Alabama: Final
Report



Author(s):
Adele Harrell ; Alexa Hirst ; Ojmarrh Mitchell ;
Douglas Marlowe ; Jeffrey Merrill

Document No.:
189244

Date Received:
08/16/2001

Award Number:
97-IJ-CX-0013


This report has not been published by the U.S. Department of Justice.
To provide better customer service, NCJRS has made this Federally-
funded grant final report available electronically in addition to
traditional paper copies.



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.



t
b
[’ . 2
TABLE OF CONTENTS
CHAPTER 1: INTRODUCTION ................................................................................................ 3
The BTC Vision ................................................................................................................... 3
BTC in Birmingham ............................................................................................................
5
Organization of Report ........................................................................................................ 7
CHAPTER 2: THE IMPLEMENTATION OF BTC IN BIRMINGHAM .............................
8
Interagency Collaboration ................................................................................................... 8
Judicial Oversight ..............................................................................................................
10
Implementation of the BTC Model ...................................................................................
11
Lessons on BTC Implementation ......................................................................................
12
Compliance Monitoring and Sanctioning .........................................................................
10
CHAPTER 3: THE IMPACT EVALUATION DESIGN AND METHODS ............
13
The Conceptual Framework ..............................................................................................
13
Data Collection ..................................................................................................................
14
Data Analysis ..................................................................................................................... 18
Selection Bias ....................................................................................................................
20
Analysis Techniques .......................................................................................................... 21
CHAPTER 4: BTC IMPACT ON OFFENDERS ...................................................................
22
Reductions in Drug Use .................................................................................................... 22
Reductions in Criminal Activity ........................................................................................ 24
Official Arrests .................................................................................................................. 27
Self-Reported Arrests ........................................................................................................
28
Self-Reported Offenses ..................................................................................................... 29
CHAPTER 5: THE IMPACT OF BREAKING THE CYCLE ON EMPLOYMENT.
FAMILY AND HEALTH PROBLEMS .........................................................................
34
Health Problems ................................................................................................................ 35
Employment Problems ...................................................................................................... 36
Conflicts with Family and Friends .................................................................................... 37
CHAPTER 6: CHANGES IN CASE PROCESSING AND OUTCOMES DURING BTC
IMPLEMENTATION ...................................................................................................... 39
Major Court Innovations During BTC ..............................................................................
39
Case Handling, Disposition, and Sentencing Before and During BTC ............................ 42
The Effect of BTC on Case Handling, Disposition, and Sentencing ................................
45
Summary ............................................................................................................................
48
CHAPTER 7: SUMMARY AND DISCUSSION .....................................................................
50
PROPERTY UF
National Crimiilai Justice Reference Service (NCJRS)
Box 6000
Rockville. MD 20849-6000
i
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
E
APPENDICES ............................................................................................................................... 53
P
Appendix A - Glossary of Research Variables
Appendix B - Analysis of Sample Attrition
Appendix C - Addiction Severity Index
Appendix D - Addiction Severity Index Follow-Up Interview
..
11
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
3
CHAPTER 1
INTRODUCTION
Brealung the Cycle (BTC) is a multi-site research and demonstration project designed to
develop and test a comprehensive strategy for addressing substance abuse among offenders. In
1996, Birmingham, Alabama was selected as the first BTC demonstration site. When fully
implemented, BTC in Birmingham would target its services to all drug-involved adults under the
supervision of criminal justice agencies. The University of Alabama at Birmingham’s (UAB)
Treatment Alternative to Street Crime (TASC) program was selected as lead agency. Since that
time, Jacksonville, Florida, and Pierce County (Tacoma), Washington have also started BTC
projects that target adult offenders. These three demonstration sites are being evaluated by the
Urban Institute (UI) and the Treatment Research Institute (TRI). UI managed the evaluation,
conducting the process evaluation site visits, collecting data on program services and justice
system activities, and analyzing the impact of BTC. TRI recruited the evaluation samples,
designed the instruments managed the surveys, and is conducting the cost-benefit analysis. One
BTC program for juveniles has begun work in Eugene, Oregon and is being evaluated by the
Research Triangle Institute. All the research and demonstration projects are managed by the
National Institute of Justice with funds provided by the Office of National Drug Control Policy.
This report presents findings on the impact of the Birmingham BTC program on
offenders and the criminal justice system and an analysis of the costs and benefits of BTC
services.
The BTC Vision
BTC is grounded in several decades of research that documents the effectiveness of
court-supervised treatment for offenders through drug courts and communi ty-based treatment
administered by Treatment Alternatives to Street Crime (TASC) programs (see Belenko, 1999;
Anglin, Longshore and Turner, 1999). However, access to these programs is generally limited to
offenders who meet selection criteria defined by charge, criminal history, or type of sentence.
Prior to BTC, no jurisdiction had offered an intervention to all felony offenders involved with
drugs (tailoring the services to the seriousness of the abuse), independent of their case and
criminal history. The goal was to ensure that criminal justice agencies focus on the challenge of
reducing drug use and drug-related crime among offenders under supervision. The following four
elements represent the core of the BTC model:
Early Intervention
The BTC model calls for identifying offenders who are eligible for drug treatment
immediately after they are arrested. An arrest can provide the best opportunity to
intervene, because it may force an individual to confront his or her substance abuse
problem. To capitalize on this moment of opportunity, BTC plans include pre-arraignment
drug testing of all offenders between arrest and first appearance. This should be followed
by a clinical assessment and timely placement in an appropriate treatment modality for
those with a positive drug test or other indicators of substance abuse.
3
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
0 Judicial Oversight
BTC requires close judicial oversight of drug treatment participation. The experience of
drug courts has shown that close judicial oversight can help reduce drug use and criminal
behavior among participants. BTC seeks to apply this lesson to all drug-using defendants
who are under any form of criminal justice supervision. In most jurisdictions, judges have
broad authority to impose and enforce conditions of pretrial release that impact public
safety. For judicial officers to exercise this authority, they need quick access to drug test
results and treatment participation information at every court hearing.
Use of Graduated Sanctions and Incentives
The agencies involved in BTC are expected to apply steady leverage to retain offenders
in treatment. Borrowing from strategies pioneered by drug courts, BTC offender
management should include consistent and timely use of sanctions and incentives. In
concordance with the drug court philosophy and social science research findings,
sanctions should be immediate and certain, graduating in severity as needed, and
incentives should be used to reward treatment progress.
Close Collaboration Between Criminal Justice Agencies and Drug Treatment
BTC requires justice agencies and treatment providers to collaborate in focusing their
expertise and mandates on changing offender behavior. As partners, these agencies
should engage in joint planning, exchange of information on offender status, and
collaborative monitoring of offender compliance. Operationally, this means setting up
procedures for three purposes: (1) to assess every drug-using defendant entering the
criminal justice system, (2) to prepare individualized treatment plans and conditions, and
(3) to encourage judicial review of treatment participation or drug testing at each court
appearance.
BTC is designed to provide an appropriate intervention for drug-involved offenders at
every stage of criminal justice supervision, from arrest through the completion of probation or
post-incarceration supervision. Figure 1.1 presents an overview of the BTC strategy, which
includes a continuum of pretrial treatment options, such as jail-based programs for those detained
and post-adjudication treatment for offenders in prison, in jail, and on probation.
4
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
D
Residential
Figure 1.1. The BTC Intervention Strategy
Residential
Outpatient
Day
Outpatient
Day
I
I
I
I
,,)
Detained (,
,,)TI
{ Case-Managed
Parole
1
Jail-Based
Treatment
Treatment
as Needed
After-Care
BTC in Birmingham
The strategy for implementing BTC in Birmingham involved the following system
changes :
a
a
a
e
Procedures for early intervention, careful case management, and proper treatment
referrals that would match the level of supervision and treatment to defendant needs.
Judicial review of all BTC defendants’ records of treatment participation and drug
testing at each court appearance as a means of improving treatment retention and
compliance with drug testing requirements.
Appropriate and consistent use of graduated sanctions to support justice system
requirements.
A continuum of services that would be provided to offenders throughout their period
of criminal justice supervision.
Ongoing collaborative planning by the justice agencies in Birmingham for the design
and enactment of global change in the criminal justice system.
An overview of drug treatment and supervision plans submitted by Birmingham for BTC
is shown in Figure 1.2.
5
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
I
~~
Figure 1.2. Birmingham Plans for BTC
Break the Cycle
Arrest
Jail Screening
Unnalysis
Development
I
Pretrial
I
I
I
TASC Case
Management
Random
UrlMlySlS
Drug Treatment
Enharcements
Jobs Program
I
I
1
I
I
Electronic
Monitonq
Day Reponing
Deterred
Prosecution
Supervised
Pretrial Release
Jail Treatment
I
Presentence
I
I
I
I
I
I
I
TASC Case
Management
Random
UfiMlySlS
DNQ Treatment
Enhancements
Jobs Program
EleCtrOnK
Monitonng
Day Reponing
Sentence
Planning
I
Drug Court
TASC Case
Management
Random
UnMlysis
I
1
I
I
I
I
Drug Treatment
Enhancements
Jobs Program
ElectrCmK
Monitonng
Day Reponing
Commumly
Semce
I
Probation
TASC Case
Management
Random
Unnalysis
I
I
I
I
I
Drug Treatment
Enhancements
Jobs Program
Eleclrcmr
Monttonng
Day Reporting
I
Servce
Community
Corrections
I
I
I
I
I
I
I
TASC Case
Management
Random
u n na I ys I s
DNQ Treatment
Enhancements
Job Program
Electrcmtc
Monttonng
Day Reporting
Community
Servce
Corrections
Treatment
Source: The Birmingham BTC Proposal to NIJ
Birmingham’s plans called for the BTC intervention to begin with screening, drug
testing, and case development shortly after arrest and to continue with services across all stages
of criminal case processing, from pretrial through community corrections. The BTC services at
each stage were to include case management, drug testing, jobs programming, drug treatment
enhancements, and a range of supervision options, such as day reporting and electronic
monitoring.
The implementation of this plan was divided into three phases to allow time for BTC to
devise and test procedures before instituting them system-wide. The Planning Phase began in
October 1996 and ended in May 1997. During Phase I of implementation - June 1997 through
mid-August 1998 - BTC offered services to defendants arrested on felony drug charges. The
design of Phase I allowed the BTC network to test new policies and procedures; to reorganize
staffing, technology, and operations as needed to support the inclusion of a large portion of the
arrestee population; and to begin the process of designing services that could continue from
arrest to the end of a defendant’s period of justice system supervision. Phase 11, which extended
BTC to all felony defendants, began August 10, 1998. This evaluation is based on BTC
operations between October 1998 and May 1999, the period of most complete implementation.
The process evaluation findings from each of the three phases are presented in earlier reports
from the Urban Institute, Baseline Report on Birmingham, Alabama ’s Breaking the Cycle
6
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Initiative (Carver, Harrell, and Cavanagh, 1998), and Process Evaluation Report on Phase I
Implementation in Birmingham, Alabama (Harrell, Cavanagh, and Hirst, 1998), and.
Implementing Systeur- Wide I~itenientions For Drug-Involved Offeeliders Iii Bin?iiiigliani,
Alabama: Evaluation Of The Breaking The Cycle Denzonstratioii (Harrell, Hirst, and Mitchell,
2000).
Organization of Report
This report examines the impact of BTC on offenders and the criminal justice system
during the project’s full implementation (Phase 11). Chapter 2 describes the services provided
during the implementation phase and summarizes the findings of the process evaluation report.
Chapter 3 presents the methodology used in the impact evaluation. Findings about effects of
BTC on offenders are presented in Chapter 4 (drug use and crime) and Chapter 5 (health,
employment, and family problems). The changes in case processing and outcomes that occurred
with the introduction of BTC are described in Chapter 6. Chapter 7 contains a summary and
discussion.
7
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
E
CHAPTER 2
THE IMPLEMENTATION OF BTC IN BIRMINGHAM
This chapter summarizes the findings from the process evaluation of BTC
implementation, providing information on the services delivered by BTC and a context for
interpreting the results of the impact analysis.
Interagency Collaboration
BTC envisioned close collaboration among justice agencies and treatment agencies, with
TASC serving as the linking agency. Under BTC, TASC would also be responsible for screening
and assessing defendants for BTC eligibility and TASC case managers would be responsible for
the supervision of BTC clients on pretrial release. During the planning phase of BTC (prior to
offering services to offenders), TASC worked to arrange agreements on joint efforts to supervise
and treat offenders, meeting individually with each justice agency and treatment provider. This
management model proved unequal to the tasks of getting agreement and action on procedures
for exchanging information and amending policies that affected a number of agencies. BTC then
established a Policy Board comprising representatives from the courts, the jail, parole and
probation, the defense bar, prosecutor’s office, TASC, and the sheriff’s department. During
Phase I and full implementation, the Policy Board met regularly to review the progress of the
project and make recommendations on program and system changes. Smaller subcommittees met
to discuss issues such as District Court judges, probation, and MIS development and reported
their findings to the full committee. Major accomplishments of the Policy ’3oard include the
following:
0 Development of a Memorandum of Understanding (MOU) during the Planning Phase
that outlined the responsibilities of each agency and demonstrated commitment to
BTC (See Attachment A for a copy). The Presiding Circuit Court Judge, the District
Attorney, the Sheriff, the Probation Supervisor, the President of the Criminal Bar
Association, directors of three drug treatment agencies, and the President of the
County Commission signed the MOU.
Implementation of a new bond condition during Phase I that required all felony
defendants to report to TASC (the BTC lead agency) upon pretrial release so that they
could be screened for BTC eligibility.
0
Implementation of methods for early identification of drug-using defendants during
Phase I. Substance abuse screening for released defendants took place at BTC
following release from jail and was facilitated when TASC opened a second office a
short distance from the jail and courthouse complex. Substance abuse screening of
defendants not immediately released on bond was scheduled for their initial court
appearance.
8
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
c
Establishment of alternative dockets for various types of dru,o-involved defendants.
These included a deferred prosecution program, a drug court, and an expedited docket
(Most of these developments occurred during Phase I of BTC implementation).
Introduction during Phase I of review hearings for probationers and BTC clients
awaiting grand jury review.
Early Intervention and Placement in Services
BTC made impressive accomplishments in Birmingham in achieving early case
identification and expanding the use of drug testing for defendants on pretrial release. BTC (1)
developed automated drug testing procedures capable of testing a large number of defendants;
(2) implemented a sophisticated management information system (MIS) for conducting client
assessments, trachng client supervision and drug test results, and generating court reports; (3)
placed case mangers near clients in offices in and near the court house; and (4) developed the
capacity for on-site drug testing in court. During the eight-month period from October 1998 to
July 1999, BTC offered intervention services to a large number of drug-involved felony
defendants, as summarized below.
Assessed and admitted 3,047 defendants into BTC services at TASC. Fifty-seven percent
of these assessments were available took place within one week of arrest.
0 Referred 2,562 BTC clients (84%) to drug treatment during pretrial release.
. 767 were referred to urine monitoring only.
. 530 were referred to day reporting or educational groups plus urine monitoring.
. 1,265 were referred to more intensive outpatient or residential treatment plus urine
monitoring.
. Sixteen percent of the BTC clients received no referral to treatment, although nearly
half of these received some drug testing.
0 BTC also placed most clients in treatment without long delays.
. 98% of the 1,297 referred to urine monitoring only or urine monitoring in
combination with educational groups entered their assigned program, most within a
few days. Over 90% of those who entered urinalysis testing only remained active in
BTC for 90 days or longer after entry.
90% of the 1,265 BTC clients who were referred to more intensive outpatient or
residential treatment entered their assigned program. The median waiting period was
just over two weeks. Over 60% of those who entered remained active in BTC for 90
days or longer after entry.
9
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
0 Drug tested 95% of the BTC clients at least once. The average number of tests scheduled
was 10.5 per client.
. Scheduled 30,922 drug tests for BTC clients during pretrial release: 52% of the tests
were negative, 23% positive, and 25% were missed.
Thirty percent of those scheduled for testing never tested positive; 12% tested
positive for heroin (alone or with other drugs); 33% tested positive for cocaine (alone
or with drugs other than heroin).
However, BTC early intervention and treatment placements were limited to defendants
released from the jail while their cases were pending. No screening, assessment, or treatment
services were provided to defendants not released from jail. Plans to set up these services were
abandoned in the face of severe space and staff shortages at the jail.
Compliance Monitoring and Sanctioning
requirements and treatment attendance for BTC clients on pretrial release and responding to
infractions with administrative sanctions. Overall, 86% of the clients with any infractions
received an administrative sanction. However, the sanctioning was not certain, swift, or severe.
BTC case managers were responsible for monitoring compliance with drug test
0 Multiple infractions tended to precede each sanction. BTC clients averaged 6.6
in frac ti ons , but on 1 y 2.6 sanc ti ons .
Sanctions occurred weeks after the first infraction in a series. The time between the first
infraction and the first sanction was over one month. This declined to three weeks
between the next infraction after the first sanction and the subsequent sanction, and then
to two weeks between the next post-sanction infraction and the subsequent sanction.
0 The sanctions were relatively mild and rarely graduated to severe penalties, despite
repeated violations. Sixty percent of the clients with infractions received an alert letter
notifying them that they were in violation of BTC requirements, 42% were subjected to a
case review by the case manager, 23% were terminated from BTC for noncompliance,
and 10% were referred to more intensive treatment. Those terminated faced no judicial
sanction for termination.
Judicial Oversight
District Court judges received reports from TASC on BTC client drug test results and
treatment compliance whenever those clients were scheduled to appear in court. Although the
MIS records on the contents of these reports were not available for analysis, it must be assumed
that many contained references to the 21,384 infractions committed by 2,509 (81%) of the BTC
clients. However, the judges did not regularly review these reports and, with the exception of
clients in drug court, few ever received a judicial sanction. In response to the lack of judicial
monitoring, BTC established new compliance hearings. Using the services of a retired judge, the
court began to hold compliance hearings for (1) defendants awaiting an indictment hearing
10
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
i
I
before a Grand Jury (typically a five month wait between the District Court waiver and the
Grand Jury), and (2) offenders placed on probation at sentencing. However, a very small
proportion of clients were referred to these hearings, appearance rates were low, and penalties for
non-appearance were rarely imposed.
368 BTC pretrial clients and 42 BTC clients on probation were scheduled for a review
hearing.
Appearance rates averaged 43% for the BTC pretrial clients and 72% for BTC clients on
probation.
72 pretrial clients (3% of the clients with any infractions) received a sanction involving
time in jail; 19 of these clients were in the drug court.
However, the availability of BTC supervision and treatment was a factor in the decision
of the court to expand the options available for drug-involved offenders by adding an expedited
docket and a diversion program to the drug court option. These programs increased the number
of drug-involved offenders released from the jail to BTC for testing and treatment and provided a
court endorsement of BTC.
Implementation of the BTC Model
BTC was successful primarily in subjecting more defendants on pretrial release to drug
testing and refemng them to treatment. Key areas in which BTC goals were not met included the
following:
0 Treatment for defendants detained at the jail before or after sentencing was part of the BTC
plan, but was not implemented due to lack of space.
Lack of coordination between TASC and the Probation Department led to a failure to drug
test and refer to treatment BTC clients who were on probation. During BTC, TASC
administered between 112 and 180 drug tests each month to a caseload of more than 2,500
probationers. A separate evaluation of the drug testing of BTC clients on probation found that
an average of 132.9 tests were conducted monthly between January 1998 and March 1999
(Yarber, 1999). Neither agency monitored drug treatment participation by these offenders.
Sanctions were not administered with a high level of certainty or speed, and did not
consistently increase in severity. This was true of administrative sanctions proposed for use
by TASC case managers and for sanctions available to the judges.
Judicial monitoring of BTC clients was minimal.
As a result, the drug interventions provided under BTC were far more similar to a pretrial
version of the TASC program in place at the start of BTC than to a model of coerced abstinence.
BTC was successful in screening defendants for substance abuse and placing drug users released
from jail in some form of intervention. Failure to use graduated sanctions was a particularly
11
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
D
serious flaw in implementation, because nearly half of BTC clients received drug testing only.
Drug testing in the absence of sanctioning (even when combined with judicial monitoring) has
been found to be of minimal effectiveness in producing drug abstinence (Harrell, Cavanagh, and
Roman 1999; Cavanagh and Harrell, 1995) Thus, the new system lacked provisions for ensuring
defendant accountability, a key element of BTC. The new system also was unable to provide a
continuum of care throughout the justice system. The problems encountered in setting up
treatment options in the jail, developing additional intensive drug treatment slots, and extending
community-based drug monitoring and treatment to BTC clients on probation limited the
program’s scope to defendants on pretrial release and may actually have decreased the level of
case management and treatment available to offenders on probation.
Lessons on BTC Implementation
envisioned by the project must be supported by technology, collaborative planning, and staff in
every participating agency. Many of the barriers to implementation can be traced to underlying
problems that made it difficult for agencies to undertake major reforms and offer expanded
servi ces :
One of the major lessons of BTC in Birmingham is that major system reform of the type
0 A severely overcrowded jail, excess case backlogs clogging the court dockets, and huge
caseloads for case managers at TASC and officers in the Probation Department. Staff
simply were not available to undertake additional responsibilities for offender
supervision.
Lack of computer systems and technology to support client trackmg and timely exchange
of interagency information.
Lack of a history of interagency collaboration around system problem solving and
agencies that were narrowly focused on their immediate budget and staffing problems.
No one agency can direct system-wide reform. Only collectively can agencies devise
ways to share resources to serve the interests of all.
The judges, the Sheriff, and the District Attorney were elected, so political considerations
reduced their willingness to take risks on reform and introduced an element of
competition between officials of different political parties.
Shortages in drug treatment slots made it difficult to find appropriate placements for
offenders in need of intensive treatment modalities.
Substantial progress was made toward addressing these bamers during BTC. Significant
reductions in jail overcrowding were achieved through alternative dockets and dispositions,
undertaken in conjunction with BTC assessment and case management. Computerized
assessment and drug testing were implemented. By establishing the Policy Board, BTC initiated
a continuing process of collaborative planning.
12
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
b
t
CHAPTER 3
THE IMPACT EVALUATION DESIGN AND METHODS
The goal of the impact analysis was to answer a series of questions about the effects of
BTC on individual offenders. Does BTC reduce criminal involvement and substance abuse?
Does BTC result in gains in legitimate employment, family cohesion, and residential stability?
How do features of the services and sanctions received affect outcomes for offenders? Does the
impact of BTC depend on characteristics of the participants? The impact evaluation also
examined the effects of BTC on the functioning of the criminal justice system by looking at
changes in the length of time required to reach a disposition, the number of hearings, and the
kinds of sentences imposed.
The analysis of the impact on individuals uses quasi-experimental design to compare a
sample of 137 offenders selected prior to full implementation to a sample of 245 offenders
eligible for the full range of BTC interventions. The impact of BTC on case processing time, use
of detention, sentencing, and compliance with court orders are evaluated using criminal justice
records of offenders entering the system before and after BTC. The analysis of the impact of
BTC on case processing is based on comparing records from criminal justice agencies on the
handling of their cases.
The Conceptual Framework
The conceptual framework guiding the study design and choice of data to be collected is
shown in Exhibit 3.1. The evaluation examined the outcomes illustrated ir the boxes on the far
right. Goals for offenders include decreased drug and alcohol use and negative consequences
associated with use, reduced criminal activity, longer time to re-arrest, improved economic well-
being and increased rates of employment, improved family and social functioning, and improved
physical and psychological health. System outcomes to be examined include the number of
hearings and number of days between arraignment and case disposition, top charge at conviction,
sentences imposed, use of alternatives to incarceration, and recidivism.
Offender characteristics that may affect both the type of services received and the
response are shown on the far left of the exhibit. These factors are used as control or
stratification variables. They include demographic or background characteristics of the offender
such as age, ethnicity, and gender; substance abuse pattern and severity; current employment and
educational status; family status and current living situation; physical and mental health; prior
criminal activity, arrests and convictions; and current charge.
The center column illustrates factors hypothesized to affect offender and system
outcomes. These include drug treatment placements, type and duration of drug treatment, drug
testing, frequency of judicial monitoring, intensity of contact with case managers or court
supervision staff, the types of incentives and sanctions, and the timeliness and consistency of
sanctioning.
13
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Exhibir 3.1 Conceptual Framework
Conceptual Framework for the Eva
Breaking the Cycle in Birming
Offender Characteristics
Demographic Characteristics
Substance Abuse History
Employment and
Educational Status
Family and Social
Functioning
Physical and Mental Health
Criminal History
Type and Seventy of Charge
uation of
iam
Outcomes
~
Interventions/Services
TASC Assessment and
Treatment Planning
Type and Duration
of Drug Treatment
and Procedures
Judicial Monitoring
Management and
Use of Incentives
and Sanctions
Services Received
I Systemchanges I
Data Collection
The impact of BTC was assessed using the results of surveys designed and managed by
TRI. The samples were interviewed shortly following arrest (baseline) and again nine months
later (follow-up) using a version of the Addiction Severity Index (McLellan, Kushner, Metzger,
Peters, et al., 1992) modified to include additional questions about illegal activities and
participation in drug treatment services. Copies of the questionnaires are provided in Appendix
A. Data on arrests were collected from criminal history records. Data on drug test results,
sanctions and infractions data, and participation in on-site drug education groups were collected
from the BTC management information system. The data sources for key domains are shown in
Exhibit 3.2. Definitions of the variables used in the analysis are shown in the Glossary at the end
of this report.
The pre-BTC sample was recruited between March 13 and May 2, 1997 by inviting
arrestees tested for the Drug Use Forecasting (DUF)’ project in the Birmingham jail to take part
in the study. Following the DUF drug test and interview, arrestees were invited by a research
recruiter to consent to be a part of the study. Those who agreed (n = 3 11) signed a consent form
14
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
B
which included agreement to their DUF drug test results to the research team with the
understanding that they would receive a $10 stipend by mail and would be contacted for the
study if they were found to be eligible. Only those who tested positive for at least one drug (n =
236,76% of those who consented) were considered eligible and included in the comparison
sample.
The BTC sample was recruited from the defendants ordered to BTC upon release. Plans
to recruit them following a drug test in the jail had to be changed when BTC dropped plans to
screen for program eligibility at the time of arrest. In lieu of in-jail drug testing, BTC required
defendants charged with felonies to report to TASC within 24 hours of release from the jail on
bond. The defendants were screened at that time for BTC eligibility using a drug test and short
self-administered questionnaire. Those who tested positive, reported drug use, or were charged
with drug felonies became eligible for BTC. The BTC sample was recruited immediately
following the TASC screening by inviting defendants found eligible for BTC to participate in the
study. Between September 8 and November 5, 1998, 596 defendants were contacted and initially
determined to be BTC eligible; 545 of these individuals agreed to participate in the study (91%)
and were sent a payment of $10. However, 171 of them were later found to be ineligible because
their charges were dropped or reduced to a misdemeanor, or they lived outside Jefferson County
and thus not eligible for BTC services, leaving a final sample of 374.
Baseline interviews were conducted with 192 pre-BTC sample members and 374 BTC
sample members. The 45-minute interviews were conducted by telephone (1 %) or in person
(99%). Pre-BTC sample baseline interviews took place approximately a month following consent
(median = 28 days); 63% took place in jail and 36% in person in the community. All BTC
sample baseline interviews were conducted in person at TASC within a day of consent.
Participants received $10 for the baseline interview. The non-random nature of the sample and
the varying sample selection criteria produced significant differences between the two groups,
particularly in areas concerning employment and cnminal histories; however, the samples were
similar on drug use variables (Table 3.1).
Although the pre-BTC had all tested positive at time of arrest, the baseline interviews
conducted some weeks later showed lower rates of drug use in the past 30 days. Their drug use
may have increased the risk of criminal behavior and detection -
leading to the arrest, and the
baseline interviews captured a 30-day period of less drug use
the comparison group may have
underreported their drug use. The somewhat higher rate of self-reported marijuana use in the
past 30 days reported by the BTC sample may have resulted from interviews conducted within a
week of arrest.
15
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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 3.2: Data Source Matrix
Demographic Characteristics
Substance Abuse History
Employment and Education Status
Family Composition and Living
I Drug Treatment
X
X
X
X
X
X
X
X
Category in Conceptual Framework
I MIS
I NClC
I AOC
I Survey
I Records
OFFENDER CHARACTERISTICS
Situation
Physical and Mental Health
Criminal History
Type and Severity of Charge
Court Assessmentrrreatment
I NTE RVE NT I ON S/S E RV I C ES
X
X
X
X
X
X
X
X
X
16
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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 3.1. Baseline Differences in Demographic Characteristics by Group and Completion Status (n=566)
BTC sample ( t i = 374)
Pre-BTC sample ( 1 2 = 192)
No
Follow-up Total
No
Follow-up (ti = Total
Follow-up
( I Z = 137)
( I Z = 192)
F o ~ ~ o w - u ~
245)
( I 1 =
(tz = 55)
(tz = 129)
374)
89%
82%
84%
828
Male
African-American
Unmarried
Mean Age in Years
Mean Years of Education
Mean ## Days Paid for Work, Past 30
Mean # Days Paid for Work. Past 6 Months
Mean Employment Income Past 30 Days
Received Public Assistance Past 6 Months
Type of offense (target arrest)'
Mean Age First Drug Use
Self-Report Drug Use Past 30 Days
Drug
Cocaine
Opiates
Marijuana
Other
Mean Baseline AS1 Composite Scores
(Range from 0 to 1)
Medical
Emplo ymenVSupport
Alcohol Use
Drug Use
Legal
Family/Social
Psvchiatric
73%
89%
30
12
7
66*(g)
$299
15%
35%
16
20%
0%
31%
6%
0.16
0.74
0.12
0.07
0.40
0.18
69%
88%
34*(f)
11
5***(9
48***(f)
*(@
$26 1
***(f)
14%
26%
16
32%
7%
30%**(f)
8 8
0.18**(f)
0.80***(9
0.14
0.08 * *( 9
0.40
0.17***(f)
708
898
32
I 1
5
54
$272
14%
16
29%
5%
30%
7%
0.18
0.79
0.13
0.08
0.40
0.17
618
90%*( g)
27
12
1 1
74
$546
7%
71%
16
28%
5%
57%
5%*(g)
0.07
0.66*(g)
0.09
0.04
0.40
0.08
7 8 8
79%
66%
64 Q
82%*(g)
85%)
29*(f)
30
12
12
13***(f)
12
82***(9
79
$673
$629
***(f)
10%
9%
66%
16
16
31%
30%
6%
6%
54%**(f)
55%
12%*(g)
9%
0. I 1 **(9
0.10
0.5 8 * **( f) *( g) 0.6 1
0.1 1
0.11
0.05 **( f)
0.05
0.4 I
0.40
0.08***(9
0.08
0.2 1
0.22***(tj
0.22
0.06
O.O7***(f)
0.07
Differences between follow-up/no-follow-up within group signified by *(e). Differences between groups with follow-up signified by *(f).
* P < .01 ** p < .01 *** P <.001
Target amestS in this table are considered preliminary. and may not be an accurate reflection of the target arrest charges for the entire sample.
Follow-up interviews that were similar to the baseline interviews were conducted by
phone. Most participants received $10 for completing the follow-up interview, although some
hard-to-contact participants received $20. Originally, two follow-up interviews were planned,
one at nine months and one at 15 months after sample recruitment. However, analysis of results
from comparison sample interviews found no significant differences in outcomes measured at the
two times so follow-up interviews for BTC sample were limited to the 9-month follow-up
interview. For the comparison sample, the 9-month follow-up was used when available (n = 113)
and the 15-month follow-up used if no 9-month follow-up was completed (n = 24). The actual
time between baseline and follow-up varied as a result, with the average length of time between
interviews was 290 days (median = 264 days). For the pre-BTC comparison sample, the time
between baseline and follow-up ranged from 92 days (for one respondent whose baseline
interview was conducted a long time after sample recruitment) to 599 days. For the BTC sample
the time between interviews ranged from 239 to 428 days.
17
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
I
The timing of sample recruitment and interviewing for the pre-BTC and BTC samples is
illustrated in Figure 3.1. As indicated, the pre-BTC sample was selected during the end of the
BTC planning phase, while the BTC sample was selected during the full implementation phase.
The figure also illustrates the extended period of baseline interviewing for the comparison
sample. For this group, the interviews were not scheduled until after test results were received.
Only at that time could efforts to locate and interview respondents begin. The BTC sample
baseline interview was conducted at the time of consent, which resulted in interviews much
closer to the time of arrest.
Figure 3.1: Sample Recruitment and Follow-up Periods
Mar
Jun
Sep
Dec
97
98
98
99
I
I_-
BTCPhaseI -!&TC
Phase I d
I
I Interview)
C
'1
Baseline
Interview
Comparison Group
Treatment Group
-Comparison
Figure 3.2 presents a pipeline analysis that illustrates sample selection and attrition rates.
The comparison sample began with 3 11 arrestees who consented to be part of the study. No data
were available on those who refused to talk with the recruiter, but very few of those who spoke
with the recruiter failed to sign a consent form. Of the 236 eligible pre-BTC sample members,
192 (81%) completed the baseline interview and 137 completed a follow-up interview (58% of
the eligible sample, 71 % of those interviewed at baseline and assigned for follow-up interview).
Of the 596 eligible sample members, 545 (91%) consented and completed a baseline interview.
However, 171 were subsequently dropped from the sample because they were later found to be
ineligible for BTC services, leaving BTC sample of 374 eligible defendants. Of these, follow-up
interviews were completed with 245 (66% of the eligible sample).
Data Analysis
Data analysis began by checking for potential biases introduced by sample attrition and
selection bias.
18
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Sample attrition
D
The attrition analysis included the traditional method of predicting response at follow-up
(yes or no) as a function of sample characteristics and methods proposed by Biglan et al. (1991)
that test the hypothesis that baseline risk scores vary significantly by group (BTC or
comparison), attrition (yes or no), or the interaction of these two factors. A more complete
description of the results of the attrition analysis is presented in Appendix B.
Figure 3.2: BTC Sample Attrition - Comparison and BTC Samples
Pre-BTC Comparison Sample
Contacted
Consented
Baseline
FoIIow-UP
Interview
Interview
BTC Client Sample
Contacted
ConsentedBaseline Remained Eligible
Follow-up
Interview
after Baseline
Interview
Interview
To examine threats to the internal validity of the comparisons, based on the survey data,
attrition analysis tested whether the attrition rate differed between group and whether the
characteristics of those who remained in the study differed as a function of group. The internal
validity of the study refers to the level of confidence that any measured differences between the
groups are due to the intervention rather than to extraneous factors. Sample attrition threatens the
internal validity of a study when subjects who drop out of one condition differ systematically
from those who drop out of another condition on characteristics that are related to the outcome
variables. The analysis found no evidence of differential attrition by group. No significant
interactions were found between group and attrition. This indicated no differential attrition by
group on key dependent variables that would have compromised the internal validity of the
study. Older subjects who completed follow-up interviews were more likely to have low alcohol
composite scores, while younger subjects interviewed at follow-up were likely to have high
alcohol composite scores. Older subjects who completed follow-up interviews were more likely
to have a drug charge, and younger subjects who completed follow-up interviews were more
likely to have a non-drug charge.
19
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
B
Overall Sample
Amition
Attrition as a Function
of Group
The external validity of the study is the degree to which the results of the comparison can
be generalized to conditions other than those under which the study was conducted. If subjects,
regardless of experimental condition, who drop out of the study are systematically different for
those who remain, then the analysis sample may not reflect the characteristics of the target
population that the original sample was designed to represent. The analysis found no significant
differences between the demographic characteristics, drug or alcohol use, or legal involvement of
respondents who stayed in the study and those who dropped out. The results of this analysis
suggest that attrition did not introduce additional differences into the study.
Differences Significant
Differences Not Significant
No differences
Baseline Drug Composite Score
Baseline Alcohol Composite Score
Baseline Legal Composite Score
No differences
Baseline Drug Composite Score
Baseline Alcohol Composite Score
Exhibit 3.3 summarizes the findings of the checks of internal and external validity.
Attrition a
-
Baseline Drug Composite Score
Baseline Alcohol Composite Score
Target Arrest (DrugNon-Drug)
I
I
Baseline Legal Composite Score
Differential Group
I No differences
1
Baseline Legal Composite Score
Selection Bias
Given the evidence of differences in the samples at baseline, the analysis uses two
strategies to control for these differences. The strategies include traditional multivariate models
that incorporate control variables to measure observed sample differences and a two-stage
estimation procedure designed to capture the effects of unmeasured sample differences
(Heckman, 1978, 1979). The two-stage method is used to assess whether unmeasured variables,
related to both treatment status &the
outcomes of interest (e.g., recidivism), lead to bias in the
estimates of BTC’s effect (Barnow, Cain, & Goldberger, 1980; Smith & Paternoster, 1990;
Winship & Mare, 1992). At the first stage, the likelihood of being a BTC sample member was
estimated using predictors believed to differentiate the two groups. The purpose of this first-stage
equation is to obtain a correction factor, which in essence is a proxy for unmeasured variables.
This correction factor is then included in a second-stage equation as an independent variable,
along with other variables hypothesized to effect the outcome of interest (see Winship & Mare,
1992; or Winship & Morgan, 1999).
The model selected for the first-stage equation was chosen on the basis of its predictive
power and parsimony (Table 2). The predictors of group membership were sex (Female = l),
number of days incarcerated during the 30 days prior to initial interview, current
probatiordparole status (Yes = l), number of self-reported crimes committed in the six months
prior to the baseline interview, lifetime number of times treated for drug abuse, usual work
20
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
P
t
pattern in the past 3 years (Full-time/Student = 1, Part-time= 2, Other is omitted from the
equation), time at current residence (in months), and number of days in the past 30 respondents
reported each of the following lunds of problems: drug problems, psychological problems, or
employment problems. These variables measure the constructs of criminal history, seriousness of
substance abuse problem, medical/psychologica1 problems, ties to the community, and
demographic factors. Collectively, these variables produce a pseudo-R‘ of .38. The addition of
more variables did not significantly improve the model fit to the data.’
Table 3.2 Sample Selection Model (First-stage Equation)
Variable
Estimate
Error
p-level
Constant
1.40
4.79
0.00
Female
0.34
1.64
0.10
-1.91
0.06
Time at current residence
-0.01
2.42
0.02
Full-time EmploymentIStudent”
0.56
2.21
0.03
Part-time Employment
0.57
Parameter
b/Std.
Days in Jail, past 30 days
-0.06
-7.84
0.00
On ParoleProbation at Sample Entry
-0.44
-2.45
0.01
Lifetime number of prior drug treatment episodes
-0.10
-1.48
0.14
Number of Self-Reported Offenses, past 6 months
-0.01
-2.41
0.02
Days experiencing Drug Problems. past 30 days
-0.02
-1.75
0.08
-0.02
-2.23
0.03
Days experiencing Employment, past 30 days
-0.03
-2.90
0.00
Days experiencing Psychological Problems, past 30 days
Model Fit
Pseudo-R2
0.38
-2LL
180.93; 11 DFp = 0.0001
N
382
responses, including “service.” “retireddisability,” “unemployed,” or “in controlled environment.”
The full-time and part-time employment variables are indicator variables; the suppressed category is all other
Analysis Tech 12 iqu es
Dichotomous drug and recidivism outcome variables were estimated using bivariate
probit analysis which simultaneously estimates the first- and second-stage models and the
correlation between the two error terms (Rho). This term corrects for selection bias (Smith &
Paternoster, 1990: 11 18).4 Models with counts as dependent variables (e.g., number of arrests),
were estimated using bivariate probit for the first stage and a separate negative binomial
regression for the second-stage. All models were estimated in LIMDEP 7.0 (Greene, 1995).
21
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
CHAPTER 4
BTC IMPACT ON OFFENDERS
This chapter utilizes the sample and the methodologies described in the previous chapter
to estimate Brealung the Cycle’s impact on client drug use and criminal activity.
Reductions in Drug Use
The analyses that follow test the general hypothesis that BTC reduced participant drug
use by comparing BTC and pre-BTC samples self-reported drug use at follow-up, controlling for
prior drug use and other factors hypothesized to affect drug use. Brealung the Cycle’s impact on
drug use was measured by self-reported drug use in the 30 days prior to the follow-up interview.’
The dependent variables include use of any drug (yesho), any stronger drugs (i.e., heroin andor
cocaine use) (yesho), and any marijuana during the 30 days prior to follow-up. Multivariate
probit analysis with and without the selection bias correction, described in the last chapter, were
used to test the hypothesis that BTC participants were less likely than the comparison sample to
report drug use on the follow-up interview. The independent variables in the models include (1)
BTC treatment, (2) number of days in jail in the 30 days prior to the follow-up interview (to
control for relative access to drugs), (3) demographic variables (sex, race, age, and education),
(4) drug use in the 30 days before the baseline interview (to control for individual differences in
severity of substance abuse problems), (5) several measures of employment and criminal history
(in order to control for differences observed between the BTC and pre-BTC samples at baseline),
and (6) any interactions between group and the above variables (in order to allow the above
variables to have varying effects conditional on BTC status). All of these measures, except
number of days in jail in the month prior to follow-up interview, were baseline measures. (See
the Glossary for definitions of the variables.)
BTC clients were less likely to report any drug use, any stronger drug use and any
marijuana use in the 30 days before follow-up than comparison sample members were. Table 4.1
displays the percentages of clients from each group who reported drug use in the 30 days prior to
the follow-up interview.’ While the absolute magnitudes of these differences are relatively small,
ranging from roughly 6% to 3%, the differences may be attenuated by the fact that respondents
from the pre-BTC sample had considerably less opportunity to use illicit drugs. The pre-BTC
sample, on average, spent 9 more days in jail than the BTC sample during the 30 days before the
follow-up interview and were still more likely to use illicit substances during that same period.
’ This drug use measure includes use of heroin, other opiates, cocaine, marijuana, amphetamines, barbiturates, other
sedatives, hallucinogens, and inhalants.
2 The sample is limited to known drug users and excludes 32 BTC sample members put in BTC for urine monitoring
because they had been charged with a felony drug offense and subsequently discharged because they did not test positive for
drugs while in BTC.
22
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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 4.1
Drug Use
(n = 213)
(n = 137)
Difference a
Any Drug Use
23%
26%
0.29
Any Stronger Drug Use
8%
12%
0.06
Any Marijuana Use
10%
16%
0.06
Selfreported Drug Use in the 30 days Prior to the Follow-up Interview
BTC sample
Pre-BTC sample
Significance of
Number of days incarcerated
1.9
11.5
0.00
a These significance tests are all one-tailed. Chi-square tests were used for the first four comparisons, and a t-test
was employed for the last comparison.
To isolate the effects of BTC, probit analysis of group differences in the likelihood of
reporting drug use in the 30 days before follow-up were conducted. The models included
variables to control for personal characteristics that might affect differences in drug use and drug
use in the 30 days before the baseline interview. Two models are shown for each drug use
outcome. The model in the first column estimates the impact of BTC on client drug use without
controlling for selection bias; the model in the second column adds a selection bias correction
factor to the model as described in Chapter 3. When the correction factor in the second model is
statistically significant, conclusions should be based on the model in that column. If the
correction factor is not significant, conclusions should be based on the model in the first column.
In general, the models with the selection bias correction did not significantly improve the models
fit to the data, suggesting that the control variables included in these models adequately control
for differences between the two groups.
Table 4.2 shows that BTC had a marginally statistically significant impact on any drug
use and use of stronger drugs in the 30 days before the follow-up interview, controlling for other
factors. Because the correction factor was not significant, these conclusions are based on the
model in the first column. Table 4.2 also shows that BTC had a more complex effect on
marijuana use. This model reveals that BTC’s effect was conditional on race; BTC had
statistically significant and substantial effects on marijuana use of African-American clients, but
no effect on the marijuana use of Whites. One way of describing the results in Table 4.2 is to say
that BTC participation is expected to reduce the latent propensity to use any drug by 0.29 and
any harder drug use by 0.35, holding all other variables constant.
Another way of describing the results is to convert these parameter estimates into
predicted probabilities (see Long, 1997) that show the likelihood of drug use, holding all
variables except BTC treatment at their means. These probabilities represent the “average” effect
of BTC on drug use, if all other variables were the same.3 Presented this way (see Figure 4. l), the
predicted probability of any drug use in the past 30 days is 17% for the BTC sample and 26% for
the pre-BTC sample, holding all other variables at their mean values. The percentage of BTC
sample members reporting any stronger drug use was half that of pre-BTC sample members (4%
versus 8%). The largest difference was in the probability of any marijuana use by African-
Americans (4% in the BTC sample versus 18% in the pre-BTC sample). However, there were no
significant differences in marijuana use among white defendants in the two samples.
3 Throughout this report parameter estimates all be converted into predicted probabilities when significant differences
emerge. These predicted probabilities should be interpreted in the manner given in the above example.
23
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
I
Table 4.2 Selj-Reported Drug Use in the 30 Days Prior to Follow-up Interview, controlling for
Defendant Characteristics and Sample Differences
Any Drug Use
Anv Hard Drug Use
Any Mariiuana Use
Probit wlo
Probit
Probit w/o
Probit
Probit w/o
Probit
Variable
Selection
w/Selection
Selection
w/Selection
Selection
w/Selection
Correction
Correction
Correction
Correction
Correction
Correction
-0.29*
-0.53*
-0.35*
-0.60
0.00
-0.53
BTC Treatmenta
Female
Black
Black*BTC Tx
Education
Employ Bother
Days Worked
Months in Jail
On Probation
Serious Offender
Prior Offenses
Prior Drug Use
Days in Jail
Constant
Selection Correctionb
Age
N
-7 r
0.01
-0.15
-0.02
-0.07*
-0.06
-0.01
0.01
0.25
-0.21
0.00
0.01
-0.05 ***
0.24
-
-
350
0.01
-0.13
-0.03
-
-0.07*
-0.09
-0.01
0.0 1
0.20
-0.22
0.00
0.01
-0.05***
0.42
0.19
350
0.03***
0.04
0.29
-0.12**
-0.09
-0.01
0.00
0.17
-0.14
0.00
0.01
-0.05***
-0.80
-
-
350
0.03**
0.05
0.27
-
-0.13
-0.1 1
-0.0 1
0.00
0.1 1
-0.16
0.00
0.0 1
-0.05 ***
-0.60
0.19
350
-0.0 1
-0.65**
0.38
-0.84**
-0.04
0.12
0.0 1
0.0 1 *
0.49**
-0.10
0.00
0.02***
-0.04***
-0.59
-
350
-0.01*
-0.62
0.32
-0.78
-0.05
0.08
0.01
0.01
0.38
-0.12
0.00
0.02**
-0.04***
-0.2 1
0.37
350
-LLL
-344.98
-6 15.05
-181.56
-45 1.89
-224.65
-493.66
Significance tests for this variable are one-tailed.
This term refers to the correlation between the error terms in the hrst- and second-stage equations (Rho).
*pHowever, examination of reports of drinlung alcohol to intoxication in the 30 days before
the follow-up interview (reported by 16% of the BTC sample and 12% of the pre-BTC sample)
did not show any significant differences as shown in the results of the multivariate modeling
testing for group differences (Table 4.3). This suggests that the focus of BTC on illegal drug use
through testing, treatment and supervision did not have a carry-over effect on alcohol abuse.
Reductions in Criminal Activity
One of the key premises of the Brealung the Cycle program was that recidivism could be
reduced if drug-involved arrestees are promptly identified and referred shortly thereafter to
appropriate treatment modalities. The following analysis utilizes official arrest and self-report
data to test the hypothesis that BTC reduced continued criminal activity. The analysis also
assesses whether BTC reduced the likelihood of any recidivism, and whether BTC reduced the
number of offenses committed in both the official and self-report data.
24
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Figure 4.1 Predicted Values for Drug Use
Drug Use* I
26%
Any Self-Reported
Cocaine Use*
BTC
' OComparison
Any Self-Reported
Mariiuana Use
I
1 goyo
Black**
10%
V Y I lllG
0%
10%
20%
30%
' p < 0.1
** p e 0.05
*** p < 0.01
Table 4.3
Drank Alcohol to Intoxication Weekly in 30 Days Prior to Follow-up
Probit w/o Selection Bias
Probit w/Selection Bias
Variable
Correction
Correction
BTC Treatment
0.16
0.12
Female
-0.23
-0.17
Black
0.08
-0.01
Education
-0.08*
-0.06
Employment Bothers
0.33
0.33
Months in Jail
0.01
0.01
On Probation
0.37*
0.29
Serious Offender
0.10
0.07
Prior Alcohol Use
0.02**
0.02**
Days in Jail past 30 days
-0.05***
-0.05 ***
Constant
-0.76
-0.8 1
Selection Correction'
-
0.00
A s
0.01
0.00
Days Worked
0.01
0.01
Prior Self-Reported Offenses
0.00
0.00
N
350
350
-2LL
-259.79
-530.03
7hs term refers to the correlation between the error terms (Rho).
*p i 0.10; **p < 0.05; ***p c 0.01
In order to test whether BTC reduced recidivism, respondents at follow-up were queried
about the number of times they had committed 14 types of offenses in the 6 months prior to the
25
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
B
I
follow-up interview, and respondents were asked how many times they were arrested for these
offenses arrested for these same 14 types of offense^.^ Additionally, official criminal justice
records were collected on the BTC and pre-BTC samples. These data were used to measure any
arrest in the 12 months after sample entry and the number of arrests in the 12 months after
sample entry. However, criminal history records could not be located for 179 of the 566 baseline
interviewees; 124 of these missing criminal histories were BTC sample members and the
remaining 55 were comparison sample members.
Table 4.4 reports the mean number of arrests for the two groups and the percentage of
each group who were arrested at least once in the six months prior to the follow-up interview.
This six-month time period is presented because it allows a direct comparison between the self-
reported and the official arrest data, allowing the data in the table to serve as a simple reliability
test of the self-report data. To the extent that self-reported arrests comport with the official
records, confidence in the self-report data is bolstered.
Table 4.4
Mean Number of Arrests in the Six Months prior to Follow-Up Interview
BTC Treatment
Comparison
Significance of
(n = 222)
(n = 137)
Differencea
Average number of total self-reported arrests
0.27
0.57
0.001
Any self-reported arrests
16%
32%
0.001
Average number of official arrests
0.13
0.61
0.001
Any official arrest
9%
39%
0.001
*These significance tests are all one-railed A t-test was used for the first comp~son,
and a chi-square test was employed for the last comparison.
The data in Table 4.4 present two key preliminary findings. First, official records and
self-report data both indicate that the BTC sample were arrested substantially less often than the
pre-BTC sample. However, as noted in Chapter 3, the BTC sample had significantly less
involvement in crime prior to BTC. For this reason, these observed differences could be due to
pre-existing differences between the two groups. The analysis that follows controls for individual
and group differences to minimize this risk.
The comparison of self-reported arrests to official arrests shows differences in the two
samples in the percentage of self-reported arrests showing up in the official arrest records. In the
pre-BTC sample, percentage reporting an arrest was similar to the percentage with an official
arrest record (32% versus 39%). The BTC sample was less likely to report arrest and less likely
to have an official arrest recorded. However, the two estimates are farther apart, with 16%
reporting an arrest, but only 9% having an arrest recorded. Two explanations suggest themselves,
either: (1) the official records did not capture all of the arrests incurred by the BTC sample, or
(2) the BTC sample counted many arrests from earlier time periods in the six-month period
before the interview (a recognized memory error known as "telescoping"). Unfortunately, it is
impossible to definitively reject one of these hypotheses. This complication vividly illustrates the
importance of having more than one data source in analyzing criminal offending outcomes. To
the degree that separate analyses of self-report and official data comport, added confidence is
given to each set of results.
4 The fourteen types of offenses were: shoplifting or vandalism; parole or probation violations; drug offenses; forgery;
weapons offenses; burglary, larceny. or breaking and entering; robbery; assault; arson; rape; homicide or manslaughter;
prostitution; contempt of court; and any other offenses.
26
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Official Arrests
b
The analysis first assesses BTC's effect on officially recorded offenses in the twelve
months after sample entry. The models control for defendant characteristics and criminal history
to examine the independent effect of BTC on recidivism. Two models are shown for each
dependent variable. The first omits the selection bias correction. The second includes the
selection bias correction described in Chapter 3. Negative binomial estimation is used for the
number of arrests and bivariate probit estimation is used for the likelihood of any arrest.
The results displayed in Table 4.5 demonstrate that BTC significantly reduced both the
likelihood of arrest and the number of arrests incurred in the 12 months after sample entry,
controlling for other variables. Table 4.5 also reveals that the rates and numbers of arrests were
significantly lower for the BTC sample than for the pre-BTC sample, regardless of race.
Table 4.5. The Probability of Arrest and Number of Arrest in the 12 months After Sample Entry
Anv Arrests
Number of Arrests
Probit w/o
Probit w/
NegBin w/o
NegBin w/
Selection
Selection
Selection
Selection
Variable
BTC Treatmenta
Age
Female
Black
Black*BTC Tx
Education
Employ Bother
Days Worked
Months in Jail
On Probation
On Probation*BTC Txb
Serious Offender
Prior Arrests
Prior Offenses
Prior Drug Use
Constant
Selection Correction'
Overdispersion Parameter
Correction
Correction
Correction
-1.25***
-0.83**
-1.61 ***
-0.01
-0.05
-0.36
0.65**
0.03
0.12
-0.0 1
0.00
-0.18
0.63**
0.05
0.02
0.00
0.00
0.43
-0.01
-0.09
-0.35
0.63**
0.03
0.17
-0.02*
0.00
-0.10
0.61**
0.07
0.02
0.00
0.00
0.13
-0.28
-0.02
-0.05
-0.33
0.91**
0.03
0.16
-0.01
0.00
0.09
0.17
0.03*
0.00
0.00
0.30
0.77***
-
-
Correction
- 1.03**
-0.02
-0.10
-0.35
0.93**
0.03
0.24
-0.02*
0.00
0.19
0.2 1
0.03**
0.00
0.00
-0.15
-0.42
-
0.74***
N
387
387
387
387
a Sipficance tests for h s vanable are one-mled
b 'llus lnterachon is meancentered. I e , the overall sample mean has been subtracted from any observahon Thus when thts term equals zero, the BIT
treatment vanable IS evaluated at the mean level of hbatlon
'
In the Bivanate Robit this term refers to the correlahon between the error terms (Rho). m the Negahve Bmomal models this term refers to the lnverje Mills
Raho
*p < 0 10. **p < 0 05, ***p < 0 01
-2LL
-221 97
-393.57
-406 94
-405.44
However, the differences were significantly larger for white than for African-American
sample members. The same results are found regardless of whether the selection bias correction
is omitted or included. The selection correction does not significantly improve the model fit to
these data; therefore, the conclusions and predicted probabilities are based on the models without
the selection correction.
27
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
B
Figure 4.2 converts the parameter estimates from the probit and negative binomial
regressions into predicted probabilities and predicted mean number of arrests in the twelve
months after sample entry. All predicted values were calculated by holding all variables, except
BTC participation and race, at their respective mean values. Figure 4.2 illustrates BTC’s main
and interaction effect on arrest. For both racial groups, BTC participation significantly reduced
the likelihood of being re-arrested; however, this effect is considerably stronger for Whites.
Likewise, while the “average” African-American in the BTC sample was predicted to have half
as many arrests as the “average” African-American in the comparison sample, this effect was
significantly smaller than the BTC’s effect on Whites.’
Figure 4.2 Predicted Values for Official Arrests in the Year after Sample Entry
Any Official Arrest
Mean Number of Arrests
0 93
W BTC
BTC
I
n
0 Corn parison
1 2 8
White***
65%
White***
0% 10% 20% 30% 40% 50% 60% 70%
0
.o
0.5
1 .o
1.5
2.0
‘p .1 “p
.05 “‘p
.01
Self-Reported Arrests
Table 4.4
(above) suggests that the criminal history record checks may not have detected
all of the arrests incurred by the BTC sample. If the official data did under-report the number of
arrests experienced by the BTC sample, then the above results from the analysis of official
arrests could be biased in the direct of finding a treatment effect, when in fact no such effect
existed. Therefore, as another measure of BTC’s effect on the likelihood of being arrested during
the follow-up time period, the following analysis assesses BTC’s effect on the number of self-
reported arrests. Unlike the above analysis of official arrests, which covered the 12 months after
sample entry, this analysis concerns arrests only in the six months prior to the follow-up
interview.
This analysis proceeds in the same manner as the analysis of official arrests. Both any
self-reported arrests and number of arrests are examined and for both outcomes, models with and
without the selection bias correction were estimated. Table 4.6 displays the results of these
models, which indicate that BTC significantly reduced the likelihood of any arrest for the BTC
sample, regardless of race. In fact, BTC lowered the risk of recidivism by more than half. In
It needs to be emphasized that In the current context the average offender is an imaginary person that has all of the average
characteristics of entire sample. See the mean values on the analysis variables are shown in Table A. I in Appendix C. These
thus estimates do not apply to any one person in the data set, but are approximations of BTC’s overall impact.
5
28
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
regards to number of arrests, however, BTC significantly reduced arrests only for White clients.
African-American BTC participants had fewer arrests than African-Americans in the pre-BTC
sample, but this difference was not significant.
Table 4.6
Self-Reported Arrests in the Six Months Prior to Follow-up Interview
Anv Arrests
Number of Arrests
Probit wlo
Probit wl
NegBin w/o
NegBin wl
Selection
Selection
Selection
Selection
Variable
Correction
Correction
Correction
Correction
BTC Treatmenta
-0.40**
-0.62**
-1.70***
-2.39***
Age
Female
Black
Black*BTC Tx
Education
Employ Bother
EmployBother*BTC Tx
Days Worked
Months in Jail
On Probation
Serious Offender
Prior Offenses
Prior Drug Use
Constant
Selection Correctionb
Overdispersion Parameter
-0.01
-0.05
0.24
-0.01
0.63***
-0.96**
0.01
0.00
0.48***
0.31*
0.00
0.01
- 1.09**
-
-0.01
-0.03
0.24
-0.01
0.59**
-0.92**
0.01
0.00
0.44**
0.29*
0.00
0.01
-0.92
0.17
-
-0.04**
0.02
-0.38
1.37**
0.97**
-1.65**
0.01
0.00
0.59*
0.40
0.00
0.02**
-0.05
-0.03
-
1.64***
-0.04**
0.09
-0.38
1.33**
-0.02
0.91*
-1.63**
0.02
0.00
0.47
0.33
0.00
0.02**
0.37
0.56
1.59***
N
382
382
382
382
-2LL
- 178.44
-337.07
-272.18
-270.96
Significance tests for this variable are one-tailed.
In the Bivariate Robit this term refers to the comeelation between the error terms (Rho); in the Negative Binomial models this term refers to the Inverse Mills
Ratio
*p < 0.10: **p < 0.05; ***p < 0.01
Self- Reported 0 ff enses
Arrests measure only detected offenses, so the analysis also tested the hypothesis that BTC
reduced criminal activity using data on self-reported criminal offenses collected in the follow-up
interviews. These self-report measures of criminality may be a more complete record of
offending, because official measures of crime record only those offenses detected by the
authorities. The majority of offenses go undetected, so these self-report measures provide
important additional insight into the effectiveness of BTC.
29
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Figure 4.3 Predicted Values for SelfReported Arrests in the Six Months Prior to Follow-up
Mean Number of Self-Reported Arrests
Any Self-Reported Arrest
I
I
Cornparison** I g'
126%
BTC
0% 10% 20% 30% 40% 50% 60% 70%
0.0
0.5
1 .o
1.5
2.0
I 'p
.1
"p .05 "'p .01
The survey instrument asked respondents to report the number of times they had
committed 14 types of offenses in the last 6 months: shoplifting or vandalism, parole or
probation violations, drug offenses, forgery, weapons offenses, burglary/larceny/brealung and
entering, robbery, assault, arson, rape, homicide or manslaughter, prostitution, contempt of court,
and any other offenses. Responses were used to create three summary measures of self-reported
crime, all covering the six months before follow-up interview: (1) the number of total offenses
committed, regardless of type of offense; (2) number of drug offenses (sales/distribution)
committed; (3) number of non-drug offenses (all types of offenses, except drug offenses). These
three summary measures were then used to create three dichotomous variables indicating any
recidivism, any drug recidivism, and any non-drug recidivism, respectively.
Table 4.7 below displays the means of each of the three measures of offending. These
results indicate that BTC participants self-reported significantly less recidivism of all types than
the pre-BTC sample, before controlling for other factors. In the analyses that follow, more
rigorous analytic techniques are employed to control for the sample differences.
Table 4.7 Self-Reported Offenses in the 6 Months Prior to Follow-up Interview
Offense Type
(n = 245)
(n = 137)
Differencea
Percent reporting any offense
21%
39%
0.001
Mean number of offenses
2.07
13.72
0.001
BTC sample
Pre-BTC sample
Significance of
Percent reporting any non-drug offense
15%
Mean number non-drug offenses
0.73
29%,
5.02
0.001
0.001
Percent reporting any drug offense
10%
23%
0.001
Mean number reporting drug offenses
I .34
8.69
0.00 1
These significance tests are all one-tailed. T-tests were used for the first five comparisons, and a chi-square test was employed for the last
comparison.
The analyses control for demographic factors, employment and offending histones. Two models
are shown for each dependent variable. The first omits the selection bias correction. The second
includes the selection bias correction described in Chapter 3. Probit models are used to estimate
30
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
t
b
BTC’s effect on the likelihood of any recidivism. The number of self-reported offenses is a count
variable and most respondents (73%) reported no offenses, so ordinary least squares regression is
inappropriate (see Long, 1997), instead negative binomial estimation is utilized.
Furthermore, because the number of self-reported offenses has a heavy positive skew
(i.e., most respondents reported no offenses but a small number of offenders reported substantial
involvement in crime) this variable was re-coded to censor the maximum number of offenses at
the 95‘h percentile (31 offenses). Thus, all offenders reporting more than 31 offenses were re-
coded to the censored maximum of 31.
The analysis of self-reported offenses is shown in Table 4.8. Similar to the results from
the arrest analyses, BTC’s effect on any self-reported offense interacts with race. White members
of the BTC sample were less likely to report criminal offending than white members of the pre-
BTC sample, but there was no significant difference between African-Americans in the two
samples. The negative binomial analyses indicate that BTC sample members did not report
significantly fewer offenses than members of the pre-BTC sample did.
Table 4.8
Selj-Reported Recidivism in the Six Months Prior to Follow-up
Any Self-Reuorted Offenses
Probit w/o
Probit
NegBin w/lo
NegBin w/
Selection
w/Selection
Selection
Selection
Variable
Correction
Correction
Correction
Correction
Number of Offenses Committed
BTC Treatmenta
-0.88 * * *
-0.97**
-0.44
-1.52*
Age
Female
Black
Black*BTC Tx
Education
Employ Bother
EmployBother*BTC Tx
Days Worked
Months in Jail
On Probation
Serious Offender
Prior Offenses
Prior Drug Use
Constant
Selection Correctionb
Overdispersion Parameter
0.00
-0.18
-0.30**
0.79*
0.36**
-0.90**
-0.02
0.00
0.26
0.08
0.00
0.01*
0.52
-0.07
-
0.00
-0.17
-0.30
0.79**
0.34
-0.89**
-0.01
0.00
0.24
0.07
0.00
0.01
0.59
0.07
-0.07
-0.01
-0.66
0.6 I
-
-0.33**
0.89
-2.96***
-0.05**
-0.01
0.80*
-0.32
0.01
0.00
4.29***
7.22***
-
-0.01
-0.52
0.49
-
-0.3 1 **
0.75
-2.93**
-0.04*
-0.01
0.56
-0.33
0.01
0.00
4.86
0.76
7.07***
N
382
382
382
382
-2LL
-402.44
-720.04
-490.1 1
-488.84
’ Significance tests for this variable are one-railed.
Inthe Bivariate Robit this term refers to the correlation between the error terms (Rho); in the Negative Binomial models this term refers to the Inverse Mills
Ratio.
*p < 0.10 **p < 0.05; ***p < 0 01
31
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Figure 4.4 illustrates the results of these analyses by showing the differences in probability of
reporting an offense holding all other variables at their means. There were no significant dlfferences
between the BTC and pre-BTC samples in the negative binomial analysis, so these results are omitted
from the figure.
Figure 4.4 Predicted Values for Self-Reported Offenses
Any Self-Reported
I
Offense
Black
White**
3 9%
Any Self-Reported
Non-Drug Offense
Black
BTC
UComparison
29%
White"
10%
16%
Any Self-Reported
Drug Offense'
0 %
10%
20%
30%
40%
50%
*pc .1
"pc .05 "'p
.01
Additional analyses were performed to determine if BTC participation affected the
likelihood of drug and non-drug recidivism because of the substantive interest in BTC's effect on
drug crimes. These two outcomes were also heavily skewed by the presence of a few high-rate
offenders. Both outcomes were re-coded censoring at the 95'h percentile (maximum = 15 for drug
offenses and 5 for non-drug offenses).
Table 4.8 displays the results from these regression analyses. These results indicate that
BTC had a marginally significant effect on drug offenses for both racial groups; however, BTC
did not significantly reduce the number of drug offenses committed. For non-drug offenses, BTC
reduced the likelihood of offending only for White BTC clients. Figure 4.4
(above) illustrates
these findings. The results from these three separate analyses of recidivism outcomes (official
arrest, self-report arrest, and self-report offenses committed) clearly indicate that BTC reduced
the likelihood of recidivism for White clients, regardless of data source and type of crime. The
consistent main effects of BTC support the conclusion that BTC reduced recidivism for African-
American clients. However, this reduction is smaller in magnitude than BTC's effect on White
clients and BTC did not appear to reduce the likelihood of non-drug offenses for African-
Americans.
3 2
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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 4.8
Self Reported Drug and Non-Drug Offenses
Any Drug Offenses
Number of Drug Offenses
Any Non-Drug Offenses
Number of Non-Drug
Offenses
Probit wlo
Probit
NegBin wlo
NegBin wl
Probit wlo
Probit
NegBin wlo
NegBin wl
Selection
w1Selection
Selection
Selection
Selection
w1Selection
Selection
Selection
Variable
Correction
Correction
Correction
Correction
Correction
Correction
Correction
Correction
BTC Treatment
-0.33*
-0.62*
-0.60
-2.35
-0.74**
-0.8
I *
0.20
-0.42
Age
Female
Black
Black*BTC Tx
Education
Employ Bother
EmployBother*BTC Tx
Days Worked
Months in Jail
On Probation
Serious Offender
Prior Offenses
Prior Drug Use
Constant
Selection Correction”
Overdispersion Parameter
-0.01
-0.20
0.2
I
-0.09*
-
0.5 1 **
- I .09**
-0.01
0.00
0.06
0.06
0.01*
0.01
0.08
-0.01
-0.17
0.2 I
-0.09*
0.44
-1.02**
-0.01
0.00
0.01
0.04
0.0 I
0.01
0.29
0.23
-
-
-0.01
-0.70
I .40
-
-0.4
I
-0.36
-
-0.05
-0.01
0.69
-0.49
0.02
0.00
4.88
14.77***
-
-0.02
-0.60
1.42
-
-0.4
I
-0.44
-0.02
-0.01
0.44
-0.47
0.0 1
0.01
5.92
1.37
14.16***
-
0.00
-0.17
-0.38
0.75**
0.32
-0.76*
-0.02**
0.00
0.42**
0.12
0.00
0.0 I
0.04
-0.05
-
-
0.00
-0.16
-0.38
0.75**
0.3
1
-0.75*
-0.02
0.00
0.4
1 **
0.1 1
0.00
0.01
0.10
0.05
-0.06
-
-0.01
-0.48
-0.09
-
-0.13
0.59
-2.01**
-0.04**
0.00
0.42
0.04
0.01
0.02
0.94
-
4.15***
-0.01
-0.45
-0.09
-
-0.13
0.57
-2.00**
-0.04**
0.00
0.39
0.02
0.01
0.02
I .04
4.14***
-
N
382
382
382
3 82
382
382
382
382
a Significance tests for this variable are one-tailed.
-2LL
-284.46
-601.33
-323.32
-32 1.50
-343.18
-660.82
-299.26
-299.2
I
In the Bivariate Probit this t e n refers to the correlation between the error tens (Rho); in the Negative Binomial models this t e n refers to the Inverse Mills Ratio.
‘p < 0.10; “p< 0.05; “‘p < 0.01
33
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
CHAPTER 5
b
THE IMPACT OF BREAKING THE CYCLE ON
EMPLOYMENT, FAMILY AND HEALTH PROBLEMS
Involvement in Brealung the Cycle was hypothesized to directly lead to reductions in
drug use and crimjnal activity, and these changes were hypothesized to lead to improvements in
social and economic well being. The BTC surveys included a number of Addiction Seventy
Index items designed to measure respondents’ perceptions of problems in a number of domains,
including health, social and family relationships, and employment. This section tests the
hypothesis that BTC participation led to reductions in problems in these areas by comparing
problems reported by the two samples in the thirty days before the follow-up interview,
controlling for sample differences and problems at baseline.
Table 5.1 compares the BTC and pre-BTC samples on health, social, and employment
problems in the 30 days before follow-up without controlling for sample differences. The results
indicate that BTC sample members were significantly less likely to report problems in the thirty
days before follow-up in every area except medical problems. The following sections test the
significance of these differences using multivariate models and controls for selection bias.
Table 5.1 Selfreported Problems in the 30 Days Prior to Follow-up Interview
BTC sample
Pre-BTC sample Significance
(n = 245)
(n = 137)
of Difference”
Health Problems
One or more days with medical problems in past
22%
21%
0.84
30 days
Number of Psychological Problems
0.32
0.54
0.03
past 30 days
One or more days with psychological problems in
18%
34%
One or more days with serious social conflicts in
9 8
20%
<0.01
SocialFamily Problems
past 30 days
Days paid for working in past 6 months
66.82
53.08
0.03
One or more days with employment problems in
22%
31%
0.04
past 30 days
‘These significance tests are all two-tailed. T-tests were used for the count variable. and chi-square tests were employed for comparisons
expressed in percentages
34
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Health Problems
Although BTC clients were more likely to report using medical services in the 30 days
before the follow-up interview than pre-BTC clients, they were not significantly less likely to
report medical problems on one or more days during the month (Table 5.2). They were also no
less likely to report having a psychological problem on one or more days in the month than the
comparison group. When asked about a list of specific symptoms of psychological distress, the
BTC sample did not report significantly fewer symptoms than the pre-BTC group during the past
30-days. The list of symptoms included serious depression, serious anxiety or tension,
hallucinations, trouble controlling violent behavior, serious thoughts of suicide, or attempted
suicide. These summary measures of perceptions of physical and mental health do not indicate
that improvements in these domains resulted from participation in BTC.
Table 5.2
Selj-Reported Health Problems in the 30-days before the Follow-up Interview
Medical Problems,
Total Psvchological
Any Psvchological
Past 30 Days
Problems,
Problems,
Past 6 Monthsa
Past 30 Daysb
Probit w/o
Probit w/Selection NegBin w/o NegBin w/
Probit w/o
Probit
Selection
Correction
Selection
Selection
Selection
w/Selection
Variable
Correction
Correction
Correction
Correction
Correction
BTC Treatment
-0.01
-0.02
0.04
0.17
-0.29
-0.15
Age
Female
Black
Education
Employ Bother
Days Worked
Months in Jail
Months Jail*BTC Tx
On Probation
Serious Offender
Prior Offenses
Prior Drug Use
Days in Jail
Prior Medical Measure
Constant
Selection Correction"
Overdispersion Parameter
N
0.02**
0.36*
-0.14
-0.04
-0.4 1 **
0.00
0.00
0.01**
0.07
0.01
0.00
-0.01
-0.03**
0.03***
-0.80*
382
0.02**
0.36
-0.14
-0.04
-0.4 1 *
0.00
0.00
0.01 **
0.07
0.01
0.00
-0.0 1
-0.03**
0.03***
-0.80
0.00
382
0.39*
0.03
-0.05
0.38
-0.01
-0.01
0.11
-0.02
0.00
0.01
-0.02
0.06**
0.32**
-0.19
-
-1.61
-
2.14***
0.03*
0.38
-0.18
-0.05
0.40
-0.01
-0.01
0.14
-0.02
0.00
0.01
-0.02
0.06**
0.33**
-
-1.73
-0.09
2.14***
0.02***
0.33*
-0.27*
-0.01
0.35*
0.00
-0.01**
0.02**
0.01
-0.17
0.00
0.00
0.01
0.02***
- 1.07**
-
0.02***
0.32
-0.26
-0.01
0.36"
0.00
-0.0 1 *
0.02*
0.03
-0.16
0.00
0.00
0.01
0.03**
-
-1.18*
-0.09
-2LL
- 176.3 1
-335.15
382
382
382
382
-287.00
-286.97
- 185.58
-344.34
In the Negative Binomial model this term refers to the Inverse Mills Ratio: in the Probit models this term refers to the comelation between the error t e r n
(Rho).
*p"This outcome is the sum of a series of questions asking respondents if they have experienced the following: serious depression; serious anxiety or tension:
experienced hallucinations: trouble controlling violent behavior. serious thoughts of suicide: or attempted suicide.
hRespondents were asked how many days they had experienced psychological problems in the last 30 days. 7his outcome was recoded as a dichotomy 10
minimize the positive skew: the higher category indicates having experienced at least one day of psychological problems.
terms (Rho).
*pIn the Negative Binomial model this term refers to the Inverse Mills Ratio: in the Bivariate Probit models this term refers to the correlation between the error
35
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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 Problems
b
i
The analysis similarly failed to find that BTC sample members were less likely to
experience an employment problem in the thirty days before follow-up and did not report
significantly more days of employment during the period (Table 5.3). The significant interaction
between group and days working in the thirty days before baseline interview is used as a control
variable to eliminate group differences in days of incarceration during the month before the
baseline and does not indicate a significant difference in BTC impact related to prior
employment.
Table 5.3
Self-Reported Employment Prior to Follow-up Interview
Days Paid for Working, in Past
OLS WIO
OLS
Probit wlo
Probit w/
Selection
wISelection
Selection
Selection
Anv Emdovment Problems in
6 Months
Past 30 Daw
Variable
Correction
Correction
Correction
Correction
BTC Treatmenta
-23.13***
-0.67
-0.20
-0.12
Age
Female
Black
Education
Education*BTC Tx
Employ Bother
Days Worked
Days Worked*BTC Tx
Months in Jail
Months in Jail*BTC Tx
On Probation
On Probation*BTC Tx
Serious Offender
Prior Offenses
Prior Drug Use
Days in Jail
Prior Work
Constant
Selection Correctiona
-0.20
6.60
-20.35***
-2.13
6.15**
-1.81*
11.07***
1.98***
0.01
6.95
2.97
0.04
-0. I3
-1.77***
0.40***
78.84***
-
-
-0.18
5.12
-20.02***
- 1.96
6.06**
13.22**
1.46**
0.05
11.31*
3.89
0.08
-0.13
-1.67***
0.41***
62.35**
-1.46**
-
-
-15.78**
-0.02**
-0.01
0.15
0.00
0.30
-0.02 * * *
-
-
-0.01
0.02***
-0.15
-
0.2 1
0.00
0.00
-0.01
0.00
0.20
-0.02**
-0.02
0.15
0.00
0.29
-0.02 * *
-0.0 1
0.02***
-0.13
0.2 1
0.00
0.00
-0.0 I
0.00
0.14
-0.06
-
-
-
N
383
382
382
382
R’I-2LL
0.36
0.36
- 194.14
-352.96
In the Ordinary Least Squares (OLS) regressions. this term refers to the Inverse Mills Ratio. In the Bivariate Probit regression. this term refers to the
correlation between the mor terms (Rho)
*p36
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.
Conflicts with Family and Others
BTC did have significant and substantial effects on social conflict, as the models in Table
5.3 show. Any days of conflict with family members in the thirty days before follow-up were
reported by 2% of the BTC sample and by 14% of the pre-BTC sample. Days of conflict with
others in the thirty days before follow-up were reported by 7% of the BTC sample and by 9% of
the pre-BTC sample. The average number of days of conflict for both groups was higher for
conflict with family members than for others, which is not surprising given the expected
frequency of contacts with family members. Thus, the results of the combined variable, shown
below, are largely made up of reduction in conflict with family members.
Table 5.3
Self-Reported SociaUFamily Conflict Outcomes
Davs ExDeriencine Serious Social
Anv Davs Experiencing Serious
Variable
Conflict, Past 30 Daysa
Conflict, Past 30 Daysb
NegBin w/o
NegBin w/
Probit w/o
Probit w/Selection
Selection
Selection
Selection
Correction
Correction
Correction
Correction
BTC Treatmenta
-2.15**
-3.78**
-0.69***
-1.23***
Age
Age*BTC Tx
Female
Black
Education
Employ Bother
Days Worked
Months in Jail
Months Jail*BTC Tx
On Probation
Serious Offender
Prior Offenses
Prior Drug Use
Days in Jail
Prior Conflicts Measure
Constant
Selection Correction'
Overdispersion Parameter
0.16*
-0.22**
0.98
-0.09
-0.2 1
0.62
0.02
0.00
0.15
-0.42
-0.03
-0.01
-0.03
0.04
-7.43
10.62***
-
-
0.14*
-0.21**
1.27
-0.05
-0.24
0.30
0.05
-0.01
-0.10
-0.34
-0.03
-0.01
-0.04
0.02
-0.95
1.17
10.1 I***
-
0.03**
0.52**
-0.13
-0.08
0.10
0.02**
-0.02*
0.03**
-0.02
0.00
0.00
0.00
-0.01
0.01
-0.9 1
-0.05**
-
0.03*
-0.05**
0.53**
-0.12
-0.08
0.00
0.03**
-0.02
0.03*
-0.12
0.01
-0.01
0.00
-0.01
0.00
-0.5 1
0.4 1
N
382
382
382
382
-2LL
-252.68
-25 I .45
- 125.64
-283.25
"This outcome is the sum of two questions asking respondents if how many days they had experienced serious conflicts with their family or with other
people, in the past 30 days.
hThis outcome is a dichotomous variable based on the preceding outcome: the higher category indicates having had experienced at least one day of social
conflicts.
'
In the Negative Binomial model this term refers to the lnverse Mills Ratio: in the Probit models this term refers to the correlation between the error terms
(Rho).
*pcO.IO **p<0.05; ***p37
This document is a research report submitted to the U.S. Department of Justice. This report has
not been published by the Department. 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.