Adult Drug Courts: Evidence Indicates Recidivism Reductions and  
Mixed Results for Other Outcomes (28-FEB-05, GAO-05-219).	 
                                                                 
Drug court programs, which were established in the late 1980s as 
a local response to increasing numbers of drug-related cases and 
expanding jail and prison populations, have become popular	 
nationwide in the criminal justice system. These programs are	 
designed to reduce defendants' repeated crime (that is, 	 
recidivism), and substance abuse behavior by engaging them in a  
judicially monitored substance abuse treatment. However,	 
determining whether drug court programs are effective at reducing
recidivism and substance use has been challenging because of a	 
large amount of weak empirical evidence. he 21st Century	 
Department of Justice Appropriations Authorization Act requires  
that GAO assess drug court program effectiveness. To meet this	 
mandate, GAO conducted a systematic review of drug court program 
research, from which it selected 27 evaluations of 39 adult drug 
court programs that met its criteria for, among other things,	 
methodological soundness. This report describes the results of	 
that review of published evaluations of adult drug court	 
programs, particularly relating to (1) recidivism outcomes, (2)  
substance use relapse, (3) program completion, and (4) the costs 
and benefits of drug court programs. DOJ reviewed a draft of this
report and had no comments. Office of National Drug Control	 
Policy reviewed a draft of this report and generally agreed with 
the findings.							 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-05-219 					        
    ACCNO:   A18317						        
  TITLE:     Adult Drug Courts: Evidence Indicates Recidivism	      
Reductions and Mixed Results for Other Outcomes 		 
     DATE:   02/28/2005 
  SUBJECT:   Alcohol or drug abuse problems			 
	     Courts (law)					 
	     Crime prevention					 
	     Crimes or offenses 				 
	     Criminals						 
	     Defendants 					 
	     Drug abuse 					 
	     Drug treatment					 
	     Drugs						 
	     Evaluation methods 				 
	     Locally administered programs			 
	     Performance measures				 
	     Program evaluation 				 
	     Recidivism 					 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-05-219

United States Government Accountability Office

GAO

                       Report to Congressional Committees

February 2005

ADULT DRUG COURTS

 Evidence Indicates Recidivism Reductions and Mixed Results for Other Outcomes

GAO-05-219

[IMG]

February 2005

ADULT DRUG COURTS

Evidence Indicates Recidivism Reductions and Mixed Results for Other Outcomes

  What GAO Found

Most of the adult drug court programs assessed in the evaluations GAO
reviewed led to recidivism reductions during periods of time that
generally corresponded to the length of the drug court program. GAO's
analysis of evaluations reporting these data for 23 programs showed the
following:

o  	Lower percentages of drug court program participants than comparison
group members were rearrested or reconvicted.

o  	Program participants had fewer recidivism events than comparison group
members.

o  	Recidivism reductions occurred for participants who had committed
different types of offenses.

o  	There was inconclusive evidence that specific drug court components,
such as the behavior of the judge or the amount of treatment received,
affected participants' recidivism while in the program.

Recidivism reductions also occurred for some period of time after
participants completed the drug court program in most of the programs
reporting these data.

Evidence about the effectiveness of adult drug court programs in reducing
participants' substance use relapse is limited to data available from
eight drug court programs. Evaluations of these eight drug court programs
reported mixed results on substance use relapse. For example, drug test
results generally showed significant reductions in use during
participation in the program, while self-reported results generally showed
no significant reductions in use.

Completion rates, which refer to the percentage of individuals who
successfully completed a program, in selected adult drug court programs
ranged from 27 to 66 percent. Other than participants' compliance with
drug court program procedures, no other program factor (such as the
severity of the sanction that would be invoked if participants failed to
complete the program) consistently predicted participants' program
completion.

A limited number of evaluations-four evaluations of seven adult drug court
programs-provided sufficient cost and benefit data to estimate their net
benefits. Although the cost of six of these programs was greater than the
costs to provide criminal justice services to the comparison group, all
seven programs yielded positive net benefits, primarily from reductions in
recidivism affecting judicial system costs and avoided costs to potential
victims. Financial cost savings for the criminal justice system (taking
into account recidivism reductions) were found in two of the seven
programs.

Figure 1: Overview of GAO Analysis of Drug Court Program Effectiveness

Source: GAO review of drug court program evaluations.

United States Government Accountability Office

Contents

  Letter 1

Background 3
Results 5
Concluding Observations 7
Agency Comments and Our Evaluation 8

Appendix I Objectives, Scope, and Methodology

Appendix II	Evaluations Used Multiple Approaches to Assess
Selection Bias

Comparison Groups Can Introduce the Problem of Selection Bias

Evaluations Addressed Selection Bias through Design

Evaluations Used a Variety of Statistical Methods to Minimize the
Possible Effects of Selection Bias

                                       9

                                       16

                                     16 17

                                       19

Appendix III GAO Criteria for Assessing Evaluations of Drug Court

Programs' Cost-Benefit Analyses 25

The Evaluations We Reviewed and the Net Benefit Measure 25
Criteria for Assessing a Drug Court Program's Cost-Benefit
Analysis 27
How the Drug Court Evaluations We Reviewed Used the Identified
Cost-Benefit Principles 32

Appendix IV Overview of Drug Court Program Characteristics and

Participants 36

Drug Court Program Components 36
Drug Court Program Participants 40

Appendix V Drug Court Programs Are Associated with Recidivism

Reductions 44

Drug Court Programs Led to Within-Program Recidivism
Reductions 44
Limited Evidence Indicates That Recidivism Reductions Endure 52

Appendix VI Limited and Mixed Evidence on Drug Court Programs'

Impacts on Substance Use Relapse 57

Limited Evidence on Drug Court Programs' Impact on Substance Use Relapse
57 Evidence on Drug Court Programs' Impact on Substance Use Relapse Is
Mixed 60

                                  Appendix VII

Evidence Is Mixed about the Factors That Predict Program Completion

Program Completion Rates Vary

Program Completion Is Associated with Compliance with Program Procedures

Effects of Drug Court Programs' Components on Program Completion Are Mixed

Participants Who Were Older and Had Less Criminal History than Other
Participants Were Most Likely to Complete Programs

                                     62 62

                                       64

                                       66

                                       69

Appendix VIII	Evidence from Four Evaluations Indicates Drug Court
Programs' Positive Net Benefits 71

Most Evaluations Found Drug Court Programs More Expensive than
Conventional Case Processing 71 Drug Court Programs' Net Benefits Derived
from Reductions in

Recidivism 72 All Seven Drug Court Programs Reported Positive Net Benefits
73 Net Benefits May Underestimate the Programs' True Benefits 74

       Appendix IX     GAO Contacts and Staff Acknowledgments              75 
                                       GAO Contacts                        75 
                                      Acknowledgments                      75 
      Bibliography                                                         76 

Tables

Table 1: Summary Information about Adult Drug Court Programs Included in
GAO's Review 11 Table 2: The Eight Drug Court Program Evaluations in Our
Cost-Benefit Review 14

Table 3: Research Designs, Comparison Groups, and Methods to Address
Selection Bias for Evaluations GAO Used in Recidivism Analysis 21

Table 4: The Eight Drug Court Program Evaluations in Our Cost-Benefit
Review 26 Table 5: Five Criteria for Assessing a Cost-Benefit Analysis of
a Drug Court Program 27 Table 6: Summary of Criteria Applied in Eight Drug
Court Program Evaluation Assessments 33 Table 7: Differences in Overall
Rearrest Rates between Drug Court Program Participants and Comparison
Group Members 46 Table 8: Differences in Reconviction Rates between Drug
Court Program Participants and Comparison Group Members 48

Table 9: Differences in Rearrest and Reconviction Rates, over Longer Time
Periods, between Drug Court Program Participants and Comparison Group
Members 53

Table 10: Post-Program Differences in Recidivism in Six New York State
Drug Court Programs 55 Table 11: Substance Use Relapse Results, by
Different Measures of Use, for Programs GAO Reviewed 58 Table 12:
Completion Rates for Selected Drug Court Programs Included in GAO's Review
63 Table 13: Reported Costs per Drug Court Program Participant in Four
Evaluations 72 Table 14: Net Benefits per Participant in Seven Drug Court
Programs 74

Abbreviations

APD antisocial personality disorder
GAO Government Accountability Office
ONDCP Office of National Drug Control Policy

This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

United States Government Accountability Office Washington, DC 20548

February 28, 2005

The Honorable Arlen Specter
Chairman
The Honorable Patrick J. Leahy
Ranking Minority Member
Committee on the Judiciary
United States Senate

The Honorable F. James Sensenbrenner, Jr.
Chairman
The Honorable John Conyers, Jr.
Ranking Minority Member
Committee on the Judiciary
House of Representatives

Drug court programs were established in the late 1980s as a local response
to increasing numbers of drug-related cases and expanding jail and prison
populations. These programs are designed to use a court's authority to
reduce crime by changing defendants' substance abuse behavior. Under
this concept, in exchange for the possibility of dismissed charges or
reduced sentences, defendants are diverted to drug court programs in
which they agree to participate in judicially monitored substance abuse
treatment. Title II of the 21st Century Department of Justice
Appropriations Authorization Act reauthorizes the award of federal grants
for drug court programs that include court-supervised drug treatment.1
The award of federal grants to drug court programs was first authorized
under Title V of the Violent Crime Control and Law Enforcement Act of
1994.2

Drug court programs have become popular nationwide in the criminal
justice system. As of September 2004, there were over 1,200 drug court
programs operating in addition to about 500 being planned. With such
expansion, it is important for policy and operational decision makers to
have definitive information about drug court programs' effectiveness in

1Pub. L. No. 107-273, 116 Stat. 1758, 1795 (2002). 2Pub. L. No. 103-322.

reducing recidivism and substance use relapse.3 In this respect, the 21st
Century Department of Justice Appropriations Authorization Act requires
that we study drug court program effectiveness.4 In response, this report
describes the results of published empirical evaluations of adult drug
court programs, particularly relating to (1) recidivism outcomes of
participants and other comparable offenders, (2) substance use relapse of
participants and other comparable offenders, (3) program completion of
participants, and (4) costs and benefits of drug court programs.

To address our objectives, we conducted a systematic review of drug court
program literature. We identified 117 evaluations of adult drug court
programs in the United States that were published between May 1997 and
January 2004 and reported recidivism, substance use relapse, or program
completion outcomes. Of these 117, we selected 27 evaluations that met
additional criteria for methodological soundness. Most of the evaluations
we selected used designs in which all drug court program participants were
compared with an appropriate group of similar offenders who did not
participate in the drug court program. In order to ensure that the groups
are similar in virtually all respects aside from the intervention, we
selected evaluations in which the comparison group was matched to the
program group as closely as possible on a number of characteristics or
that used statistical models to adjust, or control for, preexisting
differences between the program and comparison groups. Five of the 27
evaluations were experiments in which participants were randomly assigned
to groups. Random assignment works to ensure that the groups are
comparable.

The 27 evaluations we selected for our review reported information on 39
unique adult drug court programs. We systematically collected information
from these evaluations about the methodological characteristics of the
evaluations, the participants and components of the drug court programs,
the outcomes, and, where available, the costs of the drug court programs.
To assess the methodological strength of the evaluations, we used
generally accepted social science principles. Additionally, we used
standard cost-benefit criteria to screen and assess the eight evaluations
that reported cost and benefit information. Four of these contained
sufficient data on costs and benefits to allow us to assess

3We use the term recidivism to refer generally to the act of committing
new criminal offenses after having been arrested or convicted of a crime.

4Pub. L. No. 107-273, 116 Stat. 1758, 1799 (2002).

Background

net benefits. We selected the evaluations in our review according to the
strength of their methodologies; therefore, our results cannot be
generalized to all drug court programs or their evaluations. Finally, we
interviewed drug court researchers and officials at the Department of
Justice, the National Institute on Drug Abuse, and the Office of National
Drug Control Policy (ONDCP). We conducted our work from October 2003
through February 2005 in accordance with generally accepted government
auditing standards. Appendix I presents more details about our scope and
methodology, including a list of the evaluations we reviewed. Appendix II
describes the research designs and methods used in assessing recidivism in
the evaluations we reviewed, and appendix III presents our criteria for
assessing evaluations of drug court programs' cost-benefit analyses.

Of the 1,700 drug court programs operating or planned as of September
2004, about 1,040-nearly 770 operating and about 270 being planned- were
adult drug court programs, according to data collected by the Office of
Justice Programs' Drug Court Clearinghouse and Technical Assistance
Project.5 The primary purpose of these programs is to use a court's
authority to reduce crime by changing defendants' substance abuse
behavior. In exchange for the possibility of dismissed charges or reduced
sentences, eligible defendants who agree to participate are diverted to
drug court programs in various ways and at various stages in the judicial
process. These programs are typically offered to defendants as an
alternative to probation or short-term incarceration.

Drug court programs share several general characteristics but vary in
their specific policies and procedures because of, among other things,
differences in local jurisdictions and criminal justice system practices.
In general, judges preside over drug court proceedings, which are called
status hearings; monitor defendants' progress with mandatory drug testing;
and prescribe sanctions and rewards as appropriate in collaboration with
prosecutors, defense attorneys, treatment providers, and others. Drug
court programs also vary in terms of the substance abuse treatment
required. However, most programs offer a range of treatment options and
generally require a minimum of about 1 year of participation before a
defendant completes the program.

5The totals of adult drug court programs include tribal drug court
programs, but exclude family drug court programs.

In order to determine defendants' eligibility for participation, drug
court programs typically screen defendants based on their legal status and
substance use. The screening process and eligibility criteria can vary
across drug court programs.6 According to the literature, eligible drug
court program participants ranged from nonviolent offenders charged with
drug-related offenses who have substance addictions to relatively
mediumrisk defendants with fairly extensive criminal histories and failed
prior substance abuse treatment experiences. Participants were also
described as predominantly male with poor employment and educational
achievements. Appendix IV presents additional information about the
general characteristics of drug court programs and participants in the
evaluations we reviewed.

Research on drug court programs has generally focused on program
descriptions and process measures, such as program completion rates, and
presented limited empirical evidence about the effectiveness of drug court
programs in reducing recidivism and substance use. In 1997, we reported on
12 evaluations that met minimum research standards and concluded that the
evaluations showed some positive results but did not firmly establish
whether drug court programs were successful in reducing offender
recidivism and substance use relapse.7 More recently, two syntheses of
multiple drug court program evaluations have drawn positive conclusions
about the impact of drug court programs. One synthesis concluded that
criminal activity and substance use are reduced relative to other
comparable offenders while participants are engaged in the drug court
program, and that program completion rates ranged from 36 to 60 percent.8
Further, the other synthesis reported that drug offenders participating in
a drug court program are less likely to re-offend than

6Criteria for legal eligibility can include the current offense, criminal
justice histories, or supervision status (that is, whether the defendant
is currently on probation). Criteria related to substance use can include
frequency and type of current substance use, prior treatment experiences,
or motivation to seek substance abuse treatment.

7GAO, Drug Courts: Overview of Growth, Characteristics, and Results,
GAO/GGD-97-106 (Washington, D.C.: July 31, 1997).

8Belenko, S. Research on Drug Courts: A Critical Review 2001 Update. The
National Center on Addiction and Substance Abuse at Columbia University,
June 2001, p. 1. In his review, Belenko states that the reported program
completion rates are consistent with findings in other studies.

similar offenders sentenced to traditional correctional options, such as
probation.9

Some of the evaluations included in these two syntheses had methodological
limitations such as, the lack of strong comparison groups and the lack of
appropriate statistical controls. Some did not use designs that compared
all drug court program participants-including graduates, those still
active, and dropouts-with similar nonparticipants. For example, they
compared the outcomes of participants who completed the program with the
outcomes of those who did not (that is, dropouts). These evaluations, upon
finding that program graduates had better outcomes than dropouts, have
concluded that drug court programs are effective. This is a likely
overestimation of the positive effects of the intervention because the
evaluation is comparing successes to failures, rather than all
participants to nonparticipants. Additionally, other evaluations did not
use appropriate statistical methods to adjust for preexisting differences
between the program and comparison groups. Without these adjustments,
variations in measured outcomes for each group may be a function of the
preexisting differences between the groups, rather than the drug court
program.

In most of the evaluations we reviewed, adult drug court programs led to
recidivism reductions during periods of time that generally corresponded
to the length of the drug court program-that is, within-program.10 Our
analysis of evaluations reporting recidivism data for 23 programs showed
that lower percentages of drug court program participants than comparison
group members were rearrested or reconvicted. Program participants also
had fewer incidents of rearrests or reconvictions and a longer time until
rearrest or reconviction than comparison group members. These recidivism
reductions were observed for any felony offense and for drug offenses,
whether they were felonies or misdemeanors. However, we were unable to
find conclusive evidence that specific drug court program components, such
as the behavior of the judge, the amount of treatment received, the level
of supervision provided, and the sanctions for not complying with program
requirements, affect

9Wilson, D.B., O. Mitchell, and D.L. MacKenzie, A Systematic Review of
Drug Court Effects on Recidivism, (forthcoming), p. 20.

10We use within-program to refer to the period of time that all
individuals who are enrolled are expected to be participating in the drug
court program.

Results

participants' within-program recidivism. Post-program recidivism
reductions were measured for up to 1 year after participants completed the
drug court program in several evaluations, and in these the evidence
suggests that the recidivism differences observed during the program
endured. A more detailed description of the recidivism reduction results
is included in appendix V.

Evidence about the effectiveness of drug court programs in reducing
participants' substance use relapse is limited and mixed. The evidence
included in our review on substance use relapse outcomes is limited to
data available from eight drug court programs. The data include drug test
results and self-reported drug use; both measures were reported for some
programs. Drug test results generally showed significant reductions in use
during participation in the program, while self-reported results generally
showed no significant reductions in use. Appendix VI presents additional
information about the evaluations we reviewed that reported on substance
use relapse outcomes.

Completion rates, which refer to the number of individuals who
successfully completed a drug court program as a percentage of the total
number admitted, in the programs we reviewed that assessed completion
ranged from 27 to 66 percent. As might be expected, program completion was
associated with participants' compliance with program requirements.
Specifically, evaluations of 16 adult drug court programs that assessed
completion found that participants' compliance with procedures was
consistently associated with completion. These program procedures include
attending treatment sessions, engaging in treatment early in the program,
and appearing at status hearings. No other program factor, such as the
severity of the sanction that would be invoked if participants failed to
complete the program and the manner in which judges conducted status
hearings, predicted participants' program completion. Several
characteristics of the drug court program participants themselves were
also associated with an increased likelihood of program completion. These
characteristics include lower levels of prior involvement in the criminal
justice system and age, as older participants were more likely to complete
drug court programs than younger ones. Appendix VII presents additional
information about the evaluations we reviewed that reported on program
completion.

A limited number of evaluations in our review discussed the costs and
benefits of adult drug court programs. Four evaluations of seven drug
court programs provided sufficient cost and benefit data to estimate their
net benefits (that is, the benefits minus costs). The cost per drug court

Concluding Observations

program participant was greater than the cost per comparison group member
in six of these drug court programs. However, all seven programs yielded
positive net benefits, primarily from reductions in recidivism affecting
both judicial system costs and avoided costs to potential victims. Net
benefits ranged from about $1,000 per participant to about $15,000 in the
seven programs. These benefits may underestimate drug court programs' true
benefits because the evaluations did not include indirect benefits (such
as reduced medical costs of treated participants). Financial cost savings
for the criminal justice system (taking into account recidivism
reductions) were found in two of the seven programs. We provide additional
information about the reported costs and benefits of drug court programs
we reviewed in appendix VIII.

Overall, positive findings from relatively rigorous evaluations in
relation to recidivism, coupled with positive net benefit results, albeit
from fewer studies, indicate that drug court programs can be an effective
means to deal with some offenders. These programs appear to provide an
opportunity for some individuals to take advantage of a structured program
to help them reduce their criminal involvement and their substance abuse
problems, as well as potentially provide a benefit to society in general.

Although not representative of all drug court programs, our review of 27
relatively rigorous evaluations provides evidence that drug court programs
can reduce recidivism compared to criminal justice alternatives, such as
probation. These results are consistent with those of past reviews of drug
court evaluations. Positive results concerning recidivism are closely
associated with program completion. Specifically, while drug court
participation is generally associated with lower recidivism, the
recidivism of program completers is lower than for participants in
comparison or control groups. Thus, practices that encourage program
completion may enhance the success of drug court programs in relation to
recidivism.

While our review sheds little light on the specific aspects of these
programs that are linked to positive recidivism outcomes, both participant
compliance with drug court procedures and some participant characteristics
seem to be related to success. To the extent that research can help to
discern best practices for drug courts, the models for effective programs
can be enhanced. Specifically, to the extent that drug court program
managers can learn more about methods to retain participants for the
duration of the program, they may be able to further enhance the positive
impacts of drug court programs.

Agency Comments 	We requested comments on a draft of this report from the
Attorney General and the Director of ONDCP. Department of Justice
officials

and Our Evaluation 	informed us that the agency had no comments on the
report. ONDCP officials informed us that the agency generally concurred
with our findings.

We are sending copies of this report to other interested congressional
committees, the Attorney General, and the Director of the Office of
National Drug Control Policy. We will also make copies available to others
upon request. In addition, the report will be available at no charge on
GAO's Web site at http://www.gao.gov.

If you or your staff have any questions concerning this report, please
contact me at (202) 512-8777 or by e-mail at [email protected] or William
J. Sabol, Assistant Director, at (202) 512-3464, or [email protected]. Key
contributors to this report are listed in appendix IX.

Laurie E. Ekstrand, Director Homeland Security and Justice Issues

Appendix I: Objectives, Scope, and Methodology

The 21st Century Department of Justice Appropriations Authorization Act
requires that we assess drug court program effectiveness.1 Our objectives
were to assess the results of methodologically sound, published empirical
evaluations of adult drug court programs, particularly relating to (1)
recidivism outcomes of participants and other comparable offenders, (2)
substance use relapse of participants and other comparable offenders, (3)
program completion of participants, and (4) costs and benefits of drug
court programs.

To identify the universe of evaluations to include in our review, we used
a three-stage process. First, we (1) conducted key-word searches of
criminal justice and social science research databases;2 (2) searched drug
court program-related Web sites, such as those of the National Drug Court
Institute and the National Association of Drug Court Professionals; (3)
reviewed bibliographies, published summaries, meta-analyses, and prior GAO
reports on drug court programs;3 and (4) asked drug court researchers and
officials in agencies that fund drug court research to identify
evaluations. Our literary search identified over 230 documents, which
consisted of published and unpublished outcome evaluations, process
evaluations that described program objectives and operations, manuals and
guides related to drug court program operations, commentary from drug
court practitioners, and summaries of multiple program evaluations. Next,
we reviewed these documents and identified 117 evaluations of adult drug
court programs in the United States that (1) were published between May
1997 and January 20044 and (2) reported recidivism, substance use relapse,
or program completion outcomes.

Finally, to select the evaluations we used in our in-depth review, we
screened them to determine whether they met additional criteria for

1Pub. L. No. 107-273, 116 Stat. 1758, 1799 (2002).

2We searched the ERIC, Biosis Previews, Social Scisearch, National
Criminal Justice Reference Service, Dissertation Abstracts Online, Gale
Group Magazine Database, Wilson Social Science Abstracts, Gale Group Trade
& Industry Database, Gale Group Health & Wellness Database, Gale Group
Legal Resources Index, and Periodicals Abstracts PlusText.

3GAO, Drug Courts: Information on a New Approach to Address Drug-Related
Crime, GAO/GGD-95-159BR (Washington, D.C.: May 22, 1995); GAO, Drug
Courts: Overview of Growth, Characteristics, and Results, GAO/GGD-97-106
(Washington, D.C.: July 31, 1997); and GAO, Drug Courts: Better DOJ Data
Collection and Evaluation Efforts Needed to Measure Impact of Drug Court
Programs, GAO-02-434 (Washington, D.C.: April 18, 2002).

4In our last report that assessed the published research, GAO/GGD-97-106,
we completed our review in May 1997.

Appendix I: Objectives, Scope, and Methodology

methodological soundness. Specifically, to assess recidivism and substance
use relapse, we selected evaluations that used either an experimental
design in which (1) eligible offenders were randomly assigned to different
programs or conditions and (2) there was an acceptable level of attrition5
or a quasi-experimental design in which (1) all drug court program
participants were compared with an appropriate group of comparable
offenders who did not participate in the drug court program, and (2)
appropriate statistical methods were used to adjust, or control, for group
differences. If random assignment was not used, in an attempt to ensure
that the groups were similar, aside from program participation (the
intervention), the comparison group(s) should have been as alike as
possible on a range of important characteristics. Statistical analyses can
be used to further minimize differences between the program and comparison
groups. Typically, statistical analyses to control for differences such as
these are not necessary when study participants are randomly assigned to
groups.

To assess program completion, we also selected evaluations that compared
the outcomes of participants (such as program graduates and those who
dropped out) within a drug court program in order to determine what
factors, if any, are associated with program completion. We selected those
evaluations that used appropriate statistical methods to control for
differences between the participant groups. Of the 117 evaluations we
screened, we selected 27 evaluations for our in-depth review.

The 27 evaluations we selected for our review reported information on 39
unique adult drug court programs that were implemented between 1991 and
1999. Table 1 lists the drug court program evaluated, researchers, and
outcomes we used in our assessment of drug court program effectiveness.
All of the evaluations we reviewed, as well as others consulted, is
included in the bibliography.

5Over time, participants in an evaluation can drop out for various
reasons. Acceptable levels of attrition can vary depending on other
characteristics of the evaluation design, such as the sample size, the
time frames covered, and statistical methods used.

Appendix I: Objectives, Scope, and Methodology

 Table 1: Summary Information about Adult Drug Court Programs Included in GAO's
                     Review Outcomes used in our assessment

                                                                      Program 
Drug court program evaluated (researcher)  Recidivism  Drug use completion 
Bakersfield (Calif.) Municipal Drug Court                          Yes     
          (Anspach and Ferguson, 2003)                             

  Baltimore City (Md.) Drug Treatment Court (Banks and Gottfredson, 2003) Yes

(Gottfredson and Exum, 2002; Gottfredson, Najaka, and Yes Kearley, 2003)

Breaking the Cycle Program, a Birmingham, Ala.         Yes    Yes   
(Harrell and others, 2003)                                          
Breaking the Cycle Program, a Jacksonville, Fla.       Yes    Yes   
(Harrell and others, 2003)                                          
Breaking the Cycle Program, a Tacoma, Wash             Yes    Yes   
(Harrell and others, 2003)                                          
Bronx (N.Y.) Treatment Court                           Yes             Yes 
(Rempel and others, 2003)                                           

        Brooklyn (N.Y.) Treatment Court (Rempel and DeStefano, 2001) Yes

 (Rempel and others, 2003) Yes Yes Broward County (Fla.) Drug Court (Schiff and
                                Terry, 1997) Yes

                           (Senjo and Leip, 2001) Yes

Chester County (Penn.) Drug Court Program Yes Yes (Brewster, 2001)

Clark County Drug Court (Las Vegas, Nev.)

           (Goldkamp, White, and Robinson, 2001)                          Yes 
               (Miethe, Lu, and Reese, 2000)              Yes          
              Creek County (Okla.) Drug Court                             Yes 
               (Anspach and Ferguson, 2003)                            
       D.C. Superior Court Drug Intervention Program      Yes    Yes   
           (Harrell, Cavanagh, and Roman, 1998)                        
             Douglas County (Neb.) Drug Court             Yes          
    (Spohn and others, 2001; Martin, Spohn, Piper, and                 
                   Frenzel-Davis, 2001)                                
             Escambia County (Fla.) Drug Court                         

                         (Peters and Murrin, 2000) Yes

                      (Peters, Haas, and Murrin, 1999) Yes

                       (Truitt and others, 2002) Yes Yes

                 Appendix I: Objectives, Scope, and Methodology

                        Outcomes used in our assessment

                                                                      Program 
    Drug court program evaluated (researcher)  Recidivism Drug use completion 
     Hamilton County Drug Court (Cincinnati,      Yesb             
                      Ohio)                                        
           (Listwan and others, 2003)                              

        Jackson County (Mo.) Drug Court (Anspach and Ferguson, 2003) Yes

(Truitt and others, 2002)                                  Yes         Yes 
cKentucky Drug Court Programs                              Yes        
(Logan, Hoyt, and Leukefeld, 2001)                                    
Las Cruces (N. Mex.) Municipal DWI Drug Court              Yesd       
(Breckenridge and others, 2000)                                       
Los Angeles County (Calif.) Drug Court Program             Yes        
(Deschenes and others, 1999; Fielding and others, 2002)               
Maricopa County (Ariz.) First Time Drug Offender Program   Yes   Yes  
(Deschenes, and others, 1996; Turner and others, 1999)                

Multnomah County (Ore.) Sanctions, Treatment, Opportunity, and Progress
(STOP) Drug Diversion Program

         (Finigan, 1998) Yes (Goldkamp, White, and Robinson, 2001) Yes

New Castle County (Del.) Drug Court                             Yes    Yes 
(Marlowe and others, 2004; Festinger and others, 2002)               
North Carolina Drug Treatment Court                      Yes         
(Craddock, 2002)                                                     
Okaloosa County (Fla.) Drug Court                                      Yes 
(Peters and Murrin, 2000)                                            
Orange County (Calif.) Drug Court Program                Yes           Yes 
(Deschenes and others, 2001)                                         
Queens (N.Y.) Treatment Court                            Yes           Yes 
(Rempel and others, 2003)                                            

Rochester (N.Y.) Drug Treatment Court Yes (Rempel and others, 2003)

        So. San Mateo County (Calif.) Drug Court          Yes     
          (Wolfe, Guydish, and Termondt, 2002)                    
            St. Mary Parish (La.) Drug Court                              Yes 
              (Anspach and Ferguson, 2003)                        
       Suffolk County (N.Y.) Drug Treatment Court         Yes             Yes 
               (Rempel and others, 2003)                          
       Syracuse (N.Y.) Community Treatment Court          Yes             Yes 
               (Rempel and others, 2003)                          
          Washington State Drug Court Programe            Yes     
                (Barnoski and Aos, 2003)                          

Source: GAO analysis of drug court program evaluations.

Appendix I: Objectives, Scope, and Methodology

aAccording to the principal researcher of the evaluation, the Breaking the
Cycle program is technically not considered to be a drug court program.
However, we included the program in our review because it incorporates
several of the basic drug court program components, such as drug testing
and access to treatment services.

bRecidivism results for the Hamilton County evaluation were not included
in our review because upon our further review, we found that the
nonequivalent control group design did not sufficiently control for group
differences that could have explained the reported (favorable) recidivism
differences.

cThe evaluation of the Kentucky Drug Court Program included three
programs: (1) Fayette Drug Court program, (2) Jefferson County Drug Court
program, and (3) Warren Drug Court program. The data for these three
programs were combined in the analyses conducted. However, information
about the program components and participants was reported separately for
each program.

dWe included the evaluation of the Las Cruces Municipal DWI Drug Court
because it was an experiment that assessed recidivism. However, we did not
include its results in our final review of recidivism outcomes because it
is a specialty court that handles only misdemeanor driving while
intoxicated (DWI) defendants.

eThe evaluation of Washington State Drug Court Program included six
programs: those in (1) King, (2) Kitsap, (3) Pierce, (4) Skagit, (5)
Spokane, and (6) Thurston counties. Limited information about the program
components and participants were reported for each of the six programs.
However, some of the recidivism analyses combined data from multiple
counties.

To obtain information on our outcomes of interest-that is, recidivism,
substance use relapse, and program completion-we used a data collection
instrument to systematically collect information about the methodological
characteristics of the evaluations, the participants and components of the
drug court programs, and the outcomes of the participants and other
comparable groups.

To assess the methodological strength of the 27 evaluations, we used
generally accepted social science principles.6 For example, we assessed
elements such as whether data were collected during or after program
completion and the appropriateness of outcome measures, statistical
analyses, and any reported results. Each evaluation was read and coded by
a senior social scientist with training and experience in evaluation
research methods. A second senior social scientist and other members of
our evaluation team then reviewed each completed data collection
instrument to verify the accuracy of the information included. Part of our
assessment also focused on the quality of the data used in the evaluations
as reported by the researchers and our observations of any problems with
missing data, any limitations of data sources for the purposes for which

6Social science research standards are discussed in the scientific
literature. For example, see Campbell, Donald T., and Julian Stanley,
Experimental and Quasi-Experimental Designs for Research (Chicago: Rand
McNally, 1963); Cook, Thomas D., and Donald T. Campbell,
Quasi-experimentation: Design and Analysis Issues for Field Settings
(Boston: Houghton Mifflin, 1990); and GAO, Designing Evaluations,
GAO/PEMD-10.1.4 (Washington, D.C.: May 1991).

Appendix I: Objectives, Scope, and Methodology

they were used, and inconsistencies in reporting data; we incorporated any
data problems noted into our appraisals.

To assess the cost-benefit analyses of drug court program evaluations, we
reviewed all of the evaluations selected for our structured review that
reported cost or benefit information. Of the 27 evaluations included in
our in-depth review, 8 reported information about program costs and 4
about benefits. The 8 evaluations we included in our cost review are shown
in table 2.

Table 2: The Eight Drug Court Program Evaluations in Our Cost-Benefit
Review

Reported data on Drug court program evaluation Benefits Costs Included in
          cost-benefit analysis Not included in cost-benefit analysis

                                   Breaking the Cycle Program   Yes       Yes 
                                   (Harrell and others, 2003)         
                D.C. Superior Court Drug Intervention Program   Yes       Yes 
                          Harrell, Cavanagh, and Roman, 1998)         
                  Multnomah County (Ore.) STOP Drug Diversion   Yes       Yes 
                            Program (Carey and Finigan, 2003)         
                          Washington State Drug Court Program   Yes       Yes 
                                     (Barnoski and Aos, 2003)         

                                Douglas County (Neb.) Drug Court  No      Yes 
                 (Martin, Spohn, Piper, and Frenzel-Davis, 2001)       
                                    Kentucky Drug Court Programs  No      Yes 
                              (Logan, Hoyt, and Leukefeld, 2001)       
                  Los Angeles County (Calif.) Drug Court Program  No      Yes 
                                    (Deschenes and others, 1999)       
                Maricopa County (Ariz.) First Time Drug Offender  No      Yes 
                            Program (Deschenes and others, 1996)       

Source: GAO analysis of drug court program evaluations.

Four of the 8 evaluations reported sufficient data on both benefits and
costs, which allowed us to assess the reported net benefits of the drug
court programs. Specifically, we were able to determine whether the
reduction of recidivism-the benefit-would outweigh the additional costs of
a program. We used standard cost-benefit criteria to screen and assess
these evaluations that reported cost and benefit information.
Additionally, we reviewed methodologies describing approaches for
conducting cost analyses of drug court programs.

Appendix I: Objectives, Scope, and Methodology

We selected the evaluations in our review based on their methodological
strength; therefore, our results cannot be generalized to all drug court
programs or their evaluations. Although the findings of the evaluations we
reviewed are not representative of the findings of all evaluations of drug
court programs, the evaluations consist of those published evaluations we
could identify that used the strongest designs to assess drug court
program effectiveness.

Finally, we interviewed drug court program researchers and officials at
the Department of Justice, the National Institute on Drug Abuse, and the
Office of National Drug Control Policy. We conducted our work from October
2003 through February 2005 in accordance with generally accepted
government auditing standards.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

Comparison Groups Can Introduce the Problem of Selection Bias

Most of the evaluations included in our review used quasi-experimental
comparison groups.1 The use of quasi-experimental comparison groups can
result in comparisons between drug court participants and comparison group
members that differ on key variables related to recidivism. Systematic
difference between comparison groups and treatment groups, called
selection bias, threatens the validity of evaluation findings. Though
design and analysis strategies differed, each of the quasiexperimental
evaluations in our review used some combination of design and statistical
methods to address the issue of selection bias. Approaches to address
selection bias in design generally included attempts to choose comparison
groups that share key commonalities with treatment groups (usually
including likelihood of eligibility to participate in a drug court were it
available) and therefore were less likely to differ on observed and
unobserved characteristics related to recidivism. The evaluations in our
review also included attempts to minimize the effects of selection bias
through the application of various statistical methods.

The observed differences in recidivism that we discuss in this report
could arise from measured and unmeasured sources of variation between drug
court participants and comparison group members. If comparison group
members differed systematically from drug court participants on variables
or factors that are also associated with recidivism and these variables
were not accounted for by the design or analysis used in the evaluation,
then the observed differences in recidivism could be due to these sources
of variation rather than participation in the drug court program. For
example, if successful drug court participants differed systematically in
their substance abuse addiction problems from the members of the group
against which they are compared, and if the differences in substance abuse
addiction are not explicitly assessed in the evaluation, these
differences, and not necessarily participation in the drug court program,
could explain any observed recidivism differences. Similarly, if
participants differed from comparison group members in their motivation to
complete the drug court, recidivism differences could arise from these.
Evaluations generally do not have measures of variables, such as
motivation, that can be explicitly included in the analysis of drug court
outcomes.

1In contrast to experimental control groups, where participants are
deliberately randomly assigned to either treatment or control groups for
purposes of comparison, the comparison groups in quasi-experimental groups
are constructed from relatively comparable existing groups not receiving
treatment. Quasi-experimental methods are often used to assess effects in
the natural setting, where a true experiment may not be feasible.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

Evaluations Addressed Selection Bias through Design

One way to address issues of selection bias is in the design of the
comparison groups. Generally, random assignment of eligible participants
to treatment and control groups addresses selection bias by randomly
distributing individual differences between the treatment and control
group. Another way to address selection bias is statistically-by forming
comparison groups that consist of individuals that are as similar as
possible to drug court participants and by using statistical techniques to
control for observed differences, including sophisticated two-step
procedures that attempt to address differences in selection into the drug
court program. For the recidivism outcomes we reported, the evaluations in
our review used design and statistical methods to address selection bias.

Evaluations in our review were either experiments or quasi-experiments.
The experiments randomly assigned eligible defendants to a drug court
program or a control group of defendants who received conventional case
processing.2 The quasi-experiments used one of two types of comparison
groups:

o  	historical comparison group-formed from individuals who received
conventional case processing during a period of time shortly before the
drug court program was implemented and

o  	contemporaneous comparison group-formed from defendants (1) who were
eligible for drug court but received conventional case processing during
the same time period as the drug court program participants, (2) who were
from a district within a court's jurisdiction from which arrestees were
not eligible to participate in the drug court program, or (3) who had
similar charges and were matched on characteristics.

If implemented as designed, experiments can provide strong evidence for
the effectiveness (or lack thereof) of a drug court intervention. A key
presumption of random assignment is that any individual factor that could
be associated with the outcome of interest (recidivism) is randomly
distributed between the experimental and control groups. Hence, if none of
these factors are correlated systematically with assignment into groups,
then the threat of selection bias is reduced. Presuming that the
randomization is complete and that the experiment was carried out

2We use conventional case processing in this report to refer to the
general set of criminal justice procedures that comparison group members
receive.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

without high levels of attrition of subjects from either group, the
observed differences in recidivism between the two groups would likely
arise from the intervention, rather than observed or unobserved individual
factors.

Several of the evaluations used quasi-experimental designs in which the
comparison group members were chosen from defendants who appeared before
the court during a period of time shortly before the drug court program
was introduced and, based on certain observable characteristics, were
deemed to be eligible to participate in the drug court program.
Theoretically, if there were no significant changes in processes that led
to defendants appearing in a court during the pre-and drug court periods,
and if their eligibility for drug court participation could be determined,
a historical comparison group would consist of individuals that share
characteristics with drug court defendants. Alternatively, if there were a
significant difference in processing between the pre-drug court and drug
court periods (for example, such as a shift in prosecution priorities
toward particular types of offenses or a change in the nature of a
community's drug problem), then the historical comparison group could
consist of defendants that differed systematically from the drug court
participant group.

The historical comparison group in the evaluations we reviewed generally
consisted of defendants who were arrested and arraigned in the court that
offered the drug court program in a period immediately prior to the
implementation of the drug court. In one case, the comparison group
members were chosen from defendants on probation in the period prior to
the drug court. To minimize possible bias that could arise from changes in
court practices or the composition of defendants entering a court in the
period prior to the implementation of a drug court, in the evaluations we
reviewed, the prior periods generally ended within 4 months to 1 year
before the implementation of the drug court program.

Another subset of the evaluations we included in our review used
quasiexperiments with contemporaneous comparison groups. One comparison
group consisted of defendants who met the drug court program's eligibility
requirements but who were arrested in areas within a court's jurisdiction
that did not enroll individuals in the drug court. Another comparison
group was formed by selecting members from other, comparable
jurisdictions. A third consisted of individuals who, while otherwise
eligible to enroll in the drug court, were unable to participate because
of logistical reasons. Others consisted of defendants charged with similar
offenses as drug court participants and who were then matched on key
variables.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

Evaluations Used a Variety of Statistical Methods to Minimize the Possible
Effects of Selection Bias

All of the quasi-experiments that we reviewed, whether they used
historical or contemporaneous comparison groups, used various statistical
methods to control for individual differences between drug court
participants and comparison group members on observable variables. Several
evaluations used sophisticated statistical methods that not only adjusted
results for individual level differences but also attempted to correct for
selection bias by modeling unobserved differences between drug court and
comparison group members.3 These methods reduce but do not completely
eliminate potential bias from unobserved confounding factors. The extent
to which they reduce bias depends upon the richness and quality of the
control variables that are used to estimate the models.

These methods essentially use statistical models to predict individual
probabilities of participation in the drug court program (regardless of
actual participation). The models predict the probability, sometimes
called a propensity score, based on variation in individual level
characteristics (for example, criminal histories and demographic
attributes) that offenders in a given group would participate. Some of the
evaluations that we reviewed incorporated this technique as the first
stage of a two-stage model that estimated recidivism differences. Others
used it to identify the individuals to include in a comparison group.

Other studies matched comparison group members to drug court participants
on a more limited set of demographic and criminal justice variables (such
as type of charge and prior criminal history) and then, post-matching,
used various methods to control for differences on unmatched variables. A
presumption behind these matching methods is that by selecting for
inclusion in the comparison group only those defendants who matched drug
court participants on these observed characteristics, the evaluation would
create a comparison group that was similar in composition to the drug
court participants. A limited set of matching variables that is based
largely upon demographic variables may be unlikely to capture all of the
individual-level sources of variation in recidivism that could account for
differences between the drug court participants and comparison group
members. Therefore, these evaluations attempted to control for differences
in key, nonmatch variables (such as criminal justice risk-level
differences) in their analysis.

3See, for example, Heckman, James, Hidehiko Ichimura, and Petra Todd
(1997). "Matching as an Econometric Evaluation Estimator: Evidence from a
Job Training Programme," Review of Economic Studies, Vol. 64, No. 4, pp.
605-654.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

Some of the evaluations attempted to control for individual-level
differences between the drug court participants and the comparison group
members using various linear and nonlinear (e.g., logistic) regression
methods. This approach relies on the assumption that the observed
characteristics included in the regression are the key variables that
predict recidivism. The regression models adjust the results for
differences between the two samples on these characteristics and allow the
researcher to avoid making incorrect inferences about recidivism
differences because one group (either treatment or comparison) differs
systematically from the other in the presence of the key variables that
are associated with recidivism. Evaluations that used regression methods
generally made efforts to control for variables that are known to be
related to recidivism, such as criminal history, type and number of
charges, and age. Table 3 describes the comparison groups used in the
evaluations we reviewed and the methods used by these evaluations to
address selection bias.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

Table 3: Research Designs, Comparison Groups, and Methods to Address
Selection Bias for Evaluations GAO Used in Recidivism Analysis

Drug court program and design Comparison group used in evaluation Methods
to address selection bias

                              Experimental design

Baltimore City Drug Treatment Court 	Randomly assigned from the initial
pool of Random assignment to control prior to willing, eligible defendants
to a control group appearance before a drug court judge, who that received
"treatment as usual," which made final determinations for drug court
entry. included less intensive drug testing, probation, judicial
monitoring, and treatment.

    D.C. Superior Court  Randomly assigned from the     Random assignment and 
           Drug          initial pool of                   various regression 
Intervention Program          eligible drug court   methods to control for 
                          participants to a standard   individual differences 
                              (control) docket.      between drug court       
                            Participants received    participants and control 
                         services such as            group members, such as   
                         supervision, drug tests     age, gender, prior       
                            (twice weekly), judicial    criminal history,     
                                monitoring, and were  severity of drug use,   
                                                               and            
                         encouraged to seek                                   
                         community treatment.           employment status.

Maricopa County First Time Drug Offender Program

All individuals identified as eligible for the drug court program were
randomly assigned to one of three study groups at the time of their
initial probation assignment. The three study groups were:

1. probation with no testing;

2. probation with low-rate (monthly) drug testing; and

3. probation with high-rate (biweekly) drug testing.

                               Random assignment.

Quasi-experimental, historical comparison group

Breaking the Cycle Program, Historical arrestees: sample drawn from
Regression methods that control for individual differences and a two-stage
estimation procedure to account for unmeasured sample differences.
Birmingham arrestees in a drug use forecasting project in the Birmingham
jail over a 2-month period in the year prior to the start of the treatment
court. Arrestees were chosen using the same criteria to select treatment
court participants: age, county, felony charges, and involvement with
illegal drugs.

      Breaking the Cycle Program, Historical arrestees: sample drawn from

Regression methods that control for individual differences and a two-stage
estimation procedure to account for unmeasured sample differences.

Jacksonville arrestees in the Duval County jail over a 4month period that
ended 5 months prior to the start of the treatment court. Arrestees were
chosen using the same criteria to select treatment court participants:
age, county of residence (Duval), felony charges, and involvement with
illegal drugs.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

Drug court program and design Comparison group used in evaluation Methods
to address selection bias

Breaking the Cycle Program, Historical arrestees: sample drawn from
Regression methods that control for individual

Tacoma	arrestees in the Pierce County jail over a 4-differences and a
two-stage estimation month period that ended 5 months prior to procedure
to account for unmeasured sample the start of the treatment court.
Arrestees differences. were chosen using the same criteria to select
treatment court participants: age, county of residence (Pierce), felony
charges, and involvement with illegal drugs.

Bronx Treatment Court 	Historical arrestees: defendants arrested in the
Bronx in the 4 months prior to the opening of the drug court on selected
felony charges of criminal sale of a controlled substance.

Chester County Drug Court Program	Historical probationers: defendants
placed on probation during the 10-month period prior to the inception of
the drug court program who would have been eligible had the drug court
program existed (i.e., charged with nonmandatory drug offenses, had no
prior violent offenses, and were not on probation or parole).

Escambia County Drug Court 	Historical arrestees: defendants arrested and
arraigned in Escambia County over a 27month period that ended in the month
prior to the beginning of the drug court. Selected on the same type of
charges as drug court participants.

Propensity score matching methodology (to identify comparison group
members having similar predicted probabilities of participating in the
drug court as drug court participants); match model based on age, race,
specific charges, and various criminal history measures. Comparison group
members matched to the drug court participant having the nearest
propensity score.

Comparison group was similar to the drug court sample in demographic
characteristics (age, race, and gender), prior record, and prior drug
treatment characteristics. Author and county officials were unaware of
historical events that would have caused threats to validity of
comparison.

Instrumental variable approach that controls for the probability of
participation in the drug court program and models unobserved motivational
differences between participants and nonparticipants.

Jackson County Drug Court 	Historical arrestees: defendants arrested in
Instrumental variable approach that controls for Jackson County over a
3-year period that the probability of participation in the drug court
ended in the 15 months prior to the beginning program and models
unobserved motivational of the drug court. Selected on the same type
differences between participants and of charges as drug court
participants. nonparticipants.

Queens Treatment Court 	Historical arrestees: defendants arrested in
Propensity score matching methodology (to Queens in the year prior to the
opening of the identify comparison group members having drug court on a
most serious charge of similar predicted probabilities of participating in
criminal sale or possession of drugs (that met the drug court as drug
court participants); match charge eligibility for drug court). Excluded
model based on age, race, specific charges, and defendants who did not
meet the drug court's various criminal history measures. Comparison paper
eligibility criteria, i.e., current and prior group members matched to the
drug court violence and those cases in which the arrest participant having
the nearest propensity score. did not lead to a conviction.

So. San Mateo County Drug Court 	Historical arrestees: defendants arrested
and Regression methods to control for individual processed before the
start of the drug court differences on key variables, such as ethnicity,
but who would have been eligible for the drug prior convictions, primary
language, and severity court. of charges.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

Drug court program and design Comparison group used in evaluation Methods
to address selection bias

Suffolk County Drug Historical arrestees:      Propensity score matching   
        Treatment      defendants arrested in          methodology (to        
          Court          Suffolk County in the    identify comparison group   
                           year prior to the            members having        
                       opening of the drug court similar predicted            
                       on a most serious         probabilities of             
                                                 participating in             
                       charge of felony or       the drug court as drug court 
                       misdemeanor criminal              participants); match 
                       possession of a              model based on age, race, 
                       controlled substance             specific charges, and 
                       (that                     
                       met eligibility for drug  various criminal history     
                           court). Excluded      measures. Comparison         
                       defendants with current   group members matched to the 
                       violent charges or                 drug court          
                       prior violent convictions participant having the       
                       and those cases in        nearest propensity score.    
                        which the arrest did not 
                           lead to a conviction. 

Syracuse Community Treatment Historical arrestees: defendants arrested in
Propensity score matching methodology (to

Court 	Syracuse City in the year prior to the opening identify comparison
group members having of the drug court on a most serious charge of similar
predicted probabilities of participating in felony or misdemeanor criminal
possession the drug court as drug court participants); match of a
controlled substance (that met drug court model based on age, race,
specific charges, and charge eligibility). Excluded defendants with
various criminal history measures. Comparison prior violent convictions or
pending violent group members matched to the drug court charges and those
cases in which the arrest participant having the nearest propensity score.
did not lead to a conviction.

Quasi-experimental, contemporaneous comparison group

Brooklyn Treatment Court 	Contemporaneous: drug felony defendants arrested
in Brooklyn but not screened in the drug court, primarily because they
were arrested within one of the two geographic zones that were ineligible
to send defendants to the drug court. Defendants met the same paper
eligibility criteria as drug court participants. Excluded defendants with
prior violent felony or misdemeanor convictions, with current violent
offense charges, and those cases in which the arrest did not lead to a
conviction.

Propensity score matching methodology (to identify comparison group
members having similar predicted probabilities of participating in the
drug court as drug court participants); match model based on age, race,
specific charges, and various criminal history measures. Comparison group
members matched to the drug court participant having the nearest
propensity score.

Clark County Drug Court Contemporaneous arrestees: sample drawn Regression
methods to control for individual from criminal defendants who had charges
differences in gender, race/ethnicity, age, prior filed in the Clark
County District Courts but convictions, number of charges and types of did
not enter the drug court. Sample stratified charges (e.g., drug possession
or sales and on drug type and type of charge. economic crimes), prior
controlled substances types, and recidivism in the year after sentencing.

Douglas County Drug Court 	Contemporaneous: felony drug arrestees Matched
to drug court participants on most who had charges filed in court and
serious offense, gender, race and ethnicity, and traditionally
adjudicated.a Most had no more age. Risk-level data were not available for
this than one prior felony conviction, and few had group. Regression
methods to control for prior been convicted of a prior violent felony.
criminal history. Slightly less than half were characterized as medium- or
high-risk offenders.

Kentucky Drug Court Programs Contemporaneous: assessed offenders,
Regression methods to control for individual combined felony and
misdemeanor charge or differences in age, gender, race, location of
conviction. court, year of participation, and criminal justice
involvement.

Appendix II: Evaluations Used Multiple Approaches to Assess Selection Bias

Drug court program and design Comparison group used in evaluation Methods
to address selection bias

Los Angeles County Drug Court Contemporaneous: felony offenders charged

Program	with drug possession but sentenced at trial to probation. Drawn
from the same pool of offenders screened for participation in the drug
court.

Multnomah County STOP Drug Contemporaneous arrestees who were Diversion
Program eligible for the STOP program but did not Matched with the sample
of drug court defendants of a risk scale score (a composite scale based on
criminal justice and social factors and used to determine release risk),
time of hearing, gender, and age. Separate analyses done between drug
court and both comparison group members having various levels of risk.

Matched sample approach to select a group of eligible clients who were
representative of program participants on age, gender, race/ ethnicity,
and prior criminal history.

receive it. Arrested for possession of controlled substance and considered
to be eligible for STOP but did not receive it.

North Carolina Drug Treatment Court

Contemporaneous probationers who would have been eligible for the drug
court but were not admitted. Comparison group was screened for chemical
dependence and sample selected based on drug addiction (from screening)
and eligibility for drug court.

Regression methods to control for individual differences in age, gender,
ethnicity, type of offense (for example, property or drug), offense
seriousness, and graduation status.

      Orange County Drug Court Program Contemporaneous defendants who had

similar charges as the drug court participants and were sentenced to
probation. Drawn from the same pool of offenders screened for
participation in the drug court.

Matched to drug court participants on gender, race and ethnicity, and age.

Rochester Drug Treatment Court 	Contemporaneous arrestees: defendants
Propensity score matching methodology (to arrested in Rochester during the
same period identify comparison group members having of time as the drug
court was in operation but similar predicted probabilities of
participating in who were not arraigned before judges the drug court as
drug court participants); match supportive of the drug court program.
model based on age, race, specific charges, and Excluded defendants that
did not meet the various criminal history measures. Comparison drug
court's paper eligibility criteria (i.e., group members matched to the
drug court current and prior violence) and those cases participant having
the nearest propensity score. in which the arrest did not lead to a
conviction.

Quasi-experimental, both historical and contemporaneous comparison groups

Washington State Court Program Historical and contemporaneous defendants:

Historical group: defendants with cases filed in the 2 years prior to the
start of the drug court in each of the county drug court programs included
in the evaluation.

Contemporaneous group: defendants with similar cases who were processed in
counties without drug courts that were comparable to the counties that did
have drug courts.

Various statistical methods including regressions to control for
individual differences in age, gender, ethnicity, prior youth and adult
convictions, current charges; propensity score matching to identify
comparison group members with similar predicted probabilities of entering
the drug court.

Source: GAO analysis of drug court program evaluations.

aWhile these defendants consisted primarily of those arrested during the
time the drug court program was in operation (a contemporaneous
comparison), some individuals who were arrested in the 3 months prior to
the start of the drug court program were included.

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

The Evaluations We Reviewed and the Net Benefit Measure

In this appendix, we outline the criteria we used in assessing the
costbenefit analyses of drug court program evaluations, and we discuss how
the evaluations we reviewed followed these criteria in their analyses.

Cost-benefit analysis determines the costs associated with implementing or
operating a program and weighs that cost against any benefits expected
from the program. The results of a cost-benefit analysis can be
represented as either a net benefit-calculated as total benefits minus
total costs-or a benefit-to-cost ratio-calculated as total benefits
divided by total costs.

We used the net benefit-benefit minus cost-measure to represent the
cost-benefit analyses of the drug court programs we reviewed. Net benefit
can be used to evaluate the cost savings-that is, the benefits-for each
participant in a drug court program relative to the cost savings for each
offender processed by conventional case processing in the same
jurisdiction.

To assess the cost-benefit analyses of drug court program evaluations, we
reviewed all of the evaluations selected for our structured review that
reported cost or benefit information. Of the 27 evaluations included in
the structured review, 8 reported information about program costs and 4
about benefits. The 8 evaluations we included in our cost-benefit review
are shown in table 4.

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

  Table 4: The Eight Drug Court Program Evaluations in Our Cost-Benefit Review
Reported data on Drug court program evaluation Benefits Costs Included in
          cost-benefit analysis Not included in cost-benefit analysis

                                     Breaking the Cycle Program  Yes      Yes 
                                     (Harrell and others, 2003)        
                  D.C. Superior Court Drug Intervention Program  Yes      Yes 
                           (Harrell, Cavanagh, and Roman, 1998)        
                    Multnomah County (Ore.) STOP Drug Diversion  Yes      Yes 
                              Program (Carey and Finigan, 2003)        
                            Washington State Drug Court Program  Yes      Yes 
                                       (Barnoski and Aos, 2003)        

                                 Douglas County (Neb.) Drug Court  No     Yes 
                  (Martin, Spohn, Piper, and Frenzel-Davis, 2001)       
                                     Kentucky Drug Court Programs  No     Yes 
                               (Logan, Hoyt, and Leukefeld, 2001)       
                   Los Angeles County (Calif.) Drug Court Program  No     Yes 
                                     (Deschenes and others, 1999)       
                         Maricopa County First Time Drug Offender  No     Yes 
                             Program (Deschenes and others, 1996)       

Source: GAO analysis of drug court program evaluations.

Four of the 8 evaluations reported sufficient data on both costs and
benefits, which allowed us to assess the reported net benefits of the drug
court programs. Specifically, we were able to determine whether the
reduction of recidivism-the benefit-would outweigh the additional costs of
a program.

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

Criteria for Assessing a Drug Court Program's Cost-Benefit Analysis

Conducting a cost-benefit analysis is theoretically straightforward-
determine the monetary value of a program's benefits and compare that
value with the monetary value of the program's costs. However, the
analysis is more complicated in practice because of decisions that have to
be made about who to include as recipients of the benefits and how to
measure costs and benefits.1 On the basis of the general principles of
costbenefit analysis, we identified five criteria that we used in
assessing the cost-benefit analyses of the drug court programs we
reviewed. Table 5 describes these criteria; where further explanation is
needed, we discuss them in the text that follows.

Table 5: Five Criteria for Assessing a Cost-Benefit Analysis of a Drug
Court Program

Criterion Description

1. States the program's purpose In general, the purpose of a drug court
program is to reduce repeated criminal behavior-to reduce recidivism-by
reducing offenders' substance-using behavior.

2. Identifies the baseline The baseline, or alternative, is what would
happen to an offender if the drug court program did not exist.

3. Assesses all relevant costs The costs involved in a drug court program
are those associated with the program's operation and those associated
with the baseline.

4. Assesses all relevant benefits Benefits usually attributed to drug
court programs are costs avoided because of reduced recidivism; they
accrue to the criminal justice system and potential victims of crime.
Other benefits an analysis could consider include reduced medical costs
and successful program participants' increased productivity.

5. Assesses uncertainty in cost and Most cost and benefit estimates entail
uncertainty from imprecision in the data underlying

benefit estimates the analysis and the assumptions built into the
analysis. Assessing uncertainty enhances confidence in the estimates used
in evaluation.

                             Source: GAO analysis.

Identifies the Baseline 	A cost-benefit analysis should identify what the
baseline program is. In the case of drug court programs, the analysis
should state what would happen to an offender if the drug court program
did not exist. The costs and benefits of this alternative program or case
processing, called the baseline, are the standard by which the drug court
program costs and benefits are judged. There is no single baseline against
which all drug court programs are compared. For example, one jurisdiction
may offer drug court program

1Generally, a cost is the value of resources used to achieve a benefit.
Economic costs consider the opportunity cost, which is the value of the
resources in their best alternative use. The costs to and the benefits
that accrue to all members of society, rather than just to its taxpaying
members, should be included in the analysis. For example, benefits from a
decrease in crime, such as reduced victimization, accrue to all members of
society, regardless of whether they are taxpayers, and should be included.

  Appendix III: GAO Criteria for Assessing Evaluations of Drug Court Programs'
                             Cost-Benefit Analyses

participation to defendants as an alternative for jail, another as an
alternative for probation.

Assesses All Relevant Costs

Determining Relevant Costs

Measuring Drug Court Program and Baseline Program Costs

A cost-benefit analysis should enumerate and assess all relevant costs. A
cost-benefit analysis of a drug court program should consider two sets of
costs-those associated with the program's operation and those associated
with the baseline. For both sets of costs, the analysis should determine
which costs are relevant and then measure them.

Drug court programs often require additional expenses from criminal
justice system agencies, although these may differ from program to
program. The true cost of a drug court program consists of these
expenses-the resources that would have been used in their next best
alternative use-if not used for the program. Drug court programs have
several basic elements, one of which is ongoing monitoring of the
participants by a judge or other personnel in the criminal justice
system.2 Others are regular status hearings and drug testing, substance
abuse treatment, and the prescription of sanctions for noncompliance with
program requirements. While an analysis should consider all of these
costs, it should not consider costs that were incurred before the program
began (that is, sunk costs) since they are not relevant to the current
decision to fund or continue funding the program.

In addition to the program's costs, the analysis should include the costs
of the baseline. These depend on the types of offenders who are eligible
for the drug court program in each jurisdiction and the alternative
program or processing available for them. The baseline costs may include
the cost of traditional adjudication (including a judge's time during the
early phases of judicial processing), the costs of jail time served, or
costs of probation (such as the salaries of probation officers or any
required monitoring).

The various methods of measuring or estimating costs require varying
amounts of time, data, and analysis. The most straightforward way of

2In an assessment of the costs associated with the salaries of criminal
justice personnel, costs that are directly attributable to the drug court
program (such as the salary of a fulltime drug court judge) should be
included. Costs that are incurred jointly by the program and other
activities (such as costs for probation officers with shared caseloads)
should be allocated among the program and the activities, according to
some generally accepted accounting principles or standards. Finally, costs
that would be incurred in the absence of the program (such as costs for
nondedicated police personnel) should be excluded from the analysis.

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

measuring the costs of both a drug court program and the baseline is to
use budget or expenditure figures from the agencies involved to calculate
average costs.3 While this method has the advantage of simplicity, average
cost may be a misleading measure of resource costs if it does not
accurately portray the resources defendants actually used.

For example, a few very high cost program participants could skew the
average cost to such an extent that actual cost appears to be larger than
it is. This is especially important in assessing average costs associated
with the baseline program. The average baseline defendant in a
jurisdiction may be very different from participants in a drug court
program. Therefore, comparing costs associated with the average defendant
with those of the average drug court participant may not be valid.

A less simple approach, requiring a more intensive use of time and data,
is to conceptualize the cost of drug court program participation as a
series of transactions within the criminal justice system agencies.4 Since
a defendant interacts with a number of these agencies, this method
requires obtaining data about each defendant from each agency. These data
can be aggregated for individual defendants to determine the total amount
of resources used per participant, and these resources can then be
multiplied by their price. This approach has the advantage of allowing a
better determination of the true cost of drug court participation,
relative to participation in an alternative program. However, since it is
more labor intensive, jurisdictions may not have records organized in a
way that allows for tracking individuals across agencies.

Assesses and Measures All Relevant Benefits

As with costs, all relevant benefits of the drug court program should be
assessed relative to the baseline.5 The typical benefit attributed to drug

3Average costs are calculated by dividing the budget totals by the number
of drug court program cases in a given year and by the number of baseline,
or nondrug, cases in the same year.

4This is called the transaction cost approach. For more information about
this methodology, see Carey, S. M., and M. W. Finigan, A Detailed Cost
Analysis in a Mature Drug Court Setting: A Cost-Benefit Evaluation of the
Multnomah County Drug Court, (Portland, Ore.: NPC Research, July 2003).

5The present value of estimated costs and benefits should be calculated.
Researchers should identify the follow-up period in which the benefits
would accrue. The value of future benefits should be discounted by an
appropriate interest rate, to account for the fact that these benefits are
deferred.

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

Challenges in Measuring Drug Court Program Benefits

court programs is avoided costs from reductions in recidivism. Other
benefits that could be considered in the analysis include reduced medical
costs and increased worker productivity stemming from reduced drug
dependency.

Reductions in recidivism can lead to benefits, or cost savings, for
criminal justice agencies and for potential victims of crime. A reduction
in the number of arrests would result in a reduction in expenditures for
the agencies, including police, prosecutors, courts, corrections
departments, and probation agencies.

Potential victims benefit from reduced recidivism. Benefits accrued
because of reduced victimization include direct monetary costs, such as
the value of stolen property and medical expenses, and quality-of-life
costs related to pain and suffering. Not including pain and suffering
costs would underestimate the true cost of crime, but some researchers
exclude these intangibles because it is difficult to assign appropriate
dollar values to them. This results in a more conservative estimate of the
benefit of the drug court program.

In addition, the analysis should, to the extent possible, assess benefits
to society from a reduction in substance abuse. These benefits may include
avoided medical care costs, such as medical services a treated drug addict
did not require. Additionally, if drug court programs are effective, the
labor market outcomes of the participants may improve. For example,
successful participants may be unemployed less often or may earn higher
wages because of increased productivity. To the extent that they may
therefore pay higher taxes, such taxes are benefits to taxpayers in
general.6

The simplest approach to measuring the benefits of reduced recidivism is
using arrest data. This approach has the appeal that an arrest is usually
the beginning of expenditures by a criminal justice agency. Then, the
value of this benefit per arrest can be estimated in the same way that
costs are estimated-by using agency budgets to calculate the average
savings from each avoided arrest.

6These benefits may be difficult to quantify, but earnings and employment
data could be estimated through surveys of participants or could be
obtained from state unemployment insurance agencies.

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

It is possible to measure criminal victimization by using arrest data as
well. However, this approach requires the assumption that each criminal
act results in an arrest. To the extent that this is not true, the
estimation of victimization may be underestimated by a significant margin.
Another approach to measuring victimization would be to use self-reported
criminal behavior by the drug court program and comparison group
participants. However, this approach relies on the forthrightness of the
defendants, which is not ideal. Since both of these methods have
drawbacks, any measurement of criminal victimization is much more
uncertain than measures of expenditures by the agencies.

Measuring the cost of crime may also be problematic. Estimates of the
costs of crimes at the national level are available in the literature.7
For example, these can provide an estimate of the cost of a burglary, and
a researcher can apply the cost to the number of burglaries by the
participant. Ideally, an analysis would also include the costs of crimes
specific to the drug court program's jurisdiction to account for regional
differences. For example, the average value of a car stolen in Miami might
be different from the average value of a car stolen in Tallahassee.

Assesses Uncertainty in Cost and Benefit Estimates

The uncertainty in most cost and benefit estimates is the result of
imprecision in the underlying data used in the analysis and the
assumptions on which the analysis is built. Assessing uncertainty can
enhance the confidence in the estimates in the evaluation. Useful
information for an analysis to report includes the estimates of costs and
benefits and the sensitivity of the cost and benefit estimates to
assumptions made in the analysis.

Estimates of costs and benefits should take into account how likely it is
that a particular outcome (for example, arrest for a crime that will
warrant participation in the drug court program) will occur, in addition
to the value (for example, the cost to victims, the criminal justice
expenditures) of that outcome. Taking account of the likelihood of
occurrence allows the researcher to provide a better assessment of how
reliable the data-the costs and benefits-are.

7A commonly cited reference is Miller, T. R., M. A. Cohen, and B.
Wiersema, Victim Costs and Consequences: A New Look (Washington, D.C.:
National Institute of Justice, Office of Justice Programs, Department of
Justice, January 1996).

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

Uncertainty also derives from assumptions made in an analysis. In an
assessment of an analysis, it is helpful to know which assumptions can be
changed without altering the conclusion of the analysis. Sensitivity
analysis, or systematically varying the assumptions to see what effect
variations have on estimated outcomes, can be applied to several
components of a drug court program's cost analysis.8 For example, one
assumption in the criminal justice area is the numerical relationship
between the commission of a crime and the arrest rate. Not every crime
results in an arrest, but a cost-benefit analysis may have been able to
examine only arrest data, given the lack of other data available. Using
arrest data assumes, in effect, that every crime does result in an
arrest-an assumption that is likely to underestimate the change in crime
that the public experiences. A sensitivity analysis could examine this
possible understatement by increasing the study's assumed rate of crime
per arrest (for example, the Washington State study assumes that 20
percent of robberies result in an arrest) to better approximate the actual
crime rate.

We reviewed eight evaluations of 10 drug court programs for their use of
the criteria. At least five of these evaluations made a concerted effort
to assess all relevant costs as well as benefits. Only two evaluations
assessed uncertainty. However, the two evaluations that conducted
sensitivity analyses were two of the three that did not present costs and
benefits separately so as to allow for an assessment of net benefits.
Table 6 summarizes the evaluations we reviewed and their application of
the five criteria in their assessments.

How the Drug Court Evaluations We Reviewed Used the Identified
Cost-Benefit Principles

8A sensitivity analysis does not, however, provide information about the
range and distribution of the possible values of a given outcome.
Additional analysis can allow a researcher to determine the likely range
of an estimated outcome. In essence, additional analysis can provide an
upper and a lower bound of an estimate, giving an indication of how much
uncertainty there is in the outcome. For example, researchers may estimate
how much time a judge spends on a drug court program participant's case,
and the estimate may be high or low. The range of possible values of time
spent by a judge can be defined, which can help in assessing how reliable
the costs associated with a judge's time are in an analysis.

  Appendix III: GAO Criteria for Assessing Evaluations of Drug Court Programs'
                             Cost-Benefit Analyses

  Table 6: Summary of Criteria Applied in Eight Drug Court Program Evaluation
                  Assessments Assessed costs Assessed benefits

Drug court program evaluation

Stated purpose

Identified
baseline Drug court Baseline Recidivism Other

Assessed uncertainty

BTC Program (Harrell and others, 2003)

                   Birmingham, Ala. Yes Yes Yes Yes Yes No No

                  Jacksonville, Fla. Yes Yes Yes Yes Yes No No

                    Tacoma, Wash. Yes Yes Yes Yes Yes No No

     D.C. Superior Court      Yes     Yes     Yes    Yes    Yes    No      No 
      Drug Intervention                                                 
      Program (Harrell,                                                 
     Cavanagh, and Roman,                                               
            1998)                                                       

     Douglas County (Neb.)     Yes     Yes     Yes    Yes    No    No      No 
      Drug Court (Martin,                                               
       Spohn, Piper, and                                                
     Frenzel-Davis, 2001)                                               

Kentucky Drug Court Yes No Yes No Noa No Yes
Programs (Logan, Hoyt,
and Leukefeld, 2001)

     Los Angeles County      Yes     No     Yes     Yes     No    No      Yes 
     (Calif.) Drug Court                                               
     Program (Deschenes                                                
      and others, 1999)                                                

     Maricopa County First     Yes     Yes     Yes    Yes    No    No      No 
      Time Drug Offender                                                
      Program (Deschenes                                                
       and others, 1996)                                                

      Multnomah County       Yes     Yes     Yes     Yes    Yes    No      No 
      (Ore.) STOP Drug                                                  
      Diversion Program                                                 
     (Carey and Finigan,                                                
            2003)                                                       

     Washington State Drug     Yes     Yes    Yes    Yes    Yes    No      No 
         Court Program                                                  
      (Barnoski and Aos,                                                
             2003)                                                      

Source: GAO analysis of drug court program evaluations.

aThe Kentucky study does include estimates of benefits from reduced
recidivism, but does not appear to include victimization costs.

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

All Evaluations Stated the All eight evaluations provided a statement of
the purpose of the drug court

Program's Purpose 	program, although the purpose of the programs varied
somewhat one from another. Evaluations of some of the programs-for
example, those in Kentucky, Washington, D.C., and Multnomah
County-explicitly stated that the purpose was to address the source of
criminal behavior-drug addiction, dependency, and the like-in order to
reduce recidivism. None of the evaluations, however, included reduced drug
dependency in their calculation of net benefits.

All Evaluations Identified a Baseline

Of the eight evaluations we reviewed, five compared drug court program
participants with the baseline of conventional case processing in the
criminal courts in their jurisdictions. The evaluation of the D.C.
Superior Court Drug Intervention Program used a different baseline for
comparison. This program was an experiment that randomly assigned drug
felony defendants to one of three court dockets. The standard docket
offered defendants weekly drug testing, judicial monitoring, and
encouragement to seek treatment in the community. The sanctions docket
offered a program of graduated sanctions with weekly drug testing,
judicial monitoring of drug use, and referral to community-based
treatment. The treatment docket offered weekly drug testing and an
intensive daily treatment program based in D.C. Superior Court. The
evaluation studied the sanctions and treatment docket participants against
the outcomes of the standard docket participants.

The Evaluations Varied in The eight evaluations varied in the types of
costs reported. All eight

Assessing Costs 	included only costs directly attributable to drug court
programs. None attempted to allocate any shared costs incurred jointly by
a drug court program and another activity not directly related to the
program (such as policing). Further, no evaluation included sunk costs.
All eight evaluations included direct program costs such as urine
analysis. However, they varied in reporting costs for adjudication and
sanctions. Six of the eight evaluations included costs of adjudication or
sanctions.9

9Breaking the Cycle, D.C. Superior Court, Multnomah County, and Washington
State evaluations considered adjudication and sanctions cost. Douglas
County did not include sanctions costs, while Maricopa County did include
confinement and supervision costs. The Los Angeles and Kentucky drug court
program evaluations did not include adjudication or sanction costs.

Appendix III: GAO Criteria for Assessing Evaluations of Drug Court
Programs' Cost-Benefit Analyses

Including Baseline Costs

Estimating Costs

All of the evaluations, except for the Kentucky state evaluation,
attempted to include the cost of the baseline in the analysis.
Consequently, this evaluation did not report a net benefit value.

The evaluations varied in how they estimated costs. The most common method
was to compute average costs. The Multnomah County evaluation differed in
that it determined the cost to the agencies directly from the
administrative records on the drug court program participants and the
comparison group. In addition, the researchers on this evaluation followed
a smaller sample of drug court program and control group participants
through the system, tracking the use of resources in the various
transactions associated with them. For example, the researchers used a
stopwatch at arraignments to determine how much time the judge and other
drug court staff spent in this process.

The Evaluations Varied in Assessing Benefits

In their estimation of benefits attributable to the drug court programs,
five of the evaluations included estimates of drug court program and
criminal justice system costs and avoided victimization costs (Breaking
the Cycle, the D.C. Superior Court Drug Intervention Program dockets,
Kentucky, Multnomah County, and Washington State). The D.C. Superior Court
and Breaking the Cycle evaluations differed substantially from the others
in both the method of measuring victimization and the victim costs. Rather
than using arrest as the indicator of crimes committed, the evaluations
used self-reported criminal activity. In addition, the studies did not use
"quality of life" costs in their measure of cost of crime.

One evaluation estimated just drug court program costs avoided (Douglas
County, Nebraska). One evaluation (Los Angeles County) included only
criminal justice system costs avoided during the treatment period, but it
did not cover recidivism. The Maricopa County evaluation only reported
average costs.

The Evaluations' No evaluation we reviewed presented the likelihood of
estimated costs Assessment of Uncertainty and benefits, and only two
evaluations conducted sensitivity analysis. In Was Limited or the Los
Angeles County Drug Court evaluation, participants were Nonexistent
partitioned into low-risk, medium-risk, and high-risk groups, on the basis

of criminal history and other risk factors. The Kentucky Drug Court
Program evaluation placed its estimates of the benefit-to-cost ratio
within ranges that depended on whether accounting or economic costs were
included, and it included earnings improvements.

Appendix IV: Overview of Drug Court Program Characteristics and
Participants

This appendix provides a general description of drug court program
components and describes program participants in the evaluations we
reviewed. Drug court programs rely on a combination of judicial
supervision and substance abuse treatment to motivate defendants'
recovery. Judges preside over drug court proceedings, which are called
status hearings; monitor defendants' progress with mandatory drug testing;
and prescribe sanctions and rewards, as appropriate in collaboration with
prosecutors, defense attorneys, treatment providers, and others. Drug
court programs can vary in terms of the substance abuse treatment
required. However, most programs offer a range of treatment options and
generally require a minimum of about 1 year of participation before a
defendant completes the program. Drug court program participants can vary
across programs according to differences in eligibility requirements and
jurisdictions. The participants in the drug court programs we reviewed
were predominantly male, generally unemployed at the time of program
entry, and had prior involvement in the criminal justice system.

This section describes typical drug court program approaches, screening
processes and participant eligibility requirements, completion
requirements, treatment components, and sanctions.

                               Drug Court Program

Components

Drug Court Program Approaches

Drug court programs generally have taken two approaches to processing
cases: (1) deferred prosecution (diversion) and (2) post-adjudication. In
the diversion model, the courts defer prosecution dependent on the
offender's agreement to participate in the drug court program. Deferred
adjudication models do not require the defendant to plead guilty. Instead
the defendant enters the drug court before pleading to a charge.
Defendants who complete the treatment program are not prosecuted further
or their charges are dismissed. Failure to complete the program results in
prosecution for the original offense. This approach is intended to
capitalize on the trauma of arrest and offers defendants the opportunity
to obtain treatment and avoid the possibility of a felony conviction.

In contrast, offenders participating in a post-adjudication (post-plea)
drug court program plead guilty to the charge(s) and their sentences are
suspended or deferred. Upon successful completion of the program,
sentences are waived and in many cases records are expunged. This approach
provides an incentive for the defendant to rehabilitate because progress
toward rehabilitation is factored into the sentencing

Appendix IV: Overview of Drug Court Program Characteristics and
Participants

determination. Both of these approaches provide the offender with a
powerful incentive to complete the requirements of the drug court program.

Some drug court programs use both deferred prosecution and
postadjudication approaches and assign defendants to an approach depending
on the severity of the charge. Additionally, drug court programs may also
combine aspects of these models into a hybrid, or combined, approach.

Screening Process and Participant Eligibility Criteria

Defendants reach the drug court program from different sources and at
varying points in case processing. Screening defendants to determine
eligibility for a drug court program generally includes screening them for
legal and clinical eligibility. Initially, defendants are screened for
legal eligibility, based on criminal history and current case information.
Depending on the program, an assistant district or prosecuting attorney,
court clerk, or drug court coordinator typically conducts the review.
Criteria for legal eligibility typically include charging offense, prior
convictions, pending cases, and supervision status. Drug courts generally
accept defendants charged with drug possession or other nonviolent
offenses such as property crimes. Some drug court programs allow
defendants who have prior convictions to participate, and others do not.
Federal grants administered under Title II of the 21st Century Department
of Justice Appropriations Authorization Act are not supposed to be awarded
to any drug court program that allows either current or past violent
offenders to participate in its program.1

After defendants are determined to be legally eligible for the program,
treatment providers or case managers will typically determine defendants'
clinical eligibility. This can be determined through structured assessment
tests, interviews, or even preliminary drug test results. While drug
courts generally only accept defendants with substance abuse problems,
they vary in the level of addiction or type of drug to which defendants
are

1Section 2953 of Title II defines "violent offender" to mean a person who
"(1) is charged with or convicted of an offense, during the course of
which offense or conduct (A) the person carried, possessed, or used a
firearm or dangerous weapon; (B) there occurred the death of or serious
bodily injury to any person; or (C) there occurred the use of force
against the person of another, without regard to whether any of the
circumstances described in subparagraph (A) or (B) is an element of the
offense or conduct of which or for which the person is charged or
convicted; or (2) has one or more prior convictions for a felony crime of
violence involving the use or attempted use of force against a person with
the intent to cause death or serious bodily harm."

Appendix IV: Overview of Drug Court Program Characteristics and
Participants

addicted. For example, some programs do not accept defendants who only
have addictions to marijuana or alcohol, while others do.

Clinical eligibility can also include factors such as medical or mental
health barriers and motivation or treatment readiness. In several drug
court programs in our review, the drug court judge's satisfaction with or
assessment of an offender's motivation and ability to complete the program
was a factor used to screen defendants.

Program Completion Requirements

Judicial Supervision and Status Hearings

Drug Testing Requirements

Drug court programs typically require defendants to complete a 1-year
treatment program in order to graduate from or complete the program. Some
programs impose other conditions that participants must meet in addition
to treatment. These conditions could include remaining drug-free for a
minimum amount of time, not being arrested for a specified period of time,
maintaining employment or obtaining an educational degree or
certification, or performing community service.

The central element of all drug court programs is attendance at the
regularly scheduled status hearings at which the drug court judge monitors
the progress of participants. Monitoring is based on treatment provider
reports on such matters as drug testing and attendance at counseling
sessions. The judge is to reinforce progress and address noncompliance
with program requirements. The primary objectives of the status hearing
are to keep the defendant in treatment and to provide continuing court
supervision. More broadly, judicial supervision includes regular court
appearances and direct in-court interaction with the judge, as well as
scheduled case manager visits.

Monitoring participants' substance use through mandatory and frequent
testing is a core component of drug court programs. Programs vary in the
specific policies and procedures regarding the nature and frequency of
testing. For example, in some programs in our review participants were
required to call to find out whether they are required to be tested in a
given period or on a randomly selected day of the week. The frequency of
testing generally varied depending on the stage or phase of the program
that participants were in.

Treatment Components 	In most drug court programs, treatment is designed
to last at least 1 year and is generally administered on an outpatient
basis with limited inpatient treatment, as needed, to address special
detoxification or relapse situations. Many of the programs operate with
the philosophy that because

Appendix IV: Overview of Drug Court Program Characteristics and
Participants

drug addiction is a disease, relapses can occur and that the court must
respond with progressive sanctions or enhanced treatment, rather than
immediate termination.

Treatment services are generally divided into three phases.
Detoxification, stabilization, counseling, drug education, and therapy are
commonly provided during phases I and II, and in some instances,
throughout the program. Other services relating to personal and
educational development, job skills, and employment services are provided
during phases II and III, after participants have responded to initial
detoxification and stabilization. Housing, family, and medical services
are frequently available throughout the program. In some instances, a
fourth phase consisting primarily of aftercare-related services is
provided. The objectives of drug court program treatment are generally to
(1) eliminate the program participants' physical dependence on drugs
through detoxification; (2) treat the defendant's craving for drugs
through stabilization (referred to as rehabilitation stage) during which
frequent group or individual counseling sessions are generally employed;
and (3) focus on helping the defendant obtain education or job training,
find a job, and remain drug free.

Drug court programs can also either directly provide or refer participants
to a variety of other services and support, and they may include medical
or health care, mentoring, and educational or vocational programs. The use
of community-based treatment self-help groups, such as Alcoholics
Anonymous (AA) and Narcotics Anonymous (NA), and aftercare programs also
varies across drug court programs.

Sanctions for Judges generally prescribe sanctions and rewards as
appropriate in

Noncompliance 	collaboration with prosecutors, defense attorneys,
treatment providers, and others. Typical sanctions for program
noncompliance include oral warnings from the judge; transfer to an earlier
stage of the program; attendance at more frequent status hearings,
treatment sessions, or drug tests; and serving jail time for several days
or weeks. The approach or philosophy for how a drug court judge prescribes
sanctions can vary. For example, some judges use a graduated sanctions
approach, where sanctions are applied in increasing severity. Other judges
may use discretion in prescribing sanctions, assessing participants'
noncompliance on a case-by-case basis.

Appendix IV: Overview of Drug Court Program Characteristics and
Participants

Termination Criteria

Drug Court Program Participants

Drug court programs typically use various criteria for ending a
defendant's participation in the program before completion. These criteria
may include a new felony offense, multiple failures to comply with program
requirements such as not attending status hearings or treatment sessions,
and a pattern of positive drug tests.

Before terminating a defendant for continuing to use drugs, drug court
programs generally will use an array of treatment services and available
sanctions. There are no uniform standards among all programs on the number
of failed drug tests and failures to attend treatment sessions that lead
to a participant's termination. Drug court program judges generally make
decisions to terminate a program participant on a case-by-case basis,
taking into account the recommendations of others, including the treatment
provider, prosecutor, and defense counsel. Relapses are expected, and the
extent to which noncompliance results in terminations varies from program
to program. Once a defendant is terminated, he or she is usually referred
for adjudication or sentencing.

All of the evaluations we reviewed reported some basic substance use or
demographic data about the drug court program participants. However, not
every evaluation reported the same types of information or provided
equivalent levels of detail.

Participants' Substance Use Characteristics

The types of drugs participants reported using varied. Cocaine (crack or
powder) was selected by the highest percentage of participants as the
primary drug of choice in most of the programs reporting these data.2
However, in Baltimore, 77 percent reported heroin was primary drug of
choice, whereas 43 percent of participants in the Tacoma, Washington,
Breaking the Cycle program reported using methamphetamine, suggesting
regional differences in drugs of choice. Participants in evaluations we
reviewed did not always report "hard" drugs as their primary substances of
choice. In several evaluations we reviewed, participants reported that
alcohol or marijuana was their primary drug of choice. For example, in
Chester County, 47 percent of participants reported marijuana, and

2The evaluations used different measures to report drug use, such as
"primary drug of choice," "used in the past 30 days," or "used in the 90
days before program entry." Participants may indicate using more than one
substance.

Appendix IV: Overview of Drug Court Program Characteristics and
Participants

84 percent of participants in Maricopa County reported alcohol as primary
drugs of choice.

Selected Demographic and Socioeconomic Characteristics

The participants in the drug court programs we reviewed were generally in
their early 30s, predominantly male, and generally unemployed at the time
of program entry. Participants' average age at program entry ranged from
24 years (in a misdemeanor-only drug court program in New Castle County,
Delaware) to 36 years (in the Baltimore City Drug Treatment Court).3 In
the majority of these evaluations, participants were, on average, between
30 and 35.

Participants were also predominantly male. The percentage of male
defendants participating in drug court programs we reviewed ranged from 46
percent in Bakersfield, California, to 88 percent in the Jacksonville,
Florida, Breaking the Cycle program. Generally, however, about 60 to 80
percent of participants in these programs were male, which corresponds to
other reviews estimating that across multiple drug court programs about 70
percent of participants are male.

In about half of the evaluations we reviewed that reported data about
participants' race or ethnicity, the majority (50 percent or more) of the
participants were white. However, in some programs, most participants
(that is, over 75 percent of the sample) were predominantly of one racial
or ethnic background. For example, in five of the six Washington state
drug court programs, between 79 and 92 percent of participants were white;
and in the Baltimore and Washington, D.C., drug court programs, between 89
and 99 percent of participants were black. Some other programs reported
that the participants were not predominantly of one racial or ethnic
background. For example, in Los Angeles County, 23 percent of participants
were white, 30 percent black, and 43 percent Hispanic or other.

Drug court participants in the evaluations we reviewed that reported data
on employment and educational status were generally unemployed with less
than a high school education at program entry. Between 16 to 82 percent of
participants were employed at the time of entry into the

3Most evaluations we reviewed reported average age. However, some reported
median age, which ranged from 23 to 34 years.

Appendix IV: Overview of Drug Court Program Characteristics and
Participants

program.4 However, in about two-thirds of the programs reporting these
data, less than half of the participants were employed at the time of
entry into the drug court program. Similarly, in evaluations that reported
information about participants' educational status, 24 to 75 percent of
participants reported that they had less than a high school education at
the time of entry into the program. Generally, about 30 to 60 percent had
less than a high school education at the time of entry into the program.
For example, in the New York State evaluation, across the programs, the
median percentage of participants who have received a general equivalency
diploma (GED) or high school diploma is 45 percent; similarly, the median
percentage employed or in school is 34 percent.

                           Criminal Justice Histories

Participants' prior involvement in the criminal justice system, as
reported in evaluations we reviewed, was generally considerable.5 Most
participants were not first-time offenders. However the types and severity
of involvement varied. In the evaluations that reported these data, the
average number of prior arrests (of any kind) ranged from about 1 to about
13 per participant over different time periods, ranging from 1 to 5 years
before entry into the drug court program. Additionally, in the six New
York State drug court programs, the percent of participants with prior
convictions ranged from 19 percent in Queens, which says it does not
accept participants with prior felony convictions, to 69 percent in
Syracuse. However, the researchers note that less than one-third of prior
convictions in all courts were drug-related, which indicates that
participants are involved in a wider range of criminal activity. This
description of drug court participants' criminal justice system
involvement in the evaluations we reviewed was similar to descriptions
reported by other sources. For example, the Drug Court Clearinghouse and
Technical Assistance Project's 2001 survey of responding adult drug court
programs

4The 82 percent employed occurred in the Las Cruces Municipal DWI Court,
the only DWI court included in our review.

5Most evaluations reported some information on criminal justice system
involvement. However, the measures and types of information reported
varied. For example, some evaluations reported arrests, others
convictions; some reported the average number of prior events-either
arrest or conviction-while others reported the percent of participants
with at least one event.

Appendix IV: Overview of Drug Court Program Characteristics and
Participants

reported that 9 percent of participants had no prior felony convictions
and 56 percent had been previously incarcerated.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

Drug Court Programs Led to Within-Program Recidivism Reductions

In most of the evaluations we reviewed, adult drug court programs led to
statistically significant recidivism reductions during periods of time
that generally corresponded to the length of the drug court program-that
is, within-program.1 For the remaining programs, evaluations showed no
significant differences in recidivism. Our analysis of the evaluations
showed lower percentages of drug court program participants than
comparison group members were rearrested or reconvicted. Program
participants also had fewer recidivism events-that is, incidents of
rearrests or reconvictions-and a longer time until rearrest or
reconviction than comparison group participants. Recidivism reductions
were observed for any felony offense and for drug offenses, whether they
were felonies or misdemeanors. However, the evaluations did not provide
conclusive evidence that specific drug court program components, such as
the judge's characteristics or behavior, the amount of treatment received,
the level of supervision provided, and the sanctions for noncompliance
with program requirements, affect participants' within-program recidivism.
In most of the programs that reported post-program data, recidivism
reductions occurred for some period of time after participants completed
the drug court program.

The 27 evaluations we reviewed provided within-program recidivism
comparisons between drug court program participants and an appropriate
control or comparison group of non-drug court defendants in 23 different
drug court programs. Regardless of the type of comparison group used,
within-program recidivism reductions occurred for various measures of
recidivism, such as rearrests and reconvictions, and prevailed across
different types of offenses.2 Less clear, however, was the effect that
certain drug court program components, such as treatment options or
sanctions, had on participants' recidivism outcomes.

1We use within-program to refer to the period of time that all individuals
who are enrolled are expected to be participating in the drug court
program.

2Appendix II presents details on the research designs and comparison
groups used in the evaluations we reviewed.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

Drug Court Programs Were Associated with Lower Rearrest Rates for Program
Participants

Drug court programs included in our review were associated with reductions
in overall rearrest rates-that is, the percentage of a group arrested for
any new offense (felony or misdemeanor) in a given period of time.
Thirteen drug court programs reported overall rearrest data.3 Ten of these
found statistically significant reductions in overall rearrest rates for
drug court program participants.4 Across studies showing rearrest
reductions, rates of drug court program participants generally ranged from
about 10 to 30 percentage points below those of the comparison group.
Table 7 shows these differences in rearrest rates by the length of the
time frame covered.

3The evaluations we reviewed provided comparative arrest data for drug
court program participants and comparison group members for 13 drug court
programs. In 1 drug court program-the D.C. Superior Court-data were
available for two dockets, or types of drug court program interventions,
to which defendants were assigned.

4Rearrests, as well as reconvictions, do not measure all re-offending, as
every offense or violation does not lead to an arrest.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

Table 7: Differences in Overall Rearrest Rates between Drug Court Program
                   Participants and Comparison Group Members

Percentage point difference, time frame covered

Comparison group used in program Up to 1 year after Up to 2 years after
Drug court program evaluation entry entry

                        Significant reported reductions

Baltimore City Drug Treatment Court  Randomly assigned similar -16%a -15%a 
                                               defendants               
       Breaking the Cycle Program,        Historical-arrestees    -35a  
               Birminghamb                                              
bBreaking the Cycle Program, Tacoma    Historical-arrestees    -15a  
         D.C. Superior Court Drug       Randomly assigned similar  -8a  
               Intervention                    defendants               
        Program, sanctions docket                                       

          Douglas County Drug Court Contemporaneous-traditionally -19a

c

                             adjudicated offenders

     Jackson County Drug Court         Historical-arrestees              -20a 
Kentucky Drug Court Programs      Contemporaneous-assessed      -17a   -4a 
                                  offenders, combined felony and         
                                 misdemeanor charge or conviction        

Los Angeles County Drug Court Program

Contemporaneous-probationers -27a Maricopa County First Time Drug Randomly
               assigned similar -1 -11a (3 yrs) Offender Program

probationers-results for combined groups

       Orange County Drug Court Program Contemporaneous-probationers -10a

                      Nonsignificant reported differences

Breaking the Cycle Program, Historical-arrestees 7

b

Jacksonville

D.C. Superior Court Drug Intervention Randomly assigned similar defendants
-1 Program, treatment docket

Escambia County Drug Court Historical-arrestees

Source: GAO analysis of drug court program evaluations.

Note: When the difference is negative, that indicates that a drug court
program reduced rearrests; when it is positive, that indicates that drug
court program participants had higher rearrest rates than their comparison
group. Nonsignificant differences may be positive or negative, but the
absence of a statistical difference is interpreted to mean that there is
no difference in recidivism between the drug court program and comparison
group.

aThe difference was reported to be statistically significant.

bAccording to the principal researcher of the evaluation, the Breaking the
Cycle program is technically not considered to be a drug court program.
However, we included the program in our review because it incorporates
several of the basic drug court program components, such as drug testing
and access to treatment services.

cWhile these defendants consisted primarily of those arrested during the
time the drug court program was in operation (a contemporaneous
comparison), some individuals who were arrested in the 3 months prior to
the start of the drug court program were included.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

Not All Drug Court Programs Reduced Recidivism for All Offenses

In two drug court programs (Breaking the Cycle program in Jacksonville and
the D.C. Superior Court treatment docket), the program did not lead to
significant differences in overall within-program recidivism. The absence
of a significant difference suggests that these drug court programs did
not necessarily lead to recidivism reductions. In one of these programs
(the D.C. Superior Court), two types of drug court program interventions
were examined-(1) a treatment docket, consisting of an intensive
treatment-only intervention and (2) a sanctions docket, consisting of an
intervention that combined referrals to treatment with graduated
sanctions. The evaluation showed statistically significant reduction in
rearrests for the sanctions docket. However, results for the treatment
docket (treatment alone) were not statistically significant. A third drug
court program in Escambia County showed mixed results. The drug court
program showed no differences for overall rearrest rates (both felonies
and misdemeanors) but showed a significant reduction in felony rearrest
rates.

Drug Court Programs Were Associated with Lower Reconviction Rates for
Program Participants

The evaluations we reviewed showed lower reconviction rates-the percentage
of a group convicted for a new offense in a given period of time-for drug
court program participants than for comparison group members.5 Almost all
of the programs (10 of 12) with these data reported a statistically
significant reduction in reconviction rates for drug court program
participants in one of the time frames covered. The two programs that did
not show statistically significant reductions in reported results in a
direction of reconviction reduction, but not at a statistically
significant level. Table 8 displays the differences in reconviction rates
between drug court participants and comparison group members after up to 1
year and after 2 to 3 years of entry into their respective programs.

5The evaluations we reviewed measured reconviction rates during periods of
up to 3 years after entry into either the drug court program or the
comparison group.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

     Table 8: Differences in Reconviction Rates between Drug Court Program
                   Participants and Comparison Group Members

Percentage point difference, time frame covered

Comparison group used Up to 1 year 2 to 3 years Drug court program in
program evaluation after entry after entry

                        Significant reported reductions

    Bronx Treatment Courta Historical-eligible arrestees -9%b -15%b (3 yrs)

 Brooklyn Treatment Courta Contemporaneous-eligible arrestees -10b -14b (3 yrs)
      Douglas County Drug Court Contemporaneous-traditionally adjudicated

                                       c

 offenders -8b Kentucky Drug Court Programs Contemporaneous-assessed offenders,
                                    -15b -5b

combined felony and misdemeanor charge or conviction

 Maricopa County First Time Drug Randomly assigned similar probationers- -3 -6b
                  Offender Program results for combined groups

    Queens Treatment Courta Historical-eligible arrestees -21b -13b (3 yrs)
Rochester Drug Treatment Courta Contemporaneous-eligible arrestees who -10b -10b
                                    (3 yrs)

were not arraigned before a judge supportive of the program

      Suffolk County Drug                                   -21b -25b (3 yrs) 
           Treatment          Historical-eligible arrestees      
             Courta                                              
       Syracuse Community                                   -16b              
           Treatment          Historical-eligible arrestees      -14b (2 yrs)
             Courta                                              
     Washington State Drug        Historical-arrestees                    -6b 
             Court                                               
     Program, five counties                                      
            combined                                             

                       Nonsignificant reported reductions

    Baltimore City Drug Treatment   Randomly assigned similar defendants  -5  
                Court                                                     
     Washington State Drug Court            Historical-arrestees          
        Program, King County                                              

Source: GAO analysis of drug court program evaluations.

Note: When the difference is negative, that indicates that a drug court
program reduced reconvictions; when it is positive, that indicates that
drug court program participants had higher reconviction rates than the
comparison group members. Nonsignificant differences may be positive or
negative, but the absence of a statistical difference is interpreted to
mean that there is no difference in recidivism between the drug court
program and comparison group.

aThe data for these six drug court programs were provided from one
multisite evaluation of New York state drug court programs.

bThe difference was reported to be statistically significant.

cWhile these defendants consisted primarily of those arrested during the
time the drug court program was in operation (a contemporaneous
comparison), some individuals who were arrested in the 3 months prior to
the start of the drug court program were included.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

The significant differences in reconviction rates between drug court
participants and comparison group members ranged from 8 to 21 percentage
points within 1 year of entry into the drug court program. For the eight
drug court programs in which follow-up data were provided for more than 1
year, statistically significant differences in reconviction rates ranged
from between 5 and 25 percentage points. In five of these drug court
programs, the differences in reconvictions generally increased or remained
about the same as the length of the follow-up period increased.

Drug Court Participants Had Fewer Recidivism Events and Longer Times to
First Recidivism Event than Comparison Group Members Did

Consistent with reducing the percentage of participants rearrested or
reconvicted within a given period, drug court program effects may occur by
reducing the number of recidivism events-the number of arrests or
convictions within a particular period-that participants commit and by
increasing the time until participants commit an offense leading to an
arrest or reconviction. In 8 of 12 programs, drug court participants had
fewer rearrests and in 7 of 9 programs they had fewer reconvictions than
comparison group members. Per 100 drug court program and comparison group
participants, drug court program participants had between 9 and 90 fewer
arrests during the program, or between 18 and 89 fewer convictions during
the program, depending on the recidivism measure reported.

Drug court program participants also generally had longer times to first
arrest or conviction than comparison group members. In 11 of 16 programs
in which the time to first event was reported, drug court program
participants had longer times to the first recidivism event.

Recidivism Reductions Were Observed for Different Offense Types

Evidence showed recidivism reductions for different types of offenses,
specifically for felonies and drug offenses, which may indicate decreased
involvement in substance abuse. In almost all (9 of 12) of the programs
that reported data on felony offense recidivism rates (either rearrest or
reconviction), drug court participants had lower felony recidivism rates
or fewer recidivism events. Similarly, in 11 of 14 of the programs that
reported data on drug offense recidivism, drug court participants were
either rearrested or reconvicted for drug offenses at lower rates.

One drug court program that found no significant reduction in rearrests
did find a reduction in specific offense types. Specifically, the
evaluation of the drug court program in Escambia County found that there
were no significant differences in overall rearrests between drug court
participants

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

and comparison group members within 2 years. However, the evaluation found
significant reductions in felony rearrests that amounted to a 28percentage
point reduction in felony rearrest rates for drug court program
participants. Because of relatively low participation rates, the
researchers cautioned against interpreting such a large reduction in
felony rearrests as indicative of the amount of a reduction that could be
expected from the drug court program were it to be implemented for a
larger numbers of defendants.

Drug Court Program Components Were Infrequently Evaluated and Results Were
Inconclusive

Effects of the Role of the Judge

Effects of Substance Abuse Treatment and Sanctions Imposed

A limited number of evaluations reported information about whether
specific drug court program components affect participants' recidivism,
and their results were mixed. Drug court program judges, the treatment
programs prescribed for participants, and the sanctions used to enforce
compliance with drug court program procedures are three of the basic
components of drug court programs. Two evaluations that we reviewed
provided data on the effects of treatment and supervision on drug court
program participants' within-program recidivism outcomes.

None of the evaluations we reviewed explicitly studied the effect of the
judge-considered to be a critical component of a drug court program- on
participants' within-program recidivism.

In evaluations of two drug court programs, the effects of two basic
components of drug courts-substance abuse treatment and sanctions- were
assessed, and their results varied. The specific types of substance abuse
treatment provided to participants differ among programs. In general,
treatment, depending on the needs of the participant, can include
detoxification, individual and group counseling on an outpatient basis,
substance use prevention education, as well as other health, educational,
vocational, or medical services. Similarly, the types and severity of
sanctions that judges use to enforce compliance with program requirements
vary. These sanctions can include writing an essay, observing drug court
proceedings for several days from the jury box, community service, or
short jail stays. Drug court programs may use a graduated sanctions
approach, where successive infractions are met with increasingly severe
sanctions. In other programs, judges may have discretion to apply
sanctions, as needed, according to the specifics of the case.

To assess the effect of treatment, two evaluations examined recidivism
differences between drug court program participants who attended more
treatment sessions than other participants. In one of the two evaluations

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

that provided the strongest designs for measuring treatment effects-the
evaluation of the Baltimore City Drug Treatment Court-treatment
contributed to reductions in recidivism. However, in the other-the
evaluation of the D.C. Superior Court treatment docket-there were no
differences between the treatment docket and the control group, who were
assigned to a standard docket, consisting of some level of supervision,
monitoring, and access to services.6 In both of these evaluations, drug
court program participants and control group members participated in
treatment, but the drug court program participants received more treatment
and were engaged in a more structured treatment regime than were control
group members.

Findings from these two evaluations also indicate that treatment
requirements combined with supervision and sanctioning contribute the most
to recidivism reductions. While the evaluation of the Baltimore City Drug
Treatment Court measured a distinct difference between those who
participated more heavily in treatment, it also found that those who
received treatment combined with supervision had the lowest recidivism
rates. The D.C. Superior Court sanctions docket, which also combined
judicial supervision with treatment, specifically engaged participants in
a program in which judges applied an established set of graduated
sanctions for failing to comply with drug court program requirements. The
sanctions docket led to fewer rearrests for participants as compared with
the control group assigned to the standard docket. The Baltimore City and
D.C. Superior Court sanctions docket results support the contention that
treatment requirements that are combined with supervision and sanctioning
contribute the most to recidivism reductions.

6The evaluation notes that recidivism analysis does not test the effect of
a strong treatment program because the treatment docket experienced
substantial operational problems, including closure because of building
structural problems and financial problems that precluded certain types of
services from being delivered.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

Limited Evidence Indicates That Recidivism Reductions Endure

While fewer evaluations reported post-program, rather than withinprogram,
recidivism results, those that did indicate that recidivism reductions
endure beyond the time that participants are engaged in the program.7
Evaluations we reviewed reported post-program recidivism for longer
periods for 17 drug court programs. In 13 of these 17 programs, drug court
program participants had lower rearrest or reconviction rates than
comparison group members. One evaluation of 6 programs defined explicit
post-program periods of about 1 year, and drug court program participants
had lower recidivism than comparison group members of 5 of the 6 programs.
Among drug court program participants, graduates had lower post-program
recidivism than dropouts.

                               Lower Rearrest and
                               Reconviction Rates

Significantly lower rearrest and reconviction rates were reported for
participants in 13 of the 17 drug court programs that reported these data.
As shown in table 9, for 6 drug court programs reporting significant
reductions, the differences in rearrest rates between drug court program
participants and comparison group members ranged between 4 to 20
percentage points. The evaluation of the Multnomah County drug court
program reported only the mean number of arrests (recidivism events), but
it also showed post-program reduction in recidivism using this measure.
For the 9 drug court programs reporting significant reductions, the
differences in reconviction rates between drug court participants and
comparison group members ranged from 5 to 25 percentage points.

7We generally use the term post-program to refer to some period of time
after drug court participants and comparison group members completed their
respective programs. However, for some drug court programs, the
evaluations measured recidivism for a period after participants left the
program. These represent explicit post-program periods. Postprogram
recidivism outcomes were generally observed for no longer than 1 year.
Hence, it is not possible to determine whether the differences in
recidivism observed between drug court program participants and comparison
groups would persist.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

Table 9: Differences in Rearrest and Reconviction Rates, over Longer Time
Periods, between Drug Court Program Participants and Comparison Group
Members

                Recidivism measure, percentage point difference

Time frame Time frame Drug court program Rearrested covered after entry
Reconvicted covered after entry

     Significant reported reductions Nonsignificant reported differences or
                              significant increase

    Baltimore City Drug     -15a       2 years            
      Treatment Court                                     
Bronx Treatment Courtb                            -15a       Up to 4 years 
     Brooklyn Treatment                              -14a       Up to 4 years 
           Courtb                                         
    Jackson County Drug     -20a       2 years            
           Court                                          
    Kentucky Drug Court      -4a 1 year post-program  -5a 1 year post-program 
          Programs                                        
Maricopa County First    -11a                      -6a                     
     Time Drug Offender                3 years                        3 years
          Program                                         
Multnomah County STOP  -0.94a                          
       Drug Diversion                  2 years            
          Programc                                        
    North Carolina Drug     -10a 1 year post program      
      Treatment Court                                     
      Queens Treatment                               -13a       Up to 4 years 
           Courtb                                         
       Rochester Drug                                -10a       Up to 4 years 
      Treatment Courtb                                    
    Suffolk County Drug                              -25a       Up to 4 years 
      Treatment Courtb                                    
     Syracuse Community                              -14a       Up to 4 years 
      Treatment Courtb                                    
Washington State Drug                              -6a                     
    Court Program, five                                               3 years
     counties combined                                    

            Clark County Drug Court                         10a  2 years post 
                                                                      program 
          Escambia County Drug Courtd         -10  2 years       
      Southern San Mateo County Drug Court     2   2 years       
Washington State Drug Court Program, King                  -2      3 years 
                     County                                      

Source: GAO analysis of drug court program evaluations.

Note: When the difference is negative, that indicates that a drug court
program reduced recidivism; when it is positive, that indicates that drug
court program participants had higher recidivism rates than members of
their comparison group.

aThe difference was reported to be statistically significant.

bThe data for these six drug court programs were provided from one
multisite evaluation of New York state drug court programs.

cThe Multnomah County results show differences in the average number of
arrests only.

dEscambia County drug court showed a significant reduction in felony
rearrest rates but showed no significant difference for overall rearrest
rates (both felonies and misdemeanors).

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

Evidence that a gap in post-program recidivism can increase over time
comes from the evaluation of the Maricopa County drug court program. In
Maricopa County, there were no reported differences in recidivism
(specifically rearrest rates) during the 1-year within-program time frame.
However, a 3-year follow-up evaluation found a significant (11 percentage
point) difference in rearrest rates between the drug court program
participants and control group members. The difference arose largely from
an increase in the percentage of control group members that had been
rearrested by the end of the 3-year follow-up; 32 percent of the control
group were rearrested by 1 year, but by 3 years, 44 percent were
rearrested. For the drug court participants, 31 percent had been
rearrested at the end of 1 year, but only an additional 2 percent of the
sample were rearrested within 3 years.

Recidivism Reductions in For the six programs with explicit post-program
periods, participants had Explicit Post-Program significantly lower
recidivism than comparison group members in all but Periods one program,
as shown in table 10.

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

Table 10: Post-Program Differences in Recidivism in Six New York State
Drug Court Programs

Percentage point difference in New York state drug court program
Comparison group used in evaluation recidivism, 1-year post-program

                        Significant reported reductions

Bronx Treatment Court 	Historical arrestees: eligible defendants arrested
in the Bronx prior to the start of the drug court program. -13%a

Brooklyn Treatment Court 	Contemporaneous: eligible drug felony defendants
arrested in Brooklyn but not screened in the drug court, primarily because
they were arrested within one of the two geographic zones that were
ineligible to send defendants to the drug court program. -6b

Queens Treatment Court 	Historical arrestees: eligible defendants arrested
in Queens in the year prior to the drug court program's implementation.
-13a

Rochester Drug Treatment Court 	Contemporaneous arrestees: eligible
defendants arrested in Rochester during the same period of time as the
drug court was in operation but who were not arraigned before a judge
supportive of the program. -7b

Suffolk County Drug Treatment Court 	Historical arrestees: eligible
defendants arrested in Suffolk County in the year prior to the start of
the drug court program. -9b

                       Nonsignificant reported reduction

Syracuse Community Treatment Court 	Historical arrestees: eligible
defendants arrested in Syracuse City in the year prior to the start of the
drug court program.

Source: GAO analysis of drug court program evaluations.

aSignificant at at least the 1 percent level.

bSignificant at the 5 percent level.

In five of the six New York state drug court programs, drug court
participants had significantly lower post-program recidivism rates than
did their comparison group members. The differences in recidivism,
measured by the percentage with a new arrest leading to a conviction for 1
year after drug court participants left the program, ranged from 6 to 13
percentage points, as shown in table 10. These findings indicate that for
some programs at least, drug court programs can be effective at reducing
criminal behavior while participants are engaged in the program and after
they leave the program. The evaluation of these drug court programs
reports that there is evidence that the effects of drug court programs on
recidivism do not diminish over time, but because the post-program period

Appendix V: Drug Court Programs Are Associated with Recidivism Reductions

was only 1 year, that improved understanding of the longer-run effects of
drug courts on recidivism would benefit from additional research that
extends the post-program follow-up time frames.

Graduates Had Lower Post-Program Recidivism than Dropouts

Several evaluations compared post-program recidivism outcomes for drug
court program graduates with those of dropouts-those participants who were
terminated from the program either by their withdrawal or by sanctioning.
These evaluations show that the post-program recidivism differences
observed between all drug court program and comparison group participants
arise primarily from the large recidivism differences between drug court
program graduates and dropouts. Evaluations that reported these
comparisons showed large differences in the recidivism rates of drug court
program graduates compared with both dropouts and comparison group
members. For example, in three New York state drug court programs,
dropouts were four to seven times more likely to be reconvicted than
graduates. Specifically, in the Bronx Treatment Court, 29 percent of
dropouts were reconvicted during the first year after the program as
compared with only 4 percent of graduates. In addition, the post-program
recidivism rates of dropouts were generally no different from (or in some
cases greater than) the recidivism rates of comparison group members.

Appendix VI: Limited and Mixed Evidence on Drug Court Programs' Impacts on
Substance Use Relapse

Limited Evidence on Drug Court Programs' Impact on Substance Use Relapse

Evidence about the effectiveness of drug court programs in reducing
participants' substance use relapse is limited and mixed. The evidence on
substance use relapse outcomes is limited to data available from eight
drug court programs included in our review. The data include drug test
results and self-reported drug use; both measures were reported for some
programs. Drug test results generally showed significant reductions in use
during participation in the program, while self-reported results generally
showed no significant reductions in use.

The evidence on substance use relapse outcomes is limited to data
available from eight drug court programs.1 Evaluations in these eight drug
court programs used two different measures of substance use-drug test
results and self-reported use-to compare relapse between participants and
comparison group members. Three of these programs used both measures to
assess relapse. Table 11 shows the comparison group used in the programs'
evaluation and the results for the eight drug court programs that
presented substance use relapse data, grouped by the type of drug use
measure used.

1Some other evaluations we reviewed provided descriptive data on drug
court program participants' substance use while they were engaged in the
drug court program, but these evaluations did not assess substance use
relapse as an outcome by comparing the outcomes of participants against
those of a comparison group.

Appendix VI: Limited and Mixed Evidence on Drug Court Programs' Impacts on
                             Substance Use Relapse

Table 11: Substance Use Relapse Results, by Different Measures of Use, for
Programs GAO Reviewed

Comparison group used in Drug court program evaluation of program Results
of substance use relapse Drug tests

Chester County Drug Court Historical-probationers

Maricopa County First Time Drug Offender All individuals identified as
eligible for the Program drug court program were randomly

Reduction within-program: There were significantly lower rates of positive
drug test results for drug court program participants.

No reduction within-program: Among all drug court and probation control
group members, there were no significant differences in the percent that
tested positive for any drug. However, significantly lower percentages of
drug court participants tested positive for cocaine. Significantly higher
percentages of drug court participants tested positive for marijuana,
valium, and alcohol.

assigned to one of three study groups at the time of their initial
probation assignment. The three study groups were:

(1) probation with no testing;

(2) probation with low-rate (monthly) drug testing; and

(3) probation with high-rate (biweekly) drug testing.

                             Self-reported drug use

Breaking the Cycle Program, Birmingham Historical-arrestees 	Reduction
within-program: Participants reported lower use of any drug during the
past 30 days and significantly lower use of stronger drugs (such as
cocaine and heroin).

Breaking the Cycle Program, Jacksonville Historical-arrestees 	Reduction
within-program: Participants reported significantly lower use during the
past 30 days of any drug, lower heavy drug use (defined as 16 or more days
of illegal drug use), and lower marijuana use. They reported no
differences in use of stronger drugs (such as cocaine).

Breaking the Cycle Program, Tacoma Historical-arrestees 	No reduction
within-program: There was no difference in self-reported drug use on rates
or types of drugs.

                     Both drug tests and self-reported use

D.C. Superior Court Drug Intervention Program, sanctions docket

Randomly assigned eligible defendants 	Reduction within-program: Drug
tests were taken during the final month of the program-that is,
within-program. Twentyseven percent of drug court participants tested
drug-free in all drug tests compared with 11 percent of the control group.
Fiftythree percent of drug court participants had a "bad test outcome"
(negative result, missed test, or tampered result) as compared with 71
percent of the control group.

No reduction post-program: There were no differences in self-reported drug
use that occurred during the year after sentencing.

Appendix VI: Limited and Mixed Evidence on Drug Court Programs' Impacts on
Substance Use Relapse

Comparison group used in Drug court program evaluation of program Results
of substance use relapse

D.C. Superior Court Drug Intervention Program, treatment docket

Randomly assigned eligible defendants 	Reduction within-program: Drug
tests were taken during the final month of the program. Twenty-two percent
of drug court participants tested drug-free in all drug tests as compared
with 11 percent of the control group. Also, 64 percent of the drug court
participants had "bad test outcome" (negative result, missed test, or
tampered result) as compared with 71 percent of the control group.

No reduction post-program: There were no differences in self-reported drug
use that occurred during the year after sentencing.

New Castle County Drug Court Program	Drug court program participants who
were randomly assigned to biweekly versus "as needed" judicial status
hearings

Reduction within-program for subgroups: There were significant reductions
in positive drug test results for participants with mental health problems
and prior treatment who were assigned to the biweekly condition compared
with those assigned to the "as-needed" condition. There were significant
reductions in positive test results for those without mental health
problems assigned to the "as-needed" condition, but not to the biweekly
condition. However, there were no overall differences either in drug tests
or self-reported drug use between groups assigned to biweekly or
"as-needed" hearings.

Source: GAO analysis of drug court program evaluations.

All of the drug test results on substance use relapse were limited to
those obtained when participants were still engaged in the drug court
program. After drug court program participants exit the control of the
criminal justice system, researchers must obtain participants' voluntary
consent to obtain substance use relapse data, regardless of whether the
data are selfreported or based upon urinalysis or other testing methods.2

2According to researchers, collecting these data can be expensive. Also,
given likely differences between those who might volunteer and those who
would decline, the collected data might have limited value.

Appendix VI: Limited and Mixed Evidence on Drug Court Programs' Impacts on
Substance Use Relapse

Evidence on Drug Court Programs' Impact on Substance Use Relapse Is Mixed

The results on substance use relapse are mixed. Four of the five drug
court programs that used drug test results reported reductions in use.
However, self-reported data on substance use relapse show contradictory
results.

In both of the D.C. Superior Court dockets, prior to sentencing, drug
court program participants had better drug test results than did
defendants who had been randomly assigned to the control group. Similarly,
in the Chester County Drug Court, within-program drug relapse was observed
using drug test results. Drug court program participants had fewer
positive urinalysis results than did a comparison group of eligible
offenders who had been placed on probation in the 10 months prior to the
implementation of the drug court program. These results indicate decreased
relapse among participants despite the fact that drug court program
participants were tested more frequently than were the members of the
comparison group. However, in Maricopa County, there were no significant
differences in the percentage of drug court participants and the combined
control groups that tested positive.

One evaluation we reviewed provided limited evidence that judicial status
hearings reduce within-program substance use for certain types of drug
court program participants. This evaluation, an experiment in a drug court
program in New Castle County, randomly assigned eligible drug court
program defendants to one of two forms of judicial status hearing
conditions: (1) biweekly, regularly scheduled status hearings or (2)
status hearings scheduled on an as-needed basis. During the 14-week drug
court program, participants with antisocial personality disorder (APD) and
those who had prior drug treatment episodes had significantly lower levels
of substance use-as measured by drug test results-when they were assigned
to regularly scheduled biweekly judicial status hearings as compared with
when they were assigned to status hearings on an asneeded basis.3

Four of the six programs with self-reported drug use results reported no
significant reductions in use. For example, after sentencing, neither of
the D.C. Superior Court dockets' participants reported lower rates of
substance use. Alternatively, in two of the three Breaking the Cycle
programs, participants reported lower rates of substance use than did

3Preliminary data from replication experiments confirmed the results of
the original experiment, in that the researchers found evidence of the
interaction between the mental health status of participants and biweekly
status hearings.

Appendix VI: Limited and Mixed Evidence on Drug Court Programs' Impacts on
Substance Use Relapse

comparison group members. These differences in self-reported substance use
persisted even after the researchers controlled for sample differences and
selection. The use of self-reported data presents challenges related to
underreporting that we have discussed in a prior report.4

4 For more information, please see GAO, Drug Use Measurement: Strengths,
Limitations, and Recommendations for Improvement, GAO/PEMD-93-18
(Washington, D.C.: June 25, 1993).

                   Appendix VII: Evidence Is Mixed about the
                    Factors That Predict Program Completion

As discussed in appendix V, participants who completed the drug court
program (that is, graduates) had much lower recidivism rates than those
who dropped out. Further, the recidivism rates for drug court program
dropouts are comparable to the rates of comparison group members.
Completion rates-an indicator of the extent to which participants
successfully complete their drug court program requirements-for
participants in selected programs we reviewed ranged from about 30 percent
to about 70 percent.1

In evaluations of 16 drug court programs in which completion was assessed,
one factor-drug court program participants' compliance with program
procedures-was consistently associated with program completion. These
program procedures include attending treatment sessions, producing
drug-free urinalysis test results, and appearing at status hearings. No
other program factor, such as the severity of the sanction that would be
imposed if participants failed to complete the program or the manner in
which judges conducted status hearings, predicted participants' program
completion. Several characteristics of the drug court program participants
themselves were also associated with an increased likelihood of program
completion. These characteristics, while not assessed in all of the
programs, include lower levels of prior involvement in the criminal
justice system, and age, as older participants were more likely to
complete drug court programs than younger ones.

Completion rates ranged from 27 percent to 66 percent for 16 drug court
programs included in evaluations that assessed program completion. These
rates, while consistent with rates reported in other reviews of multiple
drug court programs, are not directly comparable because drug court
programs have different program completion requirements, the rates were
measured over varying time periods, and study designs can affect the
completion measures. Participants are not only required to attend
treatment, but also to appear in court on a regular basis and follow other
rules that are specific to the drug court program. These rules can
involve, but are not limited to, policies on drug testing, case manager or
probation officer visits, school or job attendance, support groups, and
graduation

1Completion rates represent the individuals who completed or were
favorably discharged from a drug court program as a percentage of the
total number admitted and not still enrolled.

Program Completion Rates Vary

Appendix VII: Evidence Is Mixed about the Factors That Predict Program
Completion

requirements. Table 12 shows completion rates for the drug court programs
whose evaluations assessed program completion.

Table 12: Completion Rates for Selected Drug Court Programs Included in
GAO's Review

                       Drug court program Completion rate

                      Bakersfield Municipal Drug Court 36%

Bronx Treatment Court

                            Brooklyn Treatment Court

(Rempel and others, 2003)

(Rempel and DeStefano, 2001)

                           Broward County Drug Court

(Schiff and Terry, 1997)

(Senjo and Leip, 2001)

                          Clark County Drug Court 52a

Creek County Drug Court

                       Escambia County Drug Court Program

(Peters and Murrin, 2000)

(Peters, Haas, and Murrin, 1999)

(Truitt and others, 2002)

                       Jackson County Drug Court Program

(Anspach and Ferguson, 2003) 29

(Truitt and others, 2002) 27 Multnomah County STOP Drug Diversion Program
51a New Castle County Drug Court

As-needed hearings 58

Biweekly hearings 49

Okaloosa County Drug Court 53

Orange County Drug Court Program 43

Queens Treatment Court 66

St. Mary Parish Drug Court 32

Suffolk County Drug Treatment Court 62

Syracuse Community Treatment Court 41 Source: GAO analysis of drug court
program evaluations.

aEvaluation reports "favorable" status, which includes defendants active
in treatment plus those who graduated.

     Appendix VII: Evidence Is Mixed about the Factors That Predict Program
                                   Completion

Program Completion Is Associated with Compliance with Program Procedures

As noted earlier, drug court programs vary in their specific program
completion requirements. Similarly, programs may terminate defendants'
program participation for a variety of reasons. Drug court program
participants can be terminated from the program if they are arrested for a
new offense, especially for felony offenses; if they regularly fail to
appear at status hearings or treatment sessions; or if they repeatedly do
not comply with program procedures. Consecutive positive drug test results
do not always lead to program termination; in some programs, this could
lead to a change in the treatment services provided or sanctions.

Drug court participants' compliance with drug court procedures, such as
appearing at treatment sessions and remaining drug-free during the
program, was generally a strong predictor of program completion. The level
of program compliance is typically indicated by the degree to which
participants follow rules and procedures determined by the drug court
program, attend required meetings and treatment sessions, and generally
progress toward recovery. Given that program completion is related to
compliance with drug court procedures, we sought to assess whether the
evaluations included in our review provided conclusive evidence about
specific aspects of compliance that were associated with program
completion. For example, in some drug court programs greater levels of
participation were associated with greater likelihood of completing the
drug court program. Similarly, participants having more within-program
arrests, more instances of warrants for failure to appear, and more
positive drug tests were generally less likely to complete the program
than those having fewer of these.

However, for some of the specific aspects of compliance, the findings were
not consistent across drug court programs and in some cases they were even
contradictory. For example, one evaluation of four drug court programs
(Bakersfield, Creek County, Jackson County, and St. Mary's Parish) found
an inconsistent relationship between drug test results and program
completion across the four drug court programs. This cross-site evaluation
assessed how various aspects of compliance predicted completion in each
program using the same measures of compliance in each drug court program.
In relation to the association between drug test results and completion,
the evaluation found that in one program (Creek County) drug test results
did not have a significant effect on completion. In another drug court
program (St. Mary's Parish), drug court program participants that had no
positive drug test results were more likely to complete the program than
were participants who had comparatively "low" or "moderate" numbers of
positive test results, whereas there were

     Appendix VII: Evidence Is Mixed about the Factors That Predict Program
                                   Completion

no differences in program completion between those participants who had
comparatively "high" numbers of positive test results and those who had no
positive test results.2 In the two remaining programs (Bakersfield and
Jackson County), participants with positive test results in the low range
actually had a higher likelihood of completing the program than
participants with no positive test results, while in Jackson County,
participants with positive test results in the high range were less likely
than those with no positive test results to complete the program. This
evaluation suggests that there may not be consistency across every drug
court program in the relationship between drug test results and program
completion.

Initial Engagement of Drug Court Participants

Some research studies indicate that drug court participants' first few
weeks in treatment are predictive of success. Early engagement in the drug
court program has been measured by whether participants attended treatment
during the first few weeks after program entry and whether participants
have fewer indications of noncompliance (such as failure to appear or
warrants) during the first month of program participation.

One evaluation of five drug court programs in New York State assessed the
role of participants' early engagement in the drug court program on their
chances of completing the program.3 It found, consistently among the drug
court programs, that early engagement by participants, measured by whether
the participant absconded from program contact within 30 days of program
entry significantly predicted program completion. Across the five New York
drug court programs, participants that received warrants within 30 days of
program entry were from about three to eight times more likely to fail to
complete the drug court program than were participants who did not receive
a warrant within 30 days of program entry.4

2The evaluation classified the percent of positive drug test results into
four categories: (1) zero positive test results; (2) "low," where the
percentage of test results that were positive ranged from 1 to 8 percent
of drug tests; (3) "moderate," where the percentage of test results that
were positive ranged from 9 to 28 percent of drug tests; and (4) "high,"
where the percentage of test results that were positive ranged from 29 to
100 percent.

3The drug court programs were in the Bronx, Brooklyn, Queens, Suffolk
County, and Syracuse.

4As defined in the evaluation of these drug court programs, a judge issues
a bench warrant when a defendant fails to make a scheduled court
appearance. Bench warrants can also be issued if defendants fail to appear
for scheduled appointments with case managers or treatment
representatives.

     Appendix VII: Evidence Is Mixed about the Factors That Predict Program
                                   Completion

Further, in the Brooklyn Treatment Court, participants' compliance with
drug court program procedures early in the program also contributed to
participants' completing 90 days of drug treatment. Those participants who
disappeared from contact, prompting the issuance of a police warrant, had
lower chances of completing 90 days of treatment than those participants
who did not. Additionally, those participants who attended at least 1 day
of treatment within 30 days of entering the program also had greater
chances of completing 90 days of treatment than those who did not attend
at least 1 day of treatment within the first 30 days of entering the
program.

Other aspects of compliance, such as treatment attendance during the
entire program, appearance at status hearings, within-program arrests, and
failure to appear, were generally but not consistently associated with
completion. For example, the percentage of expected treatment sessions
attended increased the likelihood of completion in five of seven drug
court programs.5 On the other hand, in two of the four drug court programs
in which they were measured, within-program arrests and failure to appear
decreased the likelihood of completion but had no effect in the other two
drug court programs.6 Alternatively, in both drug court programs
(Multnomah and Clark Counties) in which the effects of within-program
sanctions were assessed, an increase in the number of sanctions ordered
led to a decrease in the likelihood of completion.

In several drug court programs, the effects of various drug court program
components were examined to determine the factors that predict program
completion. Among the components that were examined were various
consequences of program failure and the role of the judge and judicial
status hearings. In a few drug court programs, efforts were made to
examine the role of individual motivation to participate in drug court
programs. In addition, evaluations of a few drug court programs examined
the role of social factors in predicting program completion. Specifically,

Effects of Drug Court Programs' Components on Program Completion Are Mixed

5The five in which treatment attendance increased the chances of
completion were (1) Bakersfield, (2) Clark County, (3) Creek County, (4)
St. Mary's Parish, and (5) Multnomah, while the two in which it did not
affect completion were Jackson County and Broward County.

6In the Bakersfield and Jackson County drug court programs, these effects
were observed, but in the Creek County and St. Mary's Parish drug court
programs, the effects were not observed.

     Appendix VII: Evidence Is Mixed about the Factors That Predict Program
                                   Completion

these evaluations assessed variables that measured individual's
attachments to other individuals and social institutions.

Legal Consequences of Program Failure

Drug Court Judge

Several drug court program evaluations in our review assessed the effects
of different sanctions, or legal consequences of failure, on program
completion. For example, if a drug court program participant fails to
complete the program, he or she may be sentenced to a predetermined
incarceration alternative. The results of these evaluations were mixed and
not directly comparable.7 For example, in the Brooklyn Treatment Court,
participants with a more serious treatment mandate-that is, those
participants that faced longer jail or incarceration terms if they failed
to comply with program requirements-were more likely to complete the drug
court program than those with a less serious treatment mandate. The
predetermined jail sentences for Brooklyn participants who failed to
complete the program ranged from 6 months to 4 1/2 years, depending on the
severity of the charge (felony or misdemeanor) and prior criminal history
(whether it was a first felony offense or not).

Alternatively, in the drug court program in Suffolk County, the length of
the incarceration sentence faced by those who failed to complete the drug
court program did not contribute to program completion. The predetermined
sentence (minimum length of the most common prison alternative) in Suffolk
County ranged from 6 months to 1 year. The researchers who conducted the
evaluation did not provide an explanation for these differing effects.8

The judge has been described in the research as a key component of a drug
court program. The presence of consistent judicial monitoring of
participants is also described as a distinguishing component of drug court

7A number of variables are used to indicate the amount of legal coercion
facing participants. These include the type and severity of charge leading
participants to enter the drug court program, the length of the
incarceration time participants face if they fail the drug court program
(only in Suffolk), jail or prison alternatives (not used in Bronx,
Suffolk, and Syracuse), the "treatment mandate," which is used in one
program (Brooklyn) as a proxy for jail or prison alternative because there
is a fairly uniform sentencing approach based on the treatment mandate.

8Rempel, M., and others, The New York State Adult Drug Court Evaluation:
Policies, Participants, and Impacts. New York: Center for Court
Innovation, 2003, p. 100.

Appendix VII: Evidence Is Mixed about the Factors That Predict Program
Completion

programs, compared with other court-based treatment programs, such as
probation. Several evaluations in our review examined the effect of drug
court judges on program completion. Each demonstrated that judges can play
an important role in contributing to participants' program completion.
However, the effect of the judge on program completion is difficult to
distinguish from the requirement to attend judicial status hearings. For
example, one evaluation of drug court program completion in Broward County
attempted to assess the types of comments that judges made and the effect
of these comments on program completion. Court monitoring comments made by
judges that the researchers classified as supportive were found to
contribute to program completion.9 On the other hand, in one evaluation of
two drug court programs (Clark and Multnomah Counties), the number of
appearances before the drug court judge was found to increase the
likelihood of program completion.

A third evaluation (in New Castle County, Delaware) assessed the
regularity of judicial status hearings on program completion. Participants
were randomly assigned to one of two different forms of judicial status
hearings: (1) biweekly hearings and (2) hearings on an as-needed basis, as
determined by court officials. There were two main findings from this
evaluation (that were also supported by the two replications of the
experiment). First, there was no "direct" effect of the different status
hearing schedules on program completion rates. Second, there was an
interaction between client characteristics and program completion. Drug
court program participants that either had prior treatment experiences or
were diagnosed as having APD were more likely to complete the program when
they were assigned to the regular biweekly status hearings as compared
with when they were assigned to status hearings on an asneeded basis.
Conversely participants who were not diagnosed as having APD or who had no
prior treatment experiences were more likely to complete the program when
they were assigned to status hearings on an as-needed basis, as compared
with the biweekly condition.

The results from the New Castle County experiment show that the regularity
of the schedule of status hearings can contribute to program completion
for distinct subpopulations of drug court program participants

9Senjo, S. R., and L. A. Leip. "Testing and Developing Theory in Drug
Court: A Four-Part Logit Model to Predict Program Completion," Criminal
Justice Policy Review, vol. 12, no. 1, 2001, p. 82.

     Appendix VII: Evidence Is Mixed about the Factors That Predict Program
                                   Completion

Participants Who Were Older and Had Less Criminal History than Other
Participants Were Most Likely to Complete Programs

and that there may not be a "one size fits all" approach to scheduling
court appearances for all participants.

The evaluations of several drug court programs examined some of the
attributes of drug court participants and how those factors were related
to program completion. Attributes such as prior substance abuse treatment,
prior criminal history, type of drug used, demographic characteristics,
and employment and education were assessed. Prior criminal history,
whether measured by the number of arrests or convictions prior to program
entry, and age, as older were generally likely to complete their drug
court programs than were younger participants to complete the program,
were related to program completion.

The other attributes were generally found not to be significant predictors
of completion, although among various minorities of drug court programs in
which they were assessed, they were significant predictors of completion.
For example, prior substance abuse treatment was a significant predictor
of completion in three of seven drug court programs, but it was not
significant in the other four drug court programs. Similarly, other
attributes such as type of drug use, race or ethnicity, gender, or
employment or education level were not observed to consistently predict
completion.

The evaluations that assessed participants' attributes were correlational-
that is, they that examined the associations among these factors and the
probability of program completion. Although not all of the evaluations
included all of the same factors in their analyses, some general patterns
about the correlates of program completion emerged. Attributes of drug
court program participants that were associated with an increased
likelihood of program completion include the following: (1) lower levels
of prior criminal history; (2) substance use other than cocaine or heroin;
(3) employment or school attendance at the time of program intake, along
with higher levels of education; and (4) age, as older participants were
more likely to complete the programs.10 One evaluation attempted to
measure the effect of motivation and readiness for treatment on program
completion. It found that those participants who were better able to

10However, not all of these factors consistently predicted program
completion in all drug court programs in which they were examined.

Appendix VII: Evidence Is Mixed about the Factors That Predict Program
Completion

recognize their problems, recognize external problems, and were ready for
treatment, were more likely to complete the drug court program.

Appendix VIII: Evidence from Four Evaluations Indicates Drug Court
Programs' Positive Net Benefits

Four evaluations in our review included sufficient information about seven
drug court programs' costs and benefits to estimate their net benefits-
that is, benefits less costs.1 All but one of the evaluations found that
drug court programs were more expensive than conventional case processing.
The costs to operate drug court programs above the costs of providing
conventional case processing services ranged from about $750 to about
$8,500 per participant. However, taking into account the drug court
programs' benefits, especially the reduced costs of crime associated with
reductions in recidivism, all four evaluations we reviewed reported net
benefits ranging from about $1,000 per participant to about $15,000,
mostly because of reduced victimization. Additionally, these benefits may
underestimate drug court programs' true benefits because the evaluations
did not include indirect benefits (such as reduced medical costs of
treated participants).

All but one of the evaluations we reviewed found drug court programs more
expensive than conventional case processing. A combination of judicial
supervision, monitoring, and treatment services that drug court programs
typically provide to their participants-services in addition to those
provided to offenders who are processed with conventional procedures
receive-result in additional expense to criminal justice agencies. Table
13 shows the net costs of drug court programs-costs above normal court
costs-including supervision, monitoring, and treatment, ranging from
somewhat less than $800 to about $8,700 per participant.

Most Evaluations Found Drug Court Programs More Expensive than
Conventional Case Processing

1Appendix III provides more information about our review of drug court
program evaluations that presented cost data and the criteria we used to
assess them.

  Appendix VIII: Evidence from Four Evaluations Indicates Drug Court Programs'
                             Positive Net Benefits

Table 13: Reported Costs per Drug Court Program Participant in Four Evaluations

      Drug court program Costs above normal costs D.C. Superior Court Drug
                              Intervention Program

Treatment docket $ 8,708

Sanctions docket 3,248

                   Washington State Drug Court Program 3,892

a

                           Breaking the Cycle Program

Tacoma, Wash. 1,461

Jacksonville, Fla. 1,133

Birmingham, Ala.

                  Multnomah County (Ore.) STOP Program (1,442)

Source: Drug court program evaluations.

aAccording to the principal researcher of the evaluation, the Breaking the
Cycle program is technically not considered to be a drug court program.
However, we included the program in our review because it incorporates
several of the basic drug court program components, such as drug testing
and access to treatment services.

The Multnomah County drug court program cost about $1,400 less than normal
court procedures. Rather than relying on administrative records and
budgets, the program's evaluation used a methodology that closely followed
participants through treatment and court adjudication. It calculated the
amount of time that each drug court participant and comparison group
participant spent in different activities, such as treatment and court
hearings. The evaluation found that the judge spent less time with
offenders in the drug court program than in normal court processing and
that this led to the estimated decreased costs.

Drug Court Programs' Net Benefits Derived from Reductions in Recidivism

The monetary benefits of reduced recidivism can be placed in two
categories: (1) reduced future expenditure by criminal justice agencies
and (2) reduced future victimization. Any arrest is a cost to a number of
criminal justice agencies, including police, prosecutors, courts,
corrections departments, and probation agencies. Reducing arrests by
reducing recidivism would benefit these agencies.2 The justice system's
benefits in the seven drug court programs we reviewed ranged from none to
about $3,800 per participant. Some of the range is due to methodological
differences among the evaluations, but some may be due

2A reduction in arrests would result in a benefit to the taxpaying members
of society, since taxpayers fund the criminal justice agencies.

Appendix VIII: Evidence from Four Evaluations Indicates Drug Court
Programs' Positive Net Benefits

to differences among the communities served by the courts. A reduction in
recidivism also benefits people who might otherwise be victimized. The
costs to potential victims of crime that are thus avoided include direct
monetary costs, such as the value of property that is not stolen and
expenses for health care that are not incurred, and quality-of-life costs,
such as costs for pain and suffering that are not experienced. Benefits to
potential victims reported in the evaluations we reviewed ranged from
about $500 to $24,000 per participant.

The evaluations we reviewed monetized the benefits from averted crime
inconsistently, but the differences in method do not explain the range of
benefits. For example, excluding the cost of pain and suffering from
victimizations that do not occur underestimates the true cost of crime.
However, the two evaluations that did not include these costs-D.C.
Superior Court and Breaking the Cycle-found the highest and lowest per
participant dollar values of reduced recidivism, at $24,000 and $500,
respectively.

All Seven Drug Court Programs Reported Positive Net Benefits

Although six of the seven drug court programs were more costly than
conventional case processing, the monetary value of the benefits from
reduced recidivism-to the justice system and potential victims-was greater
than the costs, producing positive net benefits in all seven programs. The
net benefits of the seven drug court programs we reviewed ranged from
about $1,000 to about $15,000 per participant. Table 14 presents benefits
to the justice system and to potential victims, the net costs of the
programs (costs of the drug programs above conventional case processing),
and net benefits, or benefits minus net costs.

Appendix VIII: Evidence from Four Evaluations Indicates Drug Court
Programs' Positive Net Benefits

Table 14: Net Benefits per Participant in Seven Drug Court Programs Benefits Net
               benefits (benefits less costs) Drug court program

                                 Justice system

Avoided potential victim

Costs (net drug court program costs)

                                 Justice system

                                Justice system plus avoided potential victims

  D.C. Superior Court Drug Intervention Program Treatment docketa $ 24,030b $
                                 8,708 $ 15,322

 Sanctions docket $ 19c 6,203b 3,248 ($ 3,230) 2,973 Breaking the Cycle Program

        Tacoma, Wash.         394c     7,324b     1,461     (1,067)     6,257 
      Jacksonville, Fla.       0c      2,900b     1,133     (1,133)     1,767 
       Birmingham, Ala.      1,320c     479b       767        553       1,032 
       Multnomah County                                               
     (Ore.) STOP Program      2,329     1,301    (1,442)     3,771      5,072 
    Washington State Drug                                             
        Court Program         3,759     3,020     3,891      (132)      2,888 

Source: GAO analysis of drug court program evaluations' reported costs and
benefits.

aThe D.C. Superior Court Drug Intervention Program evaluation did not draw
conclusions about the net benefits of the treatment docket because of
inconsistencies between self-reported crime and official arrest records.

bBenefits to potential victims were derived from survey data and do not
include the cost of pain and suffering.

cBenefits to the justice system were determined by official records of
arrest.

However, as shown in the last two columns of table 14, the financial cost
savings due to reductions in recidivism for the criminal justice agencies
were not always positive. Positive financial cost savings for the criminal
justice agencies were indicated for only two programs-Breaking the Cycle
Program in Birmingham and the drug court program in Multnomah County.

None of the evaluations included indirect, or secondary, benefits to
society derived from a reduction in participants' substance abuse.
Indirect benefits might include costs avoided because treated drug addicts
did not use medical services that would otherwise have been required.
After successful drug treatment, such individuals might have fewer periods
of unemployment and might be more productive, earning higher wages. To the
extent that they pay higher taxes as a result, these are benefits to
taxpaying members of society. While these benefits are difficult to
quantify in assessing a drug court program, their absence suggests that
reported net benefits are understated.

Net Benefits May Underestimate the Programs' True Benefits

Appendix IX: GAO Contacts and Staff Acknowledgments

GAO Contacts 	Laurie E. Ekstrand, (202) 512-8777 William J. Sabol, (202)
512-3464

Acknowledgments 	In addition to those named above, Mary Catherine Hult,
David P. Alexander, Michele C. Fejfar, Benjamin A. Bolitzer, Harold J.
Brumm, Jr., Wayne A. Ekblad, Ann H. Finley, Ronald La Due Lake, Jean L.
McSween, Albert Schmidt, Barry J. Seltser, Douglas M. Sloane, Shana B.
Wallace, and Kathryn G. Young made key contributions to this report.

Bibliography

Anspach, D. F., and A. S. Ferguson. Assessing the Efficacy of Treatment
Modalities in the Context of Adult Drug Courts: Final Report. Portland,
Maine: University of Southern Maine, Department of Sociology, 2003.

Aos, S., P. Phipps, R. Barnoski, and R. Lieb. The Comparative Costs and
Benefits of Programs to Reduce Crime, A Review of National Research
Findings with Implications for Washington State. Olympia, Wash.:
Washington State Institute for Public Policy, May 1999.

Banks, D. and D. C. Gottfredson. "The Effects of Drug Treatment and
Supervision on Time to Rearrest among Drug Treatment Court Participants."
Journal of Drug Issues, vol. 33, no. 2 (2003): 385-414.

Barnoski, R., and S. Aos. Washington State's Drug Courts for Adult
Defendants: Outcome Evaluation and Cost-Benefit Analysis. Olympia, Wash.:
Washington State Institute for Public Policy, 2003.

Belenko, S. Research on Drug Courts: A Critical Review 2001 Update. The
National Center on Addiction and Substance Abuse at Columbia University,
June 2001.

Breckenridge, J. F., L. T. Winfree, J. R. Maupin, and D. L. Clason. "Drunk
Drivers, DWI `Drug Court' Treatment, and Recidivism: Who Fails?" Justice
Research and Policy, vol. 2, no. 1 (2000): 87-105.

Brewster, M. P. "An Evaluation of the Chester County (PA) Drug Court
Program." Journal of Drug Issues, vol. 31, no. 1 (2001): 177-206.

Carey, S., and M. Finigan. A Detailed Cost Analysis in a Mature Drug Court
Setting: A Cost-Benefit Evaluation of the Multnomah County Drug Court.
Portland, Ore.: Northwest Professional Consortium, July 2003.

Craddock, A. North Carolina Drug Treatment Court Evaluation: Final Report.
Washington, D.C.: U.S. Department of Justice, Office of Justice Programs,
Drug Court Program Office, 2002.

Croxton, F. E., D. J. Cowden, and S. Klein. Applied General Statistics,
3rd edition. (Englewood Cliffs, N.J.: Prentice-Hall, 1967).

Deschenes, E. P., L. Cresswell, V. Emami, K. Moreno, Z. Klein, and C.
Condon. Success of Drug Courts: Process and Outcome Evaluations in Orange
County, California, Final Report. Submitted to the Superior Court of
Orange County, California, September 20, 2001.

Bibliography

Deschenes, E. P., I. Imam, T. L. Foster, L. Diaz, V. Moreno, L. Patascil,
D. Ward, and C. Condon. Evaluation of Orange County Drug Courts for Orange
County Superior Courts. Richmond, Calif.: The Center for Applied Local
Research, 1999.

Deschenes, E. P., I. Imam, E. Castellonos, T. L. Foster, C. Ha, D. Ward,
C. Coley, and K. Michaels. Evaluation of Los Angeles County Drug Courts.
Richmond, Calif.: The Center for Applied Local Research, 2000.

Deschenes, E. P., S. Turner, P. Greenwood, and J. Chiesa. An Experimental
Evaluation of Drug Testing and Treatment Interventions for Probationers in
Maricopa County, Arizona. Santa Monica, Calif.: RAND, July 1996.

Festinger, D. S., D. B. Marlowe, P. A. Lee, K. C. Kirby, G. Bovasso, and
A. T. McLellan. "Status Hearings in Drug Court: When More Is Less and Less
Is More." Drug and Alcohol Dependence, vol. 68 (2002): 151-157.

Fielding, J. E., G. Tye, P. L. Ogawa, I. J. Imam, and A. M. Long. "Los
Angeles County Drug Court Programs: Initial Results." Journal of Substance
Abuse Treatment, vol. 23 (2002): 217-224.

Finigan, M. An Outcome Program Evaluation of the Multnomah County S.T.O.P.
Drug Diversion Program. Portland, Ore.: Northwest Professional Consortium,
1998.

Goldkamp, J. S., M. D. White, and J. B. Robinson. "Do Drug Courts Work?
Getting Inside the Drug Court Black Box." Journal of Drug Issues, vol. 31,
no. 1 (2001): 27-72.

Goldkamp, J. S., M. D. White, and J. B. Robinson. From Whether to How Drug
Courts Work: Retrospective Evaluation of Drug Courts in Clark County (Las
Vegas) and Multnomah County (Portland)-Phase II Report from the National
Evaluation of Drug Courts. Philadelphia, Pa.: Crime and Justice Research
Institute, 2001.

Goldkamp, J. S., M. D. White, and J. B. Robinson. Retrospective Evaluation
of Two Pioneering Drug Courts: Phase I Findings from Clark County, Nevada,
and Multnomah County, Oregon: An Interim Report of the National Evaluation
of Drug Courts. Philadelphia, Pa.: Crime and Justice Research Institute,
2000.

Bibliography

Gottfredson, D. C., S. S. Najaka, and B. Kearley. "Effectiveness of Drug
Treatment Courts: Evidence from a Randomized Trial." Criminology and
Public Policy, vol. 2, no. 2 (2003): 171-196.

Gottfredson, D. C., and M.L. Exum. "The Baltimore City Drug Treatment
Court: One-Year Results from a Randomized Study." Journal of Research in
Crime and Delinquency, vol. 39, no. 3 (2002): 337-356.

Harrell, A., S. Cavanagh, and J. Roman. Final Report: Findings from the
Evaluation of the D.C. Superior Court Drug Intervention Program.
Washington D.C.: Urban Institute, 1998.

Harrell, A. and J. Roman. "Reducing Drug Use and Crime among Offenders:
The Impact of Graduated Sanctions." Journal of Drug Issues, vol. 31, no. 1
(2001): 207-232.

Harrell, A., O. Mitchell, J. Merrill, and D. Marlowe. Evaluation of
Breaking the Cycle. Washington D.C.: The Urban Institute, February 2003.

Harrell, A., O. Mitchell, A. Hirst, D. Marlowe, and J. Merrill. "Breaking
the Cycle of Drugs and Crime: Findings from the Birmingham BTC
Demonstration." Criminology and Public Policy, vol. 1, no. 2 (2002):
189216.

Listwan, S. J., J. L. Sundt, A. M. Holsinger, and E. J. Latessa. "Effect
of Drug Court Programming on Recidivism: The Cincinnati Experience." Crime
& Delinquency, vol. 49, no. 3 (2003): 389-441.

Logan, T. K, W. Hoyt, and C. Leukefeld. Kentucky Drug Court Outcome
Evaluation: Behavior, Costs, and Avoided Costs to Society. Lexington, Ky.:
Center on Drug and Alcohol Research, University of Kentucky, October 2001.

Marlowe, D. B., D. S. Festinger, and P. A. Lee. "The Judge Is a Key
Component of Drug Court." National Drug Court Institute Review, vol. IV,
no. 2 (2004): 1-34.

Marlowe, D. B., D. S. Festinger, P. A. Lee, M. Schepise, J. E. R. Hazzard,
J.

C. Merrill, F. D. Mulvaney, and A. T. McLellan. "Are Judicial Status
Hearings a Key Component of Drug Court? During-Treatment Data from a
Randomized Trial." Criminal Justice and Behavior, vol. 30, no. 2 (2003):
141-162.

Bibliography

Martin, T. J., C. Spohn, R. K. Piper, and E. Frenzel-Davis. Phase III
Douglas County Drug Court Evaluation: Final Report. Omaha, Neb.: Institute
for Social and Economic Development, May 2001.

Miethe, T. D., H. Lu, and E. Reese. "Reintegrative Shaming and Recidivism
Risks in Drug Court: Explanations for Some Unexpected Findings." Crime &
Delinquency, vol. 46, no. 4 (2000): 522-541.

Office of Management and Budget Circular A-94, "Guidelines and Discount
Rates for Benefit-Cost Analysis of Federal Programs." (October 29, 1992,
revised January 29, 2002)

Peters, R. H., and M. R. Murrin. "Effectiveness of Treatment-Based Drug
Courts in Reducing Criminal Recidivism." Criminal Justice and Behavior,
vol. 27, no. 1 (2000): 72-96.

Peters, R. H., and M. R. Murrin. Evaluation of Treatment-Based Drug Courts
in Florida's First Judicial Circuit. Tampa, Fla.: Department of Mental
Health Law and Policy, Louis de la Parte Florida Mental Health Institute,
University of South Florida, 1998.

Peters, R. H., A. L. Haas, and M. R. Murrin. "Predictors of Retention and
Arrest in Drug Courts." National Drug Court Institute Review, vol. II, no.
I (1999): 30-57.

Rempel, M., and C. DeStefano. "Predictors of Engagement in Court-Mandated
Treatment: Findings from the Brooklyn Treatment Court, 19962000." Journal
of Offender Rehabilitation, vol. 33, no. 4 (2001): 87-124.

Rempel, M., D. Fox-Kralstein, A. Cissner, R. Cohen, M. Labriola, D.
Farole, A. Bader, and M. Magnani. The New York State Adult Drug Court
Evaluation: Policies, Participants, and Impacts. New York: Center for
Court Innovation, 2003.

Roman, J., and A. Harrell. "Assessing the Costs and Benefits Accruing to
the Public from a Graduated Sanctions Program for Drug-Using Defendants."
Law and Policy, vol. 23, no. 2 (2001): 237-268.

Roman, J., J. Woodard, A. Harrell, and S. Riggs. Final Report: A
Methodology for Measuring Costs and Benefits of Court-Based Drug
Intervention Programs Using Findings from Experimental and
Quasiexperimental Evaluations. Washington D.C.: The Urban Institute,
December 1998.

Bibliography

Schiff, M., and W. C. Terry, III. "Predicting Graduation from Broward
County's Dedicated Drug Treatment Court." The Justice System Journal, vol.
19, no. 3 (1997): 291-310.

Senjo, S. R., and L. A. Leip. "Testing and Developing Theory in Drug
Court: A Four-Part Logit Model to Predict Program Completion." Criminal
Justice Policy Review, vol. 12, no. 1(2001): 66-87.

Spohn, C., R. K. Piper, T. Martin, and E. D. Frenzel. "Drug Courts and
Recidivism: The Results of an Evaluation Using Two Comparison Groups and
Multiple Indicators of Recidivism." Journal of Drug Issues, vol. 31, no. 1
(2001): 149-176.

Stokey, E., and Zeckhauser. A Primer for Policy Analysis. (New York: W.W.
Norton and Company, 1978).

Truitt, L., W. M. Rhodes, N. G. Hoffman, A. M. Seeherman, S. K. Jalbert,
M. Kane, C. P. Bacani, K. M. Carrigan, and P. Finn. Evaluating Treatment
Drug Courts in Kansas City, Missouri and Pensacola, Florida: Final Reports
for Phase I and Phase II. Cambridge, Mass.: Abt Associates Inc., 2002.

Turner, S., P. Greenwood, T. Fain, and E. Deschenes. "Perceptions of Drug
Court: How Offenders View Ease of Program Completion, Strengths and
Weaknesses, and the Impact on Their Lives." National Drug Court Institute
Review, vol. II, no. 1(1999): 58-81.

Wilson, D. B., O. Mitchell, and D. L. MacKenzie. "A Systematic Review of
Drug Court Effects on Recidivism." Forthcoming.

Wolfe, E., J. Guydish, and J. Termondt. "A Drug Court Outcome Evaluation
Comparing Arrests in a Two Year Follow-Up Period." Journal of Drug Issues,
vol. 2, no. 4 (2002): 1155-1172.

GAO's Mission	The Government Accountability Office, the audit, evaluation
and investigative arm of Congress, exists to support Congress in meeting
its constitutional responsibilities and to help improve the performance
and accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.

Obtaining Copies of The fastest and easiest way to obtain copies of GAO
documents at no cost

is through GAO's Web site (www.gao.gov). Each weekday, GAO postsGAO
Reports and newly released reports, testimony, and correspondence on its
Web site. To Testimony have GAO e-mail you a list of newly posted products
every afternoon, go to

www.gao.gov and select "Subscribe to Updates."

Order by Mail or Phone	The first copy of each printed report is free.
Additional copies are $2 each. A check or money order should be made out
to the Superintendent of Documents. GAO also accepts VISA and Mastercard.
Orders for 100 or more copies mailed to a single address are discounted 25
percent. Orders should be sent to:

U.S. Government Accountability Office 441 G Street NW, Room LM Washington,
D.C. 20548

To order by Phone:	Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061

To Report Fraud, Contact:
Waste, and Abuse in Web site: www.gao.gov/fraudnet/fraudnet.htm

E-mail: [email protected] Programs Automated answering system: (800)
424-5454 or (202) 512-7470

Congressional	Gloria Jarmon, Managing Director, [email protected] (202)
512-4400 U.S. Government Accountability Office, 441 G Street NW, Room 7125

Relations Washington, D.C. 20548

Public Affairs	Paul Anderson, Managing Director, [email protected] (202)
512-4800 U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548
*** End of document. ***