Justice Impact Evaluations: One Byrne Evaluation Was Rigorous;	 
All Reviewed Violence Against Women Office Evaluations Were	 
Problematic (07-MAR-02, GAO-02-309).				 
                                                                 
Discretionary grants awarded under the Bureau of Justice	 
Assistance's (BJA) Byrne Program help state and local governments
make communities safe and improve criminal justice. Discretionary
grants awarded under BJA's Violence Against Women Office (VAWO)  
programs are aimed at improving criminal justice system responses
to domestic violence, sexual assault, and stalking. The National 
Institute of Justice (NIJ) awarded $6 million for five Byrne	 
Program and five VAWO discretionary grant program evaluations	 
between 1995 and 2001. Of the 10 programs evaluated, all five	 
VAWO evaluations were designed to be both process and impact	 
evaluations of the VAWO programs. Only one of the five Byrne	 
evaluations was designed as an impact evaluation and the other	 
four evaluations were process evaluations. GAO's in-depth review 
of the four impact evaluations since fiscal year 1995 showed that
only one of these--the evaluation of the Byrne Children at Risk  
Program--was methodologically sound. The other three evaluations,
all of which examined VAWO programs, had methodological problems.
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-02-309 					        
    ACCNO:   A02714						        
  TITLE:     Justice Impact Evaluations: One Byrne Evaluation Was     
Rigorous; All Reviewed Violence Against Women Office Evaluations 
Were Problematic						 
     DATE:   03/07/2002 
  SUBJECT:   Federal aid for criminal justice			 
	     Program evaluation 				 
	     Discretionary grants				 
	     Grant monitoring					 
	     Byrne Children at Risk Program			 
	     DOJ Civil Legal Assistance Program 		 
	     DOJ Encourage Arrests Policies Program		 
	     DOJ Rural Domestic Violence and Child		 
	     Victimization Enforcement Grant Program		 
                                                                 
	     DOJ STOP Violence Against Indian Women		 
	     Discretionary Grant Program			 
                                                                 
	     Edward Byrne Memorial State and Local		 
	     Law Enforcement Assistance Program 		 
                                                                 
	     DOJ Grants to Combat Violent Crimes		 
	     Against Women on Campuses				 
                                                                 

******************************************************************
** 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-02-309
     
United States General Accounting Office

GAO

Report to Congressional Requesters

March 2002

JUSTICE IMPACT EVALUATIONS

One Byrne Evaluation Was Rigorous; All Reviewed Violence Against Women
Office Evaluations Were Problematic

GAO-02-309

Contents

Letter

Results in Brief
Background
Scope and Methodology
Number, Type, Status of Completion, and Award Amount of Byrne

and VAWO Discretionary Grant Program Evaluations Methodological Problems
Have Adversely Affected Three of Four

Impact Evaluations Conclusions Recommendations for Executive Action Agency
Comments and Our Evaluation 1

2 4 7

8

10 18 18 19

                           Appendix I Appendix II

Summaries of the Impact Evaluations of Byrne and VAWO Programs

NIJ's Guidelines  for Disseminating  Discretionary Grant  Program Evaluation
Results

NIJ's Dissemination Practices Dissemination of NIJ's Byrne Discretionary
Grants Comprehensive Communities Program and Children at Risk Program

                                     25

                                   29 29

                                     31

    Appendix III     Comments from the Department of Justice                          32
                                              GAO Comments                            43
     Appendix IV     GAO Contacts and Staff Acknowledgments                           47
                                              GAO Contacts                            47

                                             Acknowledgments                          47

Tables

Table 1:  NIJ Evaluations of  Byrne Discretionary Grant Programs  9 Table 2:
NIJ Evaluations of VAWO Discretionary Grant Programs 10

Abbreviations

BJA Bureau of Justice Assistance NIJ National Institute of Justice OJP
Office of Justice Programs VAWO Violence Against Women Office

United States General Accounting Office Washington, DC 20548

March 7, 2002

The Honorable Charles E. Grassley
Ranking Minority Member
Subcommittee on Crime and Drugs
The Honorable Jeff Sessions
Ranking Minority Member
Subcommittee on Administrative Oversight and the Courts
Committee on the Judiciary
United States Senate

This report responds to your request that we review selected aspects of
the U.S. Department of Justice's (Justice) Office of Justice Programs'
(OJP) program evaluations of discretionary grants awarded by OJP's
Bureau of Justice Assistance's (BJA) Byrne Program (Byrne)1 and the
Violence Against Women Office (VAWO). Between fiscal years 1997 and
2000, Byrne and VAWO discretionary grant awards grew, in constant fiscal
year 2000 dollars, about 85 percent-from about $105 million to
approximately $194 million. These funds were awarded directly to various
organizations, such as state and local governments, either on a competitive
basis or pursuant to legislation allocating funds through congressional
earmarks or direction. Discretionary grants awarded under the Byrne
program were designed to help state and local governments make
communities safe and improve criminal justice. Discretionary grants
awarded under VAWO programs are aimed at improving criminal justice
system responses to domestic violence, sexual assault, and stalking.
Questions have been raised, however, regarding what these and other
Justice grant programs have accomplished.

To address your request, we are reporting on (1) the number, type, status
of completion, and award amount of Byrne and VAWO discretionary grant
program evaluations during fiscal years 1995 through 2001 and (2) the
methodological rigor of the impact evaluation studies of Byrne and VAWO
discretionary grant programs during fiscal years 1995 through 2001. In
addition, information that you requested on OJP's approaches to

1The Edward Byrne Memorial State and Local Law Enforcement Assistance
Program. The Byrne discretionary grant program represents the largest single
discretionary grant program within BJA.

Results in Brief

disseminating evaluation results about Byrne and VAWO discretionary grant
programs is presented in appendix II. Our review covered discretionary grant
program evaluations managed by Justice's National Institute of Justice
(NIJ). Program evaluations are systematic studies that are conducted
periodically or ad hoc to assess how well a program is working. These
studies can include impact evaluations-designed to assess the net effect of
a program by comparing program outcomes with an estimate of what would have
happened in the absence of the program- and process evaluations-designed to
assess the extent to which a program is operating as intended. NIJ is OJP's
principal research and development agency and is responsible for evaluating
Byrne and VAWO programs. For Byrne program evaluations, NIJ awarded funding
to evaluators using its own funds and funds made available through BJA. For
VAWO program evaluations, NIJ awarded funding to evaluators using funds made
available exclusively through VAWO.

During fiscal years 1995 through 2001, NIJ awarded about $6 million for five
Byrne and five VAWO discretionary grant program evaluations. Of the 10
program evaluations, all 5 VAWO evaluations were designed to be both process
and impact evaluations of the VAWO programs. By contrast, only one of the
five Byrne evaluations was designed as an impact evaluation and the four
other Byrne evaluations were process evaluations.

Our in-depth review of the four impact evaluations that have progressed
beyond the formative stage since fiscal year 1995 showed that only one of
these, the evaluation of the Byrne Children at Risk (CAR) Program, was
methodologically sound. The other three evaluations, all of which examined
VAWO programs, had methodological problems that raise concerns about whether
the evaluations will produce definitive results. Although program evaluation
is an inherently difficult task, in all three VAWO evaluations, the effort
is particularly arduous because of variations across grantee sites in how
the programs are implemented. In addition, VAWO sites participating in the
impact evaluations have not been shown to be representative of their
programs, thereby limiting the evaluators' ability to generalize results,
and the lack of comparison groups hinders evaluators' ability to minimize
the effects of factors that are external to the program. Furthermore, data
collection and analytical problems (e.g., related to statistical tests,
assessment of change) compromise the evaluator's ability to draw appropriate
conclusions from the results. Peer review committees found methodological
problems in two of these three evaluation proposals that we examined, but it
is unclear what NIJ has done to effectively resolve those problems.

We are making a recommendation regarding the two VAWO impact evaluations in
the formative stage of development, and for all future impact evaluations,
to ensure that potential methodological design and implementation problems
are mitigated.

In commenting on a draft of this report, OJP's Assistant Attorney General
stated that she agreed with the substance of our recommendations and said
that NIJ has begun or plans to take steps to address them. Although it is
still too early to determine whether NIJ's actions will be effective in
preventing or resolving the problems we identified, they appear to be steps
in the right direction. The Assistant Attorney General also said that the
report could have gone further in acknowledging the challenges that
evaluators face when conducting research in the complex environment of
criminal justice programs and interventions. In addition, she stated that
the report contrasts the Byrne evaluation with the three VAWO evaluations
and obscures important programmatic differences that affect an evaluator's
ability to achieve GAO's conditions for methodological rigor.

Our report applies standards of methodological rigor that are well defined
in scientific literature. We recognize that there are substantive
differences in the intent, structure, and design of the various
discretionary grant programs managed by OJP and its bureaus and offices and
that impact evaluation can be an inherently difficult and challenging task,
especially since the Byrne and VAWO programs are operating in an
ever-changing, complex environment. Not all evaluation issues that can
compromise results are easily resolvable, but with more up-front attention
to design and implementation issues, there is a greater likelihood that NIJ
evaluations will provide meaningful results for policymakers. Absent this
up-front attention, questions arise as to whether NIJ is (1) positioned to
provide the definitive results expected from an impact evaluation and (2)
making sound investments given the millions of dollars spent on these
evaluations. The full text of the Assistant Attorney General's comments and
our evaluation of them are presented in appendix III and elsewhere in this
report, as appropriate.

Background

The Justice Assistance Act of 1984 (P.L. 98-473) created OJP to provide
federal leadership in developing the nation's capacity to prevent and
control crime, administer justice, and assist crime victims.2 OJP carries
out its responsibilities by providing grants to various organizations,
including state and local governments, Indian tribal governments, nonprofit
organizations, universities, and private foundations. OJP comprises five
bureaus, including BJA, and seven program offices, including VAWO.3 In
fulfilling its mission, BJA provides grants for programs and for training
and technical assistance to combat violent and drug-related crime and help
improve the criminal justice system. VAWO administers grants to help prevent
and stop violence against women, including domestic violence, sexual
assault, and stalking. During fiscal years 1995 through 2001, BJA and VAWO
awarded about $943 million to fund 700 Byrne and 1,264 VAWO discretionary
grants.

One of BJA's major grant programs is the Byrne Program.4 BJA administers the
Byrne program, just as its counterpart, VAWO, administers its programs.
Under the Byrne discretionary grants program, BJA provides federal financial
assistance to grantees for educational and training programs for criminal
justice personnel; for technical assistance to state and local units of
government; and for projects that are replicable in more than one
jurisdiction nationwide. During fiscal years 1995 through 2001, Byrne
discretionary grant programs received appropriations of about $385 million.

2This act amended the Omnibus Crime Control and Safe Streets Act of 1968 to
eliminate the Law Enforcement Assistance Administration and established OJP,
which was to be headed by an assistant attorney general.

3OJP's other four bureaus are Bureau of Justice Statistics, NIJ, Office of
Juvenile Justice and Delinquency Prevention, and Office for Victims of
Crime. OJP's other six program offices are American Indian & Alaska Native
Affairs Desk, Executive Office for Weed and Seed, the Corrections Program
Office, the Drug Courts Program Office, the Office for Domestic Preparedness
(formerly the Office for State and Local Domestic Preparedness Support), and
the Office of the Police Corps and Law Enforcement Education.

4Local law enforcement programs established by the Anti-Drug Abuse Act of
1986 (P.L. 99-570) were amended and renamed the Edward Byrne Memorial State
and Local Law Enforcement Assistance Program by the Anti-Drug Abuse Act of
1988 (P.L. 100-690). The Byrne grant program is codified at 42 U.S.C.
3750-3766b. The Byrne program is 1 of approximately 20 programs administered
by BJA. The program funds formula grants, which are awarded directly to
state governments, and discretionary grants for 28 legislatively authorized
purposes.

VAWO was created in 1995 to carry out certain programs created under the
Violence Against Women Act of 1994.5 The Victims of Trafficking and Violence
Prevention Act of 2000 reauthorized most of the existing VAWO programs and
added new programs.6 VAWO programs seek to improve criminal justice system
responses to domestic violence, sexual assault, and stalking by providing
support for law enforcement, prosecution, courts, and victim advocacy
programs across the country. During fiscal years 1995 through 2001, VAWO's
five discretionary grant programs that were subject to program evaluation
were (1) STOP (Services, Training, Officers, and Prosecutors) Violence
Against Indian Women Discretionary Grants, (2) Grants to Encourage Arrest
Policies, (3) Rural Domestic Violence and Child Victimization Enforcement
Grants, (4) Domestic Violence Victims' Civil Legal Assistance Grants, and
(5) Grants to Combat Violent Crimes Against Women on Campuses. During fiscal
years 1995 through 2001, about $505 million was appropriated to these
discretionary grant programs.7

As already mentioned, NIJ is the principal research and development agency
within OJP, and its duties include developing, conducting, directing, and
supervising Byrne and VAWO discretionary grant program evaluations. Under 42
U.S.C. 3766, NIJ is required to "conduct a reasonable number of
comprehensive evaluations" of the Byrne discretionary grant program. In
selecting programs for review under section 3766, NIJ is to consider new and
innovative approaches, program costs, potential for replication in other
areas, and the extent of public awareness and community involvement.
According to NIJ officials, the implementation of various types of
evaluations, including process and impact evaluations, fulfills this
legislative requirement. Although legislation creating VAWO does not require
evaluations of the VAWO discretionary grant programs, Justice's annual
appropriations for VAWO during fiscal years 1998 through 2002 included
monies for NIJ research and evaluations of violence against

5Title IV of the Violent Crime Control and Enforcement Act of 1994 (P.L.
103-322).

6P.L. 106-386.

7This amount does not include about $19 million for the Sex Offender
Management Training discretionary program. During the period of our review,
NIJ did not evaluate this training program. From fiscal year 1995 through
fiscal year 1999 this program was administered by VAWO. As of fiscal year
2000, responsibility for the program was shifted to OJP's Corrections
Program Office.

women.8 In addition, Justice has promulgated regulations requiring that NIJ
conduct national evaluations of two of VAWO's discretionary grant programs.9
As with the Byrne discretionary programs, NIJ is not required by statute or
Justice regulation to conduct specific types of program evaluations, such as
impact or process evaluations.

The Director of NIJ is responsible for making the final decision on which
Byrne and VAWO discretionary grant programs to evaluate; this decision is
based on the work of NIJ staff in coordination with Byrne or VAWO program
officials. Once the decision has been made to evaluate a particular program,
NIJ issues a solicitation for proposals for grant funding from potential
evaluators. When applications or proposals are received, an external peer
review panel comprising members of the research and relevant practitioner
communities is convened. Peer review panels identify the strengths,
weaknesses, and potential methodologies to be derived from competing
proposals. When developing their consensus reviews, peer review panels are
to consider the quality and technical merit of the proposal; the likelihood
that grant objectives will be met; the capabilities, demonstrated
productivity, and experience of the evaluators; and budget constraints. Each
written consensus review is reviewed and discussed with partnership agency
representatives (e.g., staff from BJA or VAWO). These internal staff reviews
and discussions are led by NIJ's Director of the Office of Research and
Evaluation who then presents the peer review consensus reviews, along with
agency and partner agency input, to the NIJ Director for consideration and
final grant award decisions. The NIJ Director makes the final decision
regarding which application to fund.

8In fiscal year 1998, Congress appropriated $7 million, and in subsequent
fiscal years 1999 through 2002, Congress appropriated $5.2 million annually
of VAWO funds for NIJ research and evaluations of violence against women.
According to NIJ's Director, Office of Research and Evaluation, none of
these funds were used to award evaluation grants of the five VAWO
discretionary grant programs discussed in this report.

928 C.F.R. 90.58 and 90.65.

Scope and Methodology

To meet our objectives, we conducted our work at OJP, BJA, VAWO, and NIJ
headquarters in Washington, D.C. We reviewed applicable laws and
regulations, guidelines, reports, and testimony associated with Byrne and
VAWO discretionary grant programs and evaluation activities. In addition, we
interviewed responsible OJP, NIJ, BJA, and VAWO officials regarding program
evaluations of discretionary grants. As agreed with your offices, we focused
on program evaluation activities associated with the Byrne and VAWO
discretionary grant programs. In particular, we focused on the program
evaluations of discretionary grants that were funded during fiscal years
1995 through 2001.

To address our first objective, regarding the number, type, status of
completion, and award amount of Byrne and VAWO discretionary grant program
evaluations, we interviewed NIJ, BJA, and VAWO officials and obtained
information on Byrne and VAWO discretionary grant programs and program
evaluations. Because NIJ is responsible for carrying out program evaluations
of Byrne and VAWO discretionary grant programs, we also obtained and
analyzed NIJ data about specific Byrne and VAWO discretionary grant program
evaluations, including information on the number of evaluations as well as
the type, cost, source of funding, and stages of implementation of each
evaluation for fiscal years 1995 through 2001. We did not independently
verify the accuracy or completeness of the data that NIJ provided.

To address the second objective, regarding the methodological rigor of the
impact evaluation studies of Byrne and VAWO discretionary grant programs
during fiscal years 1995 through 2001, we initially identified the impact
evaluations from the universe of program evaluations specified by NIJ. We
excluded from our analysis any impact evaluations that were in the formative
stage of development-that is, the application had been awarded but the
methodological design was not yet fully developed. As a result, we reviewed
four program evaluations.

For the four impact evaluations that we reviewed, we asked NIJ to provide
any documentation relevant to the design and implementation of the impact
evaluation methodologies, such as the application solicitation, the
grantee's initial and supplemental applications, progress notes, interim
reports, requested methodological changes, and any final reports that may
have become available during the data collection period. We also provided
NIJ with a list of methodological issues to be considered in our review and
requested them to submit any additional documentation that addressed these
issues. We used a data collection instrument to obtain information
systematically about each program being evaluated and about the features

of the evaluation methodology. We based our data collection and assessments
on generally accepted social science standards. We examined such factors as
whether evaluation data were collected before and after program
implementation; how program effects were isolated (i.e., the use of
nonprogram participant comparison groups or statistical controls); and the
appropriateness of sampling and outcome measures. Two of our senior social
scientists with training and experience in evaluation research and
methodology separately reviewed the evaluation documents and developed their
own assessments before meeting jointly to discuss the findings and
implications. This was done to promote a grant evaluation review process
that was both independent and objective.

To obtain information on the approaches that BJA, VAWO, and NIJ used to
disseminate program evaluation results, we requested and reviewed, if
available, relevant handbooks and guidelines on information dissemination,
including, for example, NIJ's guidelines. We also reviewed BJA, VAWO, and
NIJ's available print and electronic products as related to their proven
programs and evaluations, including two NIJ publications about Byrne
discretionary programs and their evaluation methodologies and results.

We conducted our work between February 2001 and December 2001 in accordance
with generally accepted government auditing standards.

We requested comments from Justice on a draft of this report in January
2002. The comments are discussed near the end of this letter and are
reprinted as appendix III.

Number, Type, Status of Completion, and Award Amount of Byrne and VAWO
Discretionary Grant Program Evaluations

During fiscal years 1995 through 2001, NIJ awarded about $6 million to carry
out five Byrne and five VAWO discretionary grant program evaluations. NIJ
awarded evaluation grants using mostly funds transferred from BJA and VAWO.
Specifically, of the approximately $1.9 million awarded for one impact and
four process evaluations of the Byrne discretionary program, NIJ contributed
about $299,000 (16 percent) and BJA contributed about $1.6 million (84
percent). VAWO provided all of the funding (about $4 million) to NIJ for all
program evaluations of five VAWO discretionary grant programs. According to
NIJ, the five VAWO program evaluations included both impact and process
evaluations. Our review of information provided by NIJ showed that 6 of the
10 program evaluations-all 5 VAWO evaluations and 1 Byrne
evaluation-included impact evaluations. The remaining four Byrne evaluations
were exclusively process evaluations that measured the extent to which the

programs were working as intended. 10 As of December 2001, only one of these
evaluations, the impact evaluation of the Byrne CAR Program, had been
completed. The remaining evaluations were in various stages of
implementation. Table 1 lists each of the five Byrne program evaluations and
shows whether it was a process or an impact evaluation, its stage of
implementation, the amount awarded during fiscal years 1995 through 2001,
and the total amount awarded since the evaluation was funded.

       Table 1: NIJ Evaluations of Byrne Discretionary Grant Programs

Status/stage of completion as of 12/01

Amount awarded for evaluation since FY 1995

Total amount awardeda

                        Discretionary grant programs

Impact evaluation Process evaluation

        Children at Risk Program          Y  N     Completed     $452,780    $1,034,732
        Comprehensive Communities         N  Y   Final report     482,762       782,294
                Programb                          under peer
                                                    review
   Tribal Strategies Against Violence     N  Y   Final report     280,169       280,169
               Initiative                         under peer
                                                    review
Boston's Safe Neighborhood           N   Y     Evaluation
Initiative                                      under way
Red Hook Community Court
Program

                              274,223 274,223

N Y Evaluation under way

                              374,981 374,981

           Total evaluation dollars awarded $1,864,915 $2,746,399

aTwo of the Byrne program evaluation grants began prior to FY 1995. The
evaluation of the Children at Risk Program was initially awarded in FY 1992
for $581,952. The evaluation of the Comprehensive Communities Program was
initially awarded in FY 1994 for $299,532.

bNIJ documents identify the Byrne Comprehensive Communities Program
evaluation as a process evaluation, but NIJ officials indicated that some
elements of the evaluation were impact-oriented. Our review of the
evaluation subsequently showed that this was a process evaluation because it
addressed whether the program was working as intended rather than
identifying a net effect of the program.

Source: GAO analysis of data provided by the National Institute of Justice

Table 2 lists each of the five VAWO program evaluations and shows that it
was both a process and an impact evaluation, its stage of implementation,
and the amount awarded during fiscal years 1995 through 2001, which is the
total amount awarded.

10NIJ documents identify the Byrne Comprehensive Communities Program
evaluation as a process evaluation, but NIJ officials indicated that some
elements of the evaluation were impact oriented. Our review of the
evaluation subsequently showed that this was a process evaluation, because
it addressed whether the program was working as intended rather than
identifying the net effect of the program.

    Table 2: NIJ Evaluations of VAWO Discretionary Grant Programs Process
                                 evaluation

Status/stage of completion as of 12/01

                             Impact evaluation

Total amount awarded for evaluationDiscretionary grant programs

STOP Violence Against Indian Women Discretionary Grants

Y Y Evaluation under way

                                  $468,552

Grants to Encourage Arrests Policies Y Y Evaluation under way

                                 1,130,574

Rural Domestic Violence and Child Victimization Enforcement Grants

Y Y Evaluation under way

                                  719,949

Domestic Violence Victims' Civil Legal Assistance Grants

                        Y Y Formative stage 800,154

Grants to Combat Crimes Against Women on Campuses

                        Y Y Formative stage 849,833

                 Total Evaluation Award Amounts $3,969,032

Note: We did not review the program evaluation methodologies of the Civil
Legal Assistance Program and Crimes Against Women on Campus Program because
they were in their formative stage of development. That is, the application
was awarded but the methodological design was not yet fully developed.

Source: GAO analysis of data provided by the National Institute of Justice.

Our review showed that methodological problems have adversely affected three
of the four impact evaluations that have progressed beyond the formative
stage. All three VAWO evaluations that we reviewed demonstrated a variety of
methodological limitations, raising concerns as to whether the evaluations
will produce definitive results. The one Byrne evaluation was well designed
and used appropriate data collection and analytic methods. We recognize that
impact evaluations, such as the type that NIJ is managing, can encounter
difficult design and implementation issues. In the three VAWO evaluations
that we reviewed, program variation across sites has added to the complexity
of designing the evaluations. Sites could not be shown to be representative
of the programs or of particular elements of these programs, thereby
limiting the ability to generalize results; the lack of comparison groups
hinders the ability to minimize the effects of factors external to the
program. Furthermore, data collection and analytical problems compromise the
ability of evaluators to draw appropriate conclusions from the results. In
addition, peer review committees found methodological problems in two of the
three VAWO evaluations that we considered.

The four program evaluations are multiyear, multisite impact evaluations.
Some program evaluations used a sample of grants, while others used the
entire universe of grants. For example, the Grants to Encourage Arrests
Policies Program used 6 of the original 130 grantee sites. In contrast, in
the

Methodological Problems Have Adversely Affected Three of Four Impact
Evaluations

Byrne Children at Risk impact evaluation, all five sites participated. As of
December 2001, NIJ had already received the impact findings from the Byrne
Children at Risk Program evaluation but had not received impact findings
from the VAWO discretionary grant program evaluations.

Impact Evaluations Are Difficult to Successfully Design and Implement

An impact evaluation is an inherently difficult task, since the objective is
to isolate the effects of a particular program or factor from all other
potential contributing programs or factors that could also effect change.
Given that the Byrne and VAWO programs are operating in an ever changing,
complex environment, measuring the impact of these specific Byrne and VAWO
programs can be arduous. For example, in the evaluation of VAWO's Rural
Domestic Violence Program, the evaluator's responsibility is to demonstrate
how the program affected the lives of domestic violence victims and the
criminal justice system. Several other programs or factors besides the Rural
Domestic Violence Program may be accounting for all or part of the observed
changes in victims' lives and the criminal justice system (e.g., a
co-occurring program with similar objectives, new legislation, a local
economic downturn, an alcohol abuse treatment program). Distinguishing the
effects of the Rural Domestic Violence Program requires use of a rigorous
methodological design.

Project Variation within the VAWO Programs Complicates Evaluation Design and
Implementation

All three VAWO programs permitted their grantees broad flexibility in the
development of their projects to match the needs of their local communities.
According to the Assistant Attorney General, this variation in projects is
consistent with the intent of the programs' authorizing legislation. We
recognize that the authorizing legislation provides VAWO the flexibility in
designing these programs. Although this flexibility may make sense from a
program perspective, the resulting project variation makes it more difficult
to design and implement a definitive impact evaluation of the program.
Instead of assessing a single, homogeneous program with multiple grantees,
the evaluation must assess multiple configurations of a program, thereby
making it difficult to generalize about the entire program. Although all of
the grantees' projects under each program being evaluated are intended to
achieve the same or similar goals, an aggregate analysis could mask the
differences in effectiveness among individual projects and thus not result
in information about which configurations of projects work and which do not.

The three VAWO programs exemplify this situation. The Arrest Policies
Program provided grantees with the flexibility to develop their respective
projects within six purpose areas: implementing mandatory arrest or

proarrest programs and policies in police departments, tracking domestic
violence cases, centralizing and coordinating police domestic violence
operations, coordinating computer tracking systems, strengthening legal
advocacy services, and educating judges and others about how to handle
domestic violence cases. Likewise, the STOP Grants Program encouraged tribal
governments to develop and implement culture-specific strategies for
responding to violent crimes against Indian women and provide appropriate
services for those who are victims of domestic abuse, sexual assault, and
stalking. Finally, the Rural Domestic Violence Program was designed to
provide sites with the flexibility to develop projects, based on need, with
respect to the early identification of, intervention in, and prevention of
woman battering and child victimization; with respect to increases in victim
safety and access to services; with respect to enhancement of the
investigation and prosecution of crimes of domestic violence; and with
respect to the development of innovative, comprehensive strategies for
fostering community awareness and prevention of domestic abuse. Because
participating grant sites emphasized different project configurations, the
resulting evaluation may not provide information that could be generalized
to a broader implementation of the program.

Site Selection Not Shown to be Representative within the VAWO Programs

The sites participating in the three VAWO evaluations were not shown to be
representative of their programs. Various techniques are available to help
evaluators choose representative sites and representative participants
within those sites. Random sampling of site and participant selection are
ideal, but when this is not feasible, other purposeful sampling methods can
be used to help approximate the selection of an appropriate sample (e.g.,
choosing the sample in such proportions that it reflects the larger
population--stratification). At a minimum, purposeful selection can ensure
the inclusion of a range of relevant sites.

As discussed earlier, in the case of the Arrest Policies Program, six
purpose areas were identified in the grant solicitation. The six grantees
chosen for participation in the evaluation were not however, selected on the
basis of their representativeness of the six purpose areas or the program as
a whole. Rather, they were selected on the basis of factors related solely
to program "stability;" that is; they were considered likely to receive
local funding after the conclusion of federal grant funding, and key
personnel would continue to participate in the coordinated program effort.
Similarly, the 10 Rural Domestic Violence impact evaluation grantees were
not selected for participation on the basis of program representativeness or
the specific purpose areas discussed earlier. Rather, sites were selected

by the grant evaluator on the basis of "feasibility"; specifically, whether
the site would be among those participants equipped to conduct an
evaluation.11 Similarly, the STOP Violence Against Indian Women Program
evaluation used 3 of the original 14 project sites for a longitudinal study;
these were not shown to be representative of the sites in the overall
program. For another phase of the evaluation, the principal investigator
indicated that grantee sites were selected to be geographically
representative of American Indian communities. While this methodology
provides for inclusion of a diversity of Indian tribes in the sample from
across the country, geography as a sole criterion does not guarantee
representativeness in relation to many other factors.

Lack of Comparison Groups in the VAWO Evaluations Hinders Ability to Isolate
the Effects of the Programs

Each of the three VAWO evaluations was designed without comparison groups-a
factor that hinders the evaluator's ability to isolate and minimize external
factors that could influence the results of the study. Use of comparison
groups is a standard practice employed by evaluators to help determine
whether differences between baseline and follow-up results are due to the
program under consideration or to some other programs or external factors.
For example, as we reported in 1997,12 to determine whether a drug court
program has been effective in reducing criminal recidivism and drug relapse,
it is not sufficient to merely determine whether those participating in the
drug court program show changes in recidivism and relapse rates. Changes in
recidivism and relapse variables between baseline and program completion
could be due to other external factors, irrespective of the drug court
program (e.g., the state may have developed harsher sentencing procedures
for those failing to meet drug court objectives). If, however, the drug
court participant group is matched at baseline against another set of
individuals, "the comparison group" who are experiencing similar life
circumstances but who do not qualify for drug court participation (e.g.,
because of area of residence), then the comparison group can help in
isolating the effects of the drug court program. The contrasting of the two
groups in relation to recidivism and relapse can provide an approximate
measure of the program's impact.

11According  to NIJ officials, sites would also  be selected based on VAWO's
recommendations.

12 U.S.  General  Accounting  Office,   Drug  Courts:  Overview  of  Growth,
Characteristics, and Results, GAO/GGD-97-106, (Washington, D.C.: July 1997).

All three VAWO program impact evaluations lacked comparison groups. One
issue addressed in the Arrest Policies Program evaluation, for example, was
the impact of the program on the safety and protection of the domestic
violence victim. The absence of a comparison group, however, makes it
difficult to firmly conclude that change in the safety and protection of
participating domestic abuse victims is due to the Arrest Policies Program
and not to some other external factors operating in the environment (e.g.,
economic changes, nonfederal programs such as safe houses for domestically
abused women, and church-run support programs). Instead of using comparison
groups, the Arrest Policies Program evaluation sought to eliminate potential
competing external factors by collecting and analyzing extensive historical
and interview data from subjects and by conducting cross-site comparisons;
the latter method proved unfeasible.

The STOP Violence Against Indian Women Discretionary Grant Program has
sought in part, to reduce violent crimes against Indian women by changing
professional staff attitudes and behaviors. To do this, some grantees
created and developed domestic violence training services for professional
staff participating in site activities. Without comparison groups, however,
assessing the effect of the STOP training programs is difficult. Attitudes
and behaviors may change for myriad reasons unrelated to professional
training development initiatives. If a treatment group of professional staff
receiving the STOP training had been matched with a comparison group of
professional staff that was similar in all ways except receipt of training,
there would be greater confidence that positive change could be attributed
to the STOP Program. Similarly, the lack of comparison groups in the Rural
Domestic Violence evaluation makes it difficult to conclude that a reduction
in violence against women and children in rural areas can be attributed
entirely, or in part, to the Rural Domestic program. Other external factors
may be operating.

VAWO Data Collection and Analytical Problems Evident during Grant
Implementation

All three VAWO impact evaluations involved data collection and analytical
problems that may affect the validity of the findings and conclusions. For
example, we received documentation from NIJ on the STOP Grant Program for
Reducing Violence Against Indian Women showing that only 43 percent of 127
grantees returned a mail survey.13 In addition, only 25 percent of 127
tribes provided victim outcome homicide and hospitalization rates-far less
than the percentage needed to draw broad-based conclusions about the
intended goal of assessing victim well being. In the Arrest Policies
evaluation, NIJ reported that the evaluators experienced difficulty in
collecting pre-grant baseline data from multiple sites and the quality of
the data was oftentimes inadequate, which hindered their ability to
statistically analyze change over time. In addition, evaluators were
hindered in several work areas by lack of automated data systems; data were
missing, lost, or unavailable; and the ability to conduct detailed analyses
of the outcome data was sometimes limited. For the Rural Domestic Violence
evaluation, evaluators proposed using some variables (e.g., number and type
of awareness messages disseminated to the community each month,
identification of barriers to meeting the needs of women and children, and
number of police officers who complete a training program on domestic
violence) that are normally considered to relate more to a process
evaluation than an impact evaluation. NIJ noted that outcome measurement
indicators varied by site, complicating the ability to draw generalizations.
NIJ further indicated that the evaluation team did not collect baseline data
prior to the start of the program, making it difficult to identify change
resulting from the program.

VAWO Peer Review Committees Expressed Concerns about Two Evaluations

NIJ does not require applicants to use particular evaluation
methodologies.14 NIJ employs peer review committees in deciding which
evaluation proposals to fund. The peer review committees expressed concerns
about two of the three VAWO program evaluation proposals (i.e.,

13Guidance from the Office of Management and Budget indicated that certain
statistical errors begin to arise when response rates are in the 50- to
75-percent range. Below 50 percent, these errors rise rapidly, and they are
not significant when the response rates are 75 percent or higher. Office of
Management and Budget, Office of Information and Regulatory Affairs, The
Paperwork Reduction Act of 1995: Implementing Guidance for OMB Review of
Agency Information Collection. [Washington, D.C.: 1999], 127 and 128.

14NIJ's Director of Planning and Management indicated several reasons for
the agency's decision not to specify particular evaluation methodologies,
including the need to maintain the grantee's independence, objectivity,
innovation, and methodological flexibility under varying situations.

those for the Arrest Policies and Rural Domestic Violence programs) that
were subsequently funded by NIJ. Whereas NIJ funded the Arrest Policies
evaluation as a grant, NIJ funded the Rural Domestic Violence evaluation as
a cooperative agreement so that NIJ could provide substantial involvement in
conducting the evaluation. 15

A peer-review panel and NIJ raised several concerns about the Arrest
Policies Program evaluation proposal. These concerns included issues related
to site selection, victim interviewee selection and retention in the sample,
and the need for additional impact measures and control variables. The grant
applicant's responses to these issues did not remove concerns about the
methodological rigor of the application, thus calling into question the
ability of the grantee to assess the impact of the Arrest Policies Program.
For example, the grantee stated that victim interviewee selection was to be
conducted through a quota process and that the sampling would vary by site.
This would not allow the evaluators to generalize program results. Also, the
evaluators said that they would study communities at different levels of
"coordination" when comparison groups were not feasible, but they did not
adequately explain (1) how the various levels of coordination would be
measured, (2) the procedures used to select the communities compared, and
(3) the benefits of using this method as a replacement for comparison
groups. NIJ subsequently funded this evaluation, and it is still in
progress.

A peer review committee for the Rural Domestic Violence and Child
Victimization Enforcement Grant Program evaluation also expressed concerns
about whether the design of the evaluation application, as proposed, would
demonstrate whether the program was working. In its consensus review notes,
the peer review committee indicated that the "ability to make
generalizations about what works and does not work will be limited." The
peer review committee also warned of outside factors (e.g., unavailability
of data, inaccessibility of domestic violence victims) that could imperil
the evaluation efforts of the applicant. Based on the peer review
committee's input, NIJ issued the following statement to the applicant: "As
a national evaluation of a major programmatic effort we hope to have a
research design and products on what is working, what is not working, and
why. We are not sure that the proposed design will get us to that point." We
reviewed the grant applicant's response to NIJ's concern

15According to OJP, a cooperative agreement includes substantial involvement
by the agency.

in its application addendum and found that the overall methodological design
was still not discussed in sufficient detail or depth to determine whether
the program was working. Although the Deputy Director of NIJ's Office of
Research and Evaluation asserted that this initial application was only for
process evaluation funding, our review of available documents showed that
the applicant had provided substantial information about both the process
and impact evaluation methodologies in the application and addendum. We
believe that the methodological rigor of the addendum was not substantially
improved over that of the original application. The Deputy Director told us
that, given the "daunting challenge faced by the evaluator," NIJ decided to
award the grant as a cooperative agreement. Under this arrangement, NIJ was
to have substantial involvement in helping the grantee conduct the program
evaluation. The results of that evaluation have not yet been submitted. The
evaluator's draft final report is expected no earlier than April 2002.

Byrne Evaluation Was Successfully Designed and Implemented

In contrast to the three VAWO impact evaluations, the Byrne impact
evaluation employed methodological design and implementation procedures that
met a high standard of methodological rigor, fulfilling each of the criteria
indicated above. In part, this may reflect the fact that Byrne's CAR
demonstration program, unlike the VAWO programs, was according to the
Assistant Attorney General, intended to test a research hypothesis, and the
evaluation was designed accordingly. CAR provided participants with the
opportunity to use a limited number of program services (e.g., family
services, education services, after-school activities) that were
theoretically related to the impact variables and the prevention and
reduction of drug use and delinquency. As a result, the evaluation was not
complicated by project heterogeneity. All five grantees participated in the
evaluation. High-risk youths within those projects were randomly selected
from targeted neighborhood schools, providing student representation.
Additionally, CAR evaluators chose a matched comparison group of youths with
similar life circumstances (e.g., living in distressed neighborhoods and
exposed to similar school and family risk factors) and without access to the
CAR Program. Finally, no significant data collection implementation problems
were associated with the CAR Program. The data were collected at multiple
points in time from youths (at baseline, at completion of program, and at
one year follow-up) and their caregivers (at baseline and at completion of
program). Self-reported findings from youths were supplemented by the
collection of more objective data from school, police, and court records on
an annual basis, and rigorous test procedures were used to determine whether
changes over time were statistically significant.

Conclusions

Recommendations for Executive Action

Additionally, CAR's impact evaluation used control groups, a
methodologically rigorous technique not used in the three VAWO evaluations.
To further eliminate the effects of external factors, youths in the targeted
neighborhood schools were randomly assigned either to the group receiving
the CAR Program or to a control group that did not participate in the
program. Since the CAR Program group made significant gains over the
same-school group and the matched comparison group not participating in the
program, there was good reason to conclude that the CAR Program was having a
beneficial effect on the targeted audience.

Appendix I provides summaries of the four evaluations.

Despite great interest in assessing results of OJP's discretionary grant
programs, it can be extremely difficult to design and execute evaluations
that will provide definitive information. Our in-depth review of one Byrne
and three VAWO impact evaluations that have received funding since fiscal
year 1995 has shown that, in some cases, the flexibility that can be
beneficial to grantees in tailoring programs to meet their communities'
needs has added to the complexities of designing impact evaluations that
will result in valid findings. Furthermore, the lack of site
representativeness, appropriate comparison groups, and problems in data
collection and analysis may compromise the reliability and validity of some
of these evaluations. We recognize that not all evaluation issues that can
compromise results are easily resolvable, including issues involving
comparison groups and data collection. To the extent that methodological
design and implementation issues can be overcome, however, the validity of
the evaluation results will be enhanced. NIJ spends millions of dollars
annually to evaluate OJP grant programs. More up-front attention to the
methodological rigor of these evaluations will increase the likelihood that
they will produce meaningful results for policymakers. Unfortunately, the
problematic evaluation grants that we reviewed are too far along to be
radically changed. However, two of the VAWO evaluation grants are still in
the formative stage; more NIJ attention to their methodologies now can
better ensure useable results.

We recommend that the Attorney General instruct the Director of NIJ to
assess the two VAWO impact evaluations that are in the formative stage to
address any potential methodological design and implementation problems and,
on the basis of that assessment, initiate any needed interventions to help
ensure that the evaluations produce definitive results. We further recommend
that the Attorney General instruct the

Agency Comments
and Our Evaluation

Director of NIJ to assess its evaluation process with the purpose of
developing approaches to ensure that future impact evaluation studies are
effectively designed and implemented so as to produce definitive results.

We provided a copy of a draft of this report to the Attorney General for
review and comment. In a February 13, 2002, letter, the Assistant Attorney
General commented on the draft. Her comments are summarized below and
presented in their entirety in appendix III.

The Assistant Attorney General agreed with the substance of our
recommendations and said that NIJ has begun, or plans to take steps, to
address them. Although it is still too early to tell whether NIJ's actions
will be effective in preventing or resolving the problems we identified,
they appear to be steps in the right direction. With regard to our first
recommendation-that NIJ assess the two VAWO impact evaluations in the
formative stage to address any potential design and implementation problems
and initiate any needed intervention to help ensure definitive results-the
Assistant Attorney General noted that NIJ has begun work to ensure that
these projects will provide the most useful information possible. She said
that for the Crimes Against Women on Campus Program evaluation, NIJ is
considering whether it will be possible to conduct an impact evaluation and,
if so, how it can enhance its methodological rigor with the resources
available. For the Civil Legal Assistance Program evaluation, the Assistant
Attorney General said that NIJ is working with the grantee to review site
selection procedures for the second phase of the study to enhance the
representativeness of sites. The Assistant Attorney General was silent about
any additional steps that NIJ would take during the later stages of the
Civil Legal Assistance Program process evaluation to ensure the
methodological rigor of the impact phase of the study. However, it seems
likely that as the process evaluation phase of the study continues, NIJ may
be able to take advantage of additional opportunities to address any
potential design and implementation problems.

With regard to our second recommendation-that NIJ assess its evaluation
process to develop approaches to ensure that future evaluation studies are
effectively designed and implemented to produce definitive results-the
Assistant Attorney General stated that OJP has made program evaluation,
including impact evaluations of federally funded programs, a high priority.
The Assistant Attorney General said that NIJ has already launched an
examination of NIJ's evaluation process. She also noted that, as part of its
reorganization, OJP plans to measurably strengthen NIJ's capacity to manage
impact evaluations with the goal of making them more useful for

Congress and others. She noted as an example that OJP and NIJ are building
measurement requirements into grants at the outset, requiring potential
grantees to collect baseline data and track the follow-up data through the
life of the grant. We have not examined OJP's plans for reorganizing, nor do
we have a basis for determining whether OJP's plans regarding NIJ would
strengthen NIJ's capacity to manage evaluations. However, we believe that
NIJ and its key stakeholders, such as Congress and the research community,
would be well served if NIJ were to assess what additional actions it could
take to strengthen its management of impact evaluations regardless of any
reorganization plans.

In her letter, the Assistant Attorney General pointed out that the report
accurately describes many of the challenges facing evaluators when
conducting research in the complex environment of criminal justice programs
and interventions. However, she stated that the report could have gone
further in acknowledging these challenges. The Assistant Attorney General
also stated that the report contrasts the Byrne evaluation with the three
VAWO evaluations and obscures important programmatic differences that affect
an evaluator's ability to achieve "GAO's conditions for methodological
rigor." She pointed out that the Byrne CAR Program was intended to test a
research hypothesis and that the evaluation was designed accordingly, i.e.,
the availability of baseline data were ensured; randomization of effects
were stipulated as a precondition of participation; and outcome measures
were determined in advance on the basis of the theories to be tested. She
further stated that, in contrast, all of the VAWO programs were (1) highly
flexible funding streams, in keeping with the intention of Congress, that
resulted in substantial heterogeneity at the local level and (2) well into
implementation before the evaluation started. The Assistant Attorney General
went on to say that it is OJP's belief that evaluations under less than
optimal conditions can provide valuable information about the likely impact
of a program, even though the conditions for methodological strategies and
overall rigor of the CAR evaluation were not available.

We recognize that there are substantive differences in the intent,
structure, and design of the various discretionary grant programs managed by
OJP and its bureaus and offices. And, as stated numerous times in our
report, we acknowledge not only that impact evaluation can be an inherently
difficult and challenging task but also that measuring the impact of these
specific Byrne and VAWO programs can be arduous, given that they are
operating in an ever changing, complex environment. We agree that not all
evaluation issues that can compromise results are easily resolvable, but we
firmly believe that, with more up-front attention to design and

implementation issues, there is a greater likelihood that NIJ evaluations
will provide meaningful results for policymakers. Absent this up-front
attention, questions arise as to whether NIJ is (1) positioned to provide
the definitive results expected from an impact evaluation and (2) making
sound investments given the millions of dollars spent on these evaluations.

The Assistant Attorney General also commented that although our report
discussed "generally accepted social science standards," it did not specify
the document that articulates these standards or describe our elements of
rigor. As a result, the Assistant Attorney General said, OJP had to infer
that six elements had to be met to achieve what "GAO believes" is necessary
to "have a rigorous impact evaluation." Specifically, she said that she
would infer that, for an impact evaluation to be rigorous would require (1)
selection of homogenous programs, (2) random or stratified site sampling
procedures (or selection of all sites), (3) use of comparison groups, (4)
high response rates, (5) available and relevant automated data systems that
will furnish complete and accurate data to evaluators in a timely manner,
and (6) funding sufficient to accomplish all of the above. Furthermore, the
Attorney General said that it is rare to encounter all of these conditions
or be in a position to engineer all of these conditions simultaneously; and
when all of these conditions are present, the evaluation would be rigorous.
She also stated that it is possible to glean useful, if not conclusive,
evidence of the impact of a program from an evaluation that does not rise to
the standard recommended by GAO because of the unavoidable absence of "one
or more elements."

We agree that our report did not specify particular documents that
articulate generally accepted social science standards. However, the
standards that we applied are well defined in scientific literature.16 All
assessments of the impact evaluations we reviewed were completed by social
scientists with extensive experience in evaluation research. Throughout our
report, we explain our rationale and the criteria we used in measuring the
methodological rigor of NIJ's impact evaluations. Furthermore, our report
does not suggest that a particular standard or set

16Donald T. Campbell and Julian C. Stanley, Experimental and
Quasi-Experimental Designs for Research (Chicago: Rand McNally & Company,
1963); Carol H. Weiss, Evaluation Research: Methods for Assessing Program
Effectiveness (Englewood Cliffs: Prentice-Hall, Inc., 1972); Edward A.
Suchman, Evaluative Research: Principles and Practice in Public Service &
Social Action Programs (New York: Russell Safe Foundation, 1967); U.S.
General Accounting Office, Designing Evaluations, GAO/PEMD-10.1.4
(Washington, D.C.: May 1991).

of standards is necessary to achieve rigor, nor does it suggest that other
types of evaluations, such as comprehensive process evaluations, are any
less useful in providing information on how a program is operating. In this
context, it is important to point out that the scope of our work covered
impact evaluations of Byrne and VAWO discretionary grant programs- those
designed to assess the net effect of a program by comparing program outcomes
with an estimate of what would have happened in the absence of the program.

We differ with the Assistant Attorney General with respect to the six
elements cited as necessary elements for conducting an impact evaluation.
Contrary to the Assistant Attorney General's assertion, our report did not
state that a single homogeneous program is a necessary element for
conducting a rigorous impact evaluation. Rather, we pointed out that
heterogeneity or program variation is a challenge that adds to the
complexity of designing an evaluation. In addition, contrary to her
assertion, the report did not assert that random sampling or stratification
was a necessary element for conducting a rigorous evaluation; instead it
stated that when random sampling is not feasible, other purposeful sampling
methods can be used. With regard to comparison groups, the Assistant
Attorney General's letter asserted that GAO standards required using groups
that do not receive program benefits as a basis of comparison with those
that do receive such benefits. In fact, we believe that the validity of
evaluation results can be enhanced through establishing and tracking
comparison groups. If other ways exist to effectively isolate the impacts of
a program, comparison groups may not be needed. However, we saw no evidence
that other methods were effectively used in the VAWO impact evaluations we
assessed.

The Assistant Attorney General also suggested that we used a 75 percent or
greater response rate for evaluation surveys as a standard of rigor. In
fact, we did not-we simply pointed out that NIJ documents showed a 43
percent response rate on one of the STOP Grant Program evaluation surveys.
This is below OMB's threshold response rate level-the level below which OMB
particularly believes nonresponse bias and statistical problems could affect
surveys. Given OMB guidance, serious questions could be raised about program
conclusions drawn from the results of a survey with a 43 percent response
rate. In addition, the Assistant Attorney General suggested that, by GAO
standards, she would have to require state, local, or tribal government
officials to furnish complete and accurate data in a timely manner. In fact,
our report only points out that NIJ reported that evaluators were hindered
in carrying out evaluations because of the lack of automated data systems or
because data were

missing, lost, or unavailable-again, challenges to achieving
methodologically rigorous evaluations that could produce meaningful and
definitive results.

Finally, the Assistant Attorney General's letter commented that one of the
elements needed to meet "all of GAO's conditions" of methodological rigor is
sufficient funding. She stated that more rigorous impact evaluations cost
more than those that provide less scientific findings, and she said that OJP
is examining the issue of how to finance effective impact evaluations. We
did not assess whether funding is sufficient to conduct impact evaluations,
but we recognize that designing effective and rigorous impact evaluations
can be expensive-a condition that could affect the number of impact
evaluations conducted. However, we continue to believe that with more
up-front attention to the rigor of ongoing and future evaluations, NIJ can
increase the likelihood of conducting impact evaluations that produce
meaningful and definitive results.

In addition to the above comments, the Assistant Attorney General made a
number of suggestions related to topics in this report. We have included the
Assistant Attorney General's suggestions in the report, where appropriate.
Also, the Assistant Attorney General provided other comments in response to
which we did not make changes. See appendix III for a more detailed
discussion of the Assistant Attorney General's comments.

We are sending copies of this report to the Chairman and the Ranking
Minority Member of the Senate Judiciary Committee; to the Chairman and
Ranking Minority Member of the House Judiciary Committee; to the Chairman
and Ranking Minority Member of the Subcommittee on Crime, House Committee on
the Judiciary; to the Chairman and the Ranking Minority Member of the House
Committee on Education and the Workforce; to the Attorney General; to the
OJP Assistant Attorney General; to the NIJ Director; to the BJA Director; to
the VAWO Director; and to the Director, Office of Management and Budget. We
will also make copies available to others on request.

If you or your staff have any questions about this report, please contact
John F. Mortin or me at (202) 512-8777. Key contributors to this report are
acknowledged in appendix IV.

Laurie E. Ekstrand Director, Justice Issues

Appendix I: Summaries of the Impact Evaluations of Byrne and VAWO Programs

Evaluation  National Evaluation  of  the Rural  Domestic Violence  and Child
Victimization Grant Program

Principal investigator COSMOS Corporation

Program evaluated The Violence Against Women Office's (VAWO) Rural Domestic
Violence Program, begun in fiscal year 1996, has funded 92 grantees through
September 2001. The primary purpose of the program is to enhance the safety
of victims of domestic abuse, dating violence, and child abuse. The program
supports projects that implement, expand, and establish cooperative efforts
between law enforcement officers, prosecutors, victim advocacy groups, and
others in investigating and prosecuting incidents of domestic violence,
dating violence, and child abuse; provide treatment, counseling, and
assistance to victims; and work with the community to develop educational
and prevention strategies directed toward these issues.

Evaluation components The impact evaluation began in July 2000, with a final
report expected no earlier than April 2002. Initially, 10 grantees were
selected to participate in the impact evaluation; 9 remain in the
evaluation. Two criteria were used in the selection of grant participants:
the "feasibility" of earlier site-visited grantees to conduct an outcome
evaluation and VAWO recommendations based on knowledge of grantee program
activities. Logic models were developed, as part of the case study approach,
to show the logical or plausible links between a grantee's activities and
the desired outcomes. The specified outcome data were to be collected from
multiple sources, using a variety of methodologies during 2- to-3-day site
visits (e.g., multiyear criminal justice, medical, and shelter statistics
were to be collected from archival records; community stakeholders were to
be interviewed; and grantee and victim service agency staff were to
participate in focus groups). At the time of our review, this evaluation was
funded at $719,949. The National Institute of Justice (NIJ) could not
separate the cost of the impact evaluation from the cost of the process
evaluation.

                  Evaluation findings Too early to assess.

Assessment of evaluation This evaluation has several limitations. (1) The
choice of the 10 impact sites is skewed toward the technically developed
evaluation sites and is not representative of either all Rural Domestic
Violence Program Grantees, particular types of projects, or delivery styles.
(2) The lack of comparison groups will make it difficult to exclude the
effect of external factors, such as victim safety and improved access to
services, on perceived change. (3) Several so-called short-term outcome
variables are in fact process variables (e.g., number of police officers who
complete a training program on domestic violence, identification of barriers
to meeting the needs of women and children). (4) It is not clear how
interview and focus group participants are to be selected, (5) Statistical
procedures to be used in the analyses have not been sufficiently identified.
The NIJ peer review committee had concerns about whether the evaluation
could demonstrate that the program was working. NIJ funded the application
as a cooperative agreement because a substantial amount of agency
involvement was deemed necessary to meet the objectives of the evaluation.

Appendix I: Summaries of the Impact Evaluations of Byrne and VAWO Programs

Evaluation National Evaluation of the Arrest Policies Program

Principal investigator Institute for Law and Justice (ILJ)

Program evaluated The purpose of VAWO's Arrest Policies Program is to
encourage states, local governments, and Indian tribal governments to treat
domestic violence as a serious violation of criminal law. The program
received a 3-year authorization (fiscal years 1996 through 1998) at
approximately $120 million to fund grantees under six purpose areas:
implementing mandatory arrest or proarrest programs and policies in police
departments, tracking domestic violence cases, centralizing and coordinating
police domestic violence operations, coordinating computer tracking systems,
strengthening legal advocacy services, and educating judges and others about
how to handle domestic violence cases. Grantees have flexibility to work in
several of these areas. At the time the NIJ evaluation grant was awarded,
130 program grantees had been funded; the program has since expanded to 190
program grantees.

Evaluation components The impact evaluation began in August 1998, with a
draft final report due in March 2002. Six grantees were chosen to
participate in the impact evaluation. Each of the six sites was selected on
the basis of program "stability," not program representativeness. Within
sites, both quantitative and qualitative data were to be collected and
analyzed to enable better understanding of the impact of the Arrest Program
on offender accountability and victim well being. This process entailed
reviewing data on the criminal justice system's response to domestic
violence; tracking a random sample of 100 offender cases, except in rural
areas, to determine changes in offender accountability; conducting content
analyses of police incident reports to assess change in police practices and
documentation; and interviewing victims or survivors at each site to obtain
their perceptions of the criminal justice system's response to domestic
violence and its impact on their well-being. ILJ had planned cross-site
comparisons and the collection of extensive historical and interview data to
test whether competing factors could be responsible for changes in arrest
statistics. At the time of our review, this evaluation was funded at
$1,130,574. NIJ could not separate the cost of the impact evaluation from
the cost of the process evaluation.

                  Evaluation findings Too early to assess.

Assessment of evaluation This evaluation has several limitations: the
absence of a representative sampling frame for site selection, the lack of
comparison groups, the inability to conduct cross-site comparisons, and the
lack of a sufficient number of victims in some sites to provide a
perspective on the changes taking place in domestic violence criminal
justice response patterns and victim well-being. In addition, there was
difficulty collecting pre-grant baseline data, and the quality of the data
was oftentimes inadequate, limiting the ability to measure change over time.
Further, automated data systems were not available in all work areas, and
data were missing, lost, or unavailable. An NIJ peer review committee also
expressed some concerns about the grantee's methodological design.

Appendix I: Summaries of the Impact Evaluations of Byrne and VAWO Programs

Evaluation Impact  Evaluation of  STOP Grant Programs  for Reducing Violence
Against Indian Women

Principal investigator The University of Arizona

Program evaluated VAWO's STOP (Services, Training, Officers, and
Prosecutors) Grant Programs for Reducing Violence Against Indian Women
Discretionary Grant Program was established under Title IV of the Violent
Crime Control and Law Enforcement Act of 1994. The program's principal
purpose is to reduce violent crimes against Indian women. The program, which
began in fiscal year 1995 with 14 grantees, encourages tribal governments to
develop and implement culture-specific strategies for responding to violent
crimes against Indian women and providing appropriate services for those who
are victims of domestic abuse, sexual assault, and stalking. In this effort,
the program provided funding for the services and training, and required the
joint coordination, of nongovernmental service providers, law enforcement
officers, and prosecutors hence the name, the STOP Grant Programs for
Reducing Violence Against Indian Women.

Evaluation components The University of Arizona evaluation began in October
1996 with an expected final report due in March 2002. The basic analytical
framework of this impact evaluation involves the comparison of quantitative
and qualitative pre-grant case study histories of participating tribal
programs with changes taking place during the grant period. Various data
collection methodologies have been adopted (at least in part, to be
sensitive to the diverse Indian cultures): 30-minute telephone interviews,
mail surveys, and face-to-face 2- to 3 day site visits. At the time of our
review, this evaluation was funded at $468,552. NIJ could not separate the
cost of the impact evaluation from the cost of the process evaluation.

                  Evaluation findings Too early to assess.

Assessment of evaluation Methodological design and implementation issues may
cause difficulties in attributing program impact. A number of methodological
aspects of the study remain unclear: the site selection process for
"in-depth case study evaluations;" the methodological procedures for
conducting the longitudinal evaluation; the measurement, validity, and
reliability of the outcome variables; the procedures for assessing impact;
and the statistical tests to be used for determining significant change.
Comparison groups are not included in the methodological design. In
addition, only 43 percent of the grantees returned the mail survey, only 25
percent could provide the required homicide and hospitalization rates; and
only 26 victims of domestic violence and assault could be interviewed
(generally too few to measure statistical change). Generalization of
evaluation results to the entire STOP Grant Programs for Reducing Violence
Against Indian Women will be difficult, given these problems.

Appendix I: Summaries of the Impact Evaluations of Byrne and VAWO Programs

Longitudinal Impact  Evaluation of the Strategic  Intervention for High Risk
Youth (a.k.a. The Children at Risk Program)

Evaluation

                 Principal investigator The Urban Institute

Program evaluated The Children at Risk (CAR) Program, a comprehensive drug
and delinquency prevention initiative funded by the Bureau of Justice
Assistance (BJA), the Office of Juvenile Justice and Delinquency Prevention
(OJJDP), the Center on Addiction and Substance Abuse, and four private
foundations, was established to serve as an experimental demonstration
program from 1992 to 1996 in five grantee cities. Low-income youths (11 to
13 years old) and their families, who lived in severely distressed
neighborhoods at high-risk for drugs and crime, were targeted for
intervention. Eight core service components were identified: case
management, family services, education services, mentoring, after-school and
summer activities, monetary and nonmonetary incentives, community policing,
and criminal justice and juvenile intervention (through supervision and
community service opportunities). The goals of the program were to reduce
drug use among targeted families and improve the safety and overall quality
of life in the community.

Evaluation components The evaluation actually began in 1992, and the final
report was submitted in May 1998. The study used both experimental and
quasi-experimental evaluation designs. A total of 671 youths in target
neighborhood schools were randomly assigned to either a treatment group
(which received CAR services and the benefit of a safer neighborhood) or to
a control group (which received only a safer neighborhood). Comparison
groups (n=203 youths) were selected from similar high-risk neighborhoods by
means of census tract data; comparison groups did not have access to the CAR
Program. Interviews were conducted with youth participants at program entry
(baseline), program completion (2 years later), and 1-year after program
completion. A parent or caregiver was interviewed at program entry and
completion. Records from schools, police, and courts were collected annually
for each youth in the sample as a means of obtaining more objective data.
The total evaluation funding was $1,034,732.

Evaluation findings Youths participating in CAR were significantly less
likely than youths in the control group to have used gateway and serious
drugs, to have sold drugs, or to have committed violent crimes in the year
after the program ended. CAR youths were more likely than youths in the
control and comparison groups to report attending drug and alcohol abuse
programs. CAR youths received more positive peer support than controls,
associated less frequently with delinquent peers, and were pressured less
often by peers to behave in antisocial ways. CAR households used more
services than control group households, but the majority of CAR households
did not indicate using most of the core services available.

Assessment of evaluation CAR is a methodologically rigorous evaluation in
both its design and implementation. The evaluation findings demonstrate the
value of the program as a crime and drug prevention initiative.

Appendix II: NIJ's Guidelines for Disseminating Discretionary Grant Program
Evaluation Results

NIJ's Dissemination Practices

NIJ has the primary role of disseminating Byrne and VAWO discretionary grant
program evaluation results of evaluations managed by NIJ, according to NIJ,
BJA, and VAWO officials, because NIJ is responsible for conducting these
types of evaluations.1 NIJ is authorized to share the results of its
research with federal, state, and local governments.2 NIJ also disseminates
information on methodology designs.

NIJ's practices for disseminating program evaluation results are specified
in its guidelines. According to the guidelines, once NIJ receives a final
evaluation report from the evaluators and the results of peer reviews have
been incorporated, NIJ grant managers are to carefully review the final
product and, with their supervisor, recommend to the NIJ Director which
program results to disseminate and the methods for dissemination. Before
making a recommendation, grant managers and their supervisors are to
consider various criteria, including policy implications, the nature of the
findings and research methodology, the target audience and their needs, and
the cost of various forms of dissemination. Upon receiving the
recommendation, the Director of NIJ is to make final decisions about which
program evaluation results to disseminate. NIJ's Director of Planning and
Management said that NIJ disseminates program evaluation results that are
peer reviewed, are deemed successful, and add value to the field.

Once the decision has been made to disseminate program evaluation results
and methodologies with researchers and practitioners, NIJ can choose from a
variety of publications, including its Research in Brief; NIJ Journal-At a
Glance: Recent Research Findings; Research Review; NIJ Journal-Feature
Article; and Research Report. In addition, NIJ provides research results on
its Internet site and at conferences. For example, using its Research in
Brief publication, NIJ disseminated impact evaluation results on the Byrne
Children at Risk (CAR) program to 7,995 practitioners and researchers,
including state and local government and law enforcement officials; social
welfare and juvenile justice professionals; and criminal justice
researchers. In addition, using the same format, NIJ stated that it
distributed the results of its process evaluation of the Byrne Comprehensive
Communities Program (CCP) to 41,374 various constituents, including local
and state criminal and juvenile justice agency

1BJA and VAWO also disseminate evaluation results of their discretionary
grant programs.

242 U.S.C. 3721.

Appendix II: NIJ's Guidelines for Disseminating Discretionary Grant Program
Evaluation Results

administrators, mayors and city managers, leaders of crime prevention
organizations, and criminal justice researchers. NIJ and other OJP offices
and bureaus also disseminated evaluation results during NIJ's annual
conference on criminal justice research and evaluation. The July 2001
conference was attended by 847 public and nonpublic officials, including
criminal justice researchers and evaluation specialists from academic
institutions, associations, private organizations, and government agencies;
federal, state, and local law enforcement, court, and corrections officials;
and officials representing various social service, public housing, school,
and community organizations.

In addition to NIJ's own dissemination activities, NIJ's Director of
Planning and Management said that it allows and encourages its evaluation
grantees to publish their results of NIJ-funded research via nongovernmental
channels, such as in journals and through presentations at professional
conferences. Although NIJ requires its grantees to provide advance notice if
they are publishing their evaluation results, it does not have control over
its grantees' ability to publish these results. NIJ does, however, require a
Justice disclaimer that the "findings and conclusions reported are those of
the authors and do not necessarily reflect the official position or policies
of the U.S. Department of Justice." For example, although NIJ has not yet
disseminated the program evaluation results of the three ongoing VAWO impact
evaluations that we reviewed, one of the evaluation grantees has already
issued, on its own Internet site, 9 of 20 process evaluation reports on the
Arrests Policies evaluation grant. The process evaluations were a component
of the NIJ grantee's impact evaluation of the Arrest Policies Program.
Because the evaluations were not completed, NIJ required that the grantee's
publication of the process evaluations be identified as a draft report
pending final NIJ review.

Appendix II: NIJ's Guidelines for Disseminating Discretionary Grant Program
Evaluation Results

Dissemination of NIJ's Byrne Discretionary Grants: Comprehensive Communities
Program and Children at Risk Program

As discussed earlier, NIJ publishes the results of its evaluations in
several different publications. For example, NIJ used the Research in Brief
format to disseminate evaluation results for two of the five Byrne
discretionary grant programs Comprehensive Communities Program (CCP) and
Children at Risk Program (CAR) that were evaluated during fiscal years 1995
through 2001. Both publications summarize information including each
program's evaluation results, methodologies used to conduct the evaluations,
information about the implementation of the programs themselves, and
services that the programs provided.3 CCP's evaluation results were based on
a process evaluation. Although a process evaluation does not assess the
results of the program being evaluated, it can provide useful information
that explains the extent to which a program is operating as intended. The
NIJ Research in Brief on the Byrne CAR Discretionary Grant Program provides
a summary of issues and findings regarding the impact evaluation. That
summary included findings reported one year after the end of the program, in
addition to a summary of the methodology used to conduct the evaluation, the
outcomes, the lessons learned, and a major finding from the evaluation.

3Although Justice funds these grants and publishes them, its publications
carry a disclaimer that says "findings and conclusions of the research
reported here are those of the authors and do not necessarily reflect the
official position or policies of the U.S. Department of Justice."

Appendix III: Comments from the Department of Justice

Note: GAO comments supplementing those in the report text appear at the end
of this appendix.

Appendix III: Comments from the Department of Justice

Appendix III: Comments from the Department of Justice

                                Now on p. 8.

                               Now on p. 11.

                               Now on p. 12.

                        Now on p. 13. Now on p. 15.

                               Now on p. 15.

Appendix III: Comments from the Department of Justice

Appendix III: Comments from the Department of Justice

                               See comment 1.

                               See comment 2.

Appendix III: Comments from the Department of Justice

Now on pp.15-16. See comment 3.

Appendix III: Comments from the Department of Justice

See comment 4.

Now on p. 11.

Now on pp. 14 and 18. See comment 5.

See comment 6.

Appendix III: Comments from the Department of Justice

                        Now on p. 28. Now on p. 17.

                               Now on p. 13.

                               See comment 7.

Appendix III: Comments from the Department of Justice

Now on p. 14. See comment 8.

Now on p. 15. See comment 9.

Appendix III: Comments from the Department of Justice

Now on p. 5.

See comment 10.

Now on p. 25. See comment 11.

Now on pp. 25-27.

See comment 12.

Now on p. 26. Now on p. 25. Now on p. 27.

See comment 13.

Appendix III: Comments from the Department of Justice

Now on p. 31. See comment 14.

Appendix III: Comments from the Department of Justice

GAO Comments

Following are GAO's comments on the Department of Justice's February 13,
2002, letter.

1. We have amended the text to further clarify that BJA administers the
Byrne program, just as its counterpart, VAWO, administers its programs (see
page 4). However it is important to point out that regardless of the issues
raised by OJP, the focus of our work was on the methodological rigor of the
evaluations we reviewed, not the purpose and structure of the programs being
evaluated. As discussed in our Scope and Methodology section, our work
focused on program evaluation activities associated with Byrne and VAWO
discretionary grant programs generally and the methodological rigor of
impact evaluation studies associated with those programs in particular. To
make our assessment, we relied on NIJ officials to identify which of the
program evaluations of Byrne and VAWO grant programs were, in fact, impact
evaluation studies. We recognize that there are substantial differences
among myriad OJP programs that can make the design and implementation of
impact evaluations arduous. But, that does not change the fact that impact
evaluations, regardless of differences in programs, can benefit from
stronger up-front attention to better ensure that they provide meaningful
and definitive results.

2. We disagree with OJP's assessment of our report's treatment of program
variation. As discussed earlier, the scope of our review assessed impact
evaluation activities associated with Byrne and VAWO discretionary grant
programs, not the programs themselves. We examined whether the evaluations
that NIJ staff designated as impact evaluations were designed and
implemented with methodological rigor. In our report we observe that
variations in projects funded through VAWO programs complicate the design
and implementation of impact evaluations.

According to the Assistant Attorney General, this variation in projects is
consistent with the intent of the programs' authorizing legislation. We
recognize that the authorizing legislation provides VAWO the flexibility in
designing these programs. In fact, we point out that although such
flexibility may make sense from a program perspective, project variation
makes it much more difficult to design and implement a definitive impact
evaluation. This poses sizable methodological problems because an aggregate
analysis, such as one that might be constructed for an impact evaluation,
could mask the differences in effectiveness among individual projects and
therefore not result in information about which configurations of projects
work and which do not.

3. We have amended the Results in Brief to clarify that peer reviews
evaluated proposals. However, it is important to note that while the peer
review committees may have found the two VAWO grant applications to be the
most superior, this does not necessarily imply that the impact evaluations
resulting from these

Appendix III: Comments from the Department of Justice

applications were well designed and implemented. As discussed in our report,
the peer review panel for each of the evaluations expressed concerns about
the proposals that were submitted, including issues related to site
selection and the need for additional impact measures and control variables.
Our review of the documents NIJ made available to us, including evaluators'
responses to peer review comments, led to questions about whether the
evaluators' proposed methodological designs were sufficient to allow the
evaluation results to be generalized and to determine whether the program
was working.

4. We have amended the Background section of the report to add this
information (see page 6).

5. As discussed in OJP's comments, we discussed external factors that could
account for changes that the Rural Program evaluation observed in victims'
lives and the criminal justice system. We did so not to critique or endorse
activities that the program was or was not funding, but to demonstrate that
external factors may influence evaluation findings. To the extent that such
factors are external, the Rural Program evaluation methodology should
account for their existence and attempt to establish controls to minimize
their affect on results (see page 14). We were not intending to imply that
alcohol is a cause for domestic violence, as suggested by the Assistant
Attorney General, but we agree that it could be an exacerbating factor that
contributes to violence against women.

6. As discussed earlier, we recognize that there are substantive differences
in the intent, structure, and design of the various discretionary grant
programs managed by OJP and its bureaus and offices. Also, as stated
numerous times in our report, we acknowledge not only that impact evaluation
can be an inherently difficult and challenging task but also that measuring
the impact of these specific Byrne and VAWO programs can be arduous given
that they are operating in an ever changing, complex environment. We agree
that not all evaluation issues that can compromise results are easily
resolvable, but we firmly believe that with more up-front attention to
design and implementation issues, there is a greater likelihood that NIJ
impact evaluations will provide meaningful results for policymakers.

Regarding the representativeness of sites, NIJ documents that were provided
during our review indicated that sites selected during the Rural Program
evaluation were selected on the basis of feasibility, as discussed in our
report- specifically, whether the site would be among those participants
equipped to conduct an evaluation. In its comments, OJP stated that the 6
sites selected for the impact evaluation were chosen to maximize
geographical and purpose area diversity while focusing on sites with high
program priority. OJP did not provide any additional information that would
further indicate that the sites were selected on a representative basis. OJP
did, however, point out that the report does not address how immensely
expensive the Arrest evaluation would have become if it had included all 130
sites. We did not address specific evaluation site costs

Appendix III: Comments from the Department of Justice

because we do not believe that there is a requisite number of sites needed
for any impact evaluation to be considered methodologically rigorous.
Regarding OJP's comment about the flexibility given to grantees in
implementing VAWO grants, our report points out that project variation
complicates evaluation design and implementation. Although flexibility may
make sense from a program perspective, it makes it difficult to generalize
about the impact of the entire program.

7. We used the drug court example to illustrate, based on our past work, how
comparison groups can be used in evaluation to isolate and minimize external
factors that could influence the study results. We did not, nor would we,
suggest that any particular unit of analysis is appropriate for VAWO
evaluations since the appropriate unit of analysis is dependent upon the
specific circumstances of the evaluation. We were only indicating that since
comparison groups were not utilized in the studies, the evaluators were not
positioned to demonstrate that change took place as a result of the program.

8. We do not dispute that VAWO grant programs may provide valuable outputs
over the short term. However, as we have stated previously, the focus of our
review was on the methodological rigor of impact evaluations--those
evaluations that are designed to assess the net effect of a program by
comparing program outcomes with an estimate of what would have happened in
the absence of the program. Given the methodological issues we found, it is
unclear whether NIJ will be able to discern long-term effects due to the
program.

9. As stated in our report, we acknowledge not only that impact evaluation
can be an inherently difficult and challenging task, but that measuring the
impact of Byrne and VAWO programs can be arduous given the fact that they
are operating in an ever changing, complex environment. We agree that not
all evaluation issues that can compromise results are easily resolvable, but
we firmly believe that, with more up-front attention to design and
implementation issues, there is a greater likelihood that NIJ evaluations
will provide meaningful results for policymakers. As we said before, absent
this up-front attention, questions arise as to whether NIJ is (1) positioned
to provide the definitive results expected from an impact evaluation and (2)
making sound investments given the millions of dollars spent on these
evaluations. If NIJ believes that the circumstances of a program are such
that it cannot be evaluated successfully (in relation to impact) they should
not proceed with an impact evaluation.

10. We have amended the footnote to state that from fiscal year 1995 through
fiscal year 1999, this program was administered by VAWO. As of fiscal year
2000, responsibility for the program was shifted to OJP's Corrections
Program Office (see page 5).

11. In regard to the number of grants, we have amended the text to reflect
that the information NIJ provided during our review is the number of
grantees, not the

Appendix III: Comments from the Department of Justice

number of grants (see pages 25 and 26). We have also amended our report to
reflect some of the information provided in VAWO's description of the Rural
Domestic Violence Program to further capture the essence of the program (see
page 25).

12. We disagree. We believe that separating the cost of the impact and
process evaluations is more than a matter of bookkeeping. Even though the
work done during the process phase of an evaluation may have implications
for the impact evaluation phase of an evaluation, it would seem that, given
the complexity of impact evaluations, OJP and NIJ would want to have in
place appropriate controls to provide reasonable assurance that the
evaluations are being effectively and efficiently carried out at each phase
of the evaluation. Tracking the cost of these evaluation components would
also help reduce the risk that OJP's, NIJ's, and, ultimately, the taxpayer's
investment in these impact evaluations is not wasted.

13. As discussed earlier, we recognize that there are substantive
differences in the intent, structure, and design of the various
discretionary grant programs managed by OJP and its bureaus and offices,
including those managed by VAWO. Our report focuses on the rigor of impact
evaluations of grant programs administered by VAWO and not on the program's
implementing legislations. Although flexibility may make sense from a
program perspective, it makes it difficult to develop a well designed and
methodologically rigorous evaluation that produces generalizeable results
about the impact of the entire program.

14. Our report does not suggest that other types of evaluations, such as
comprehensive process evaluations, are any less useful in providing
information about how well a program is operating. The scope of our review
covered impact evaluations of Byrne and VAWO discretionary grant
programs-those designed to assess the net effect of a program by comparing
program outcomes with an estimate of what would have happened in the absence
of the program.

Appendix IV: GAO Contacts and Staff Acknowledgments

GAO Contacts Acknowledgments

(440025)

Laurie E. Ekstrand, (202) 512-8777 John F. Mortin, (202) 512-8777

In addition to the above, Wendy C. Simkalo, Jared A. Hermalin, Chan My J.
Battcher, Judy K. Pagano, Grace A. Coleman, and Ann H. Finley made key
contributions to this report.

GAO's Mission

Obtaining Copies of GAO Reports and Testimony

The General Accounting Office, the 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.

The fastest and easiest way to obtain copies of GAO documents is through the
Internet. GAO's Web site (www.gao.gov) contains abstracts and full-text
files of current reports and testimony and an expanding archive of older
products. The Web site features a search engine to help you locate documents
using key words and phrases. You can print these documents in their
entirety, including charts and other graphics.

Each day, GAO issues a list of newly released reports, testimony, and
correspondence. GAO posts this list, known as "Today's Reports," on its Web
site daily. The list contains links to the full-text document files. To have
GAO e-mail this list to you every afternoon, go to www.gao.gov and select
"Subscribe to daily e-mail alert for newly released products" under the GAO
Reports heading.

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. General Accounting Office P.O. Box 37050 Washington, D.C. 20013

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

Visit GAO's Document GAO Building

Distribution  Center  Room 1100,  700 4th  Street, NW  (corner of 4th  and G
Streets, NW) Washington, D.C. 20013

To Report Fraud, Contact: Web site: www.gao.gov/fraudnet/fraudnet.htm,

Waste, and Abuse in E-mail: [email protected], or

Federal Programs 1-800-424-5454 or (202) 512-7470 (automated answering
system).

Jeff   Nelligan,  Managing   Director,   [email protected]   (202)  512-4800

Public Affairs U.S.  General Accounting Office, 441 G. Street NW, Room 7149,
Washington, D.C. 20548
*** End of document. ***