School Vouchers: Publicly Funded Programs in Cleveland and	 
Milwaukee (31-AUG-01, GAO-01-914).				 
								 
This report reviews publicly funded school vouchers, which	 
provide money for families to send their children to private	 
schools, in the Cleveland and Milwaukee voucher programs. State  
laws and regulations govern the participation of students and	 
schools in the Cleveland and Milwaukee programs. Both programs	 
were primarily designed to provide educational opportunity for	 
students of low-income families residing within the Cleveland	 
school district or the city of Milwaukee and are similar in their
design. Both programs require participating private schools to be
located within the city or the city's school district and to	 
adhere to state standards for private schools, such as those	 
covering health and safety. Among the research reports addressing
student characteristics, the most commonly reported data	 
(excluding race) focused on family income, the family's living	 
arrangement, and parents' education. In both Cleveland and	 
Milwaukee, voucher students were more likely than public school  
students to come from families that had less income and were	 
headed by a parent who was single or not married to the person he
or she were living with. Some information about the racial and	 
ethnic composition of Cleveland's and Milwaukee's public school  
and voucher students is available, but whether the composition	 
changed as a result of the voucher programs is unclear. Ohio and 
Wisconsin use different methods to provide state funds for the	 
Cleveland and Milwaukee voucher programs and spend less on	 
voucher students than on public school students. The Cleveland	 
voucher program is funded with Disadvantaged Pupil Impact Aid	 
funds up to a limit established by the Ohio Legislature.	 
Wisconsin funds the Milwaukee voucher program with general state 
aid funds based on the number of students participating in the	 
program in a given year. The contracted evaluations of voucher	 
students' academic achievement in Cleveland and Milwaukee found  
little or no difference in voucher and public school students'	 
performance, but studies by other investigators found that	 
voucher students did better in some of the subject areas tested. 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-01-914 					        
    ACCNO:   A01676						        
    TITLE:   School Vouchers: Publicly Funded Programs in Cleveland   
             and Milwaukee                                                    
     DATE:   08/31/2001 
  SUBJECT:   Academic achievement				 
	     Aid for education					 
	     Disadvantaged persons				 
	     Elementary schools 				 
	     Eligibility criteria				 
	     Funds management					 
	     Private schools					 
	     Quality control					 
	     Secondary schools					 
	     Cleveland (OH)					 
	     Milwaukee (WI)					 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Testimony.                                               **
**                                                              **
** 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-01-914
     
Report to the Honorable Judd Gregg, U. S. Senate

United States General Accounting Office

GAO

August 2001 SCHOOL VOUCHERS Publicly Funded Programs in Cleveland and
Milwaukee

GAO- 01- 914

Page i GAO- 01- 914 Publicly Funded Voucher Programs Letter 1

Results in Brief 2 Background 5 Cleveland and Milwaukee Voucher Programs
Have Similar Student

and Private School Eligibility Criteria 9 Voucher Family Characteristics, as
Well as Voucher School

Attributes, Differ From Those of Public School Families and Their Schools 12
Minority Group Members Predominate in School Populations, but

Whether Voucher Programs Have Changed Schools? Racial Composition is Unclear
17 States Fund Voucher Programs Differently and Spend Less on Each

Voucher Student Than on Each Public School Student 22 Contract Researchers
Found Little or No Significant Improvement

in Voucher Students? Achievement, but Other Investigators Found Some
Positive Effects 27 Concluding Observations 31 Agency and Other Comments 32

Appendix I Scope and Methodology 34

Appendix II Information on Publicly Funded School Voucher Programs in
Florida, Maine, and Vermont 38

Appendix III Additional Data on Racial and Ethnic Composition of Cleveland
Schools 42

Appendix IV Strengths and Limitations of Studies Analyzing Cleveland
Students? Academic Achievement 43

Appendix V Strengths and Limitations of Studies Analyzing Milwaukee
Students? Academic Achievement 45 Contents

Page ii GAO- 01- 914 Publicly Funded Voucher Programs Appendix VI GAO
Contacts and Staff Acknowledgments 47

Bibliography 48

Tables

Table 1: Characteristics of Cleveland Families With Students in the Voucher
Program or in the Public Schools 14 Table 2: Characteristics of Milwaukee
Families With Students in

the Voucher Program or in the Public Schools 15 Table 3: Comparison of
Cleveland Private Voucher and Public

School Characteristics 16 Table 4: Racial and Ethnic Composition of
Cleveland Public School

and Voucher Students, School Year 1998- 99 17 Table 5: Racial and Ethnic
Composition of Milwaukee Public

School and Voucher Students 20 Table 6: Data Quality Criteria 35 Table 7:
Criteria for Assessing Studies of Academic Achievement 35 Table 8: Racial
and Ethnic Composition of Public School and

Voucher Students in Cleveland 42

Figures

Figure 1: Strengths and Limitations of Studies Analyzing Cleveland Students?
Academic Achievement 43 Figure 2: Strengths and Limitations of Studies
Analyzing

Milwaukee Students? Academic Achievement 45

Abbreviations

GED General Education Development

Page 1 GAO- 01- 914 Publicly Funded Voucher Programs

August 31, 2001 The Honorable Judd Gregg United States Senate

Dear Senator Gregg: School voucher programs have generated considerable
debate nationwide. Only three states- Florida, Ohio, and Wisconsin- have
initiated programs within the last decade to provide public funding that
allows families to send their children to private elementary and secondary
schools. These states created their voucher programs largely in reaction to
state and parental dissatisfaction with the quality of public schools.
Opinions vary widely about whether voucher programs are a viable solution to
public school problems. Some advocates have claimed that voucher programs
will provide poor families with educational choices like those available to
affluent families and will promote greater competition among schools,
forcing them to become more effective in order to remain viable. Opponents
have stressed that voucher programs will hurt public school funding by
steering tax dollars to private schools, and that evidence of student
achievement gains is inconclusive. Opponents have also warned of the
potential for increased segregation by race and income as voucher schools
may attract the best students from the traditional public schools.

You indicated your interest in both publicly and privately funded school
voucher programs. For this report you requested that we provide information
on the publicly funded school voucher programs in Cleveland and Milwaukee,
primarily based on research conducted on these programs. 1 As of the 1999-
2000 school year, the Cleveland program had about 3,400 voucher students
enrolled in 52 private schools, and the Milwaukee program had about 7,600
students enrolled in 91 private schools. We focused our work on answering
the following questions:

1 We were asked to focus on the publicly funded Ohio and Wisconsin voucher
programs because the Florida voucher program, with first- year
implementation in the 1999- 2000 school year, was too new for our review. In
addition, we did not include the publicly funded voucher programs in Maine
and Vermont because those programs primarily focus on rural communities
without adequate public school capacity. While this review was not intended
to include privately funded voucher programs, such as those in the District
of Columbia, New York City, Dayton, and San Antonio, we do plan to review
such programs. We are aware that several studies of these privately funded
voucher programs have been conducted.

United States General Accounting Office Washington, DC 20548

Page 2 GAO- 01- 914 Publicly Funded Voucher Programs

 What are the key eligibility criteria for student and private school
participation in the Cleveland and Milwaukee voucher programs?

 How does research characterize students and private schools participating
in these voucher programs compared to public schools and their students in
the two cities?

 What is known about the racial composition of public school and voucher
students and whether this composition has changed as a result of the voucher
programs?

 What is known about how these voucher programs are funded and the amount
spent on public school students compared to the amount spent on voucher
students?

 What is known about student academic achievement within these voucher
programs?

Most of the data we used to address these questions were collected from
studies of the Cleveland and Milwaukee voucher programs. We selected studies
for review that met two or more of the following criteria: (1) were
performed under contract to Ohio or Wisconsin state departments of
education; (2) had been published in a peer- reviewed journal; (3) had been
issued by a research institution that reviews work prior to release; or (4)
employed quantitative analysis to examine student academic achievement. The
studies we selected included assessments of the student and school
characteristics, racial and ethnic composition of public school and voucher
students, and evaluations of Cleveland and Milwaukee students? academic
achievement. We examined the studies to determine if the data they contained
met criteria for reliability and completeness and if analyses of the voucher
programs? effects employed impact evaluation methods, such as the use of
comparison groups and multivariate analysis procedures. We did not include
data in our analyses from studies that did not meet our criteria. In
addition to these studies, we reviewed state laws and regulations, school
district and voucher program office records, state budget data, and a
variety of school voucher- related reports and articles. We conducted site
visits to Ohio and Wisconsin and interviewed officials from the state
departments of education, the administering program offices, the public
school districts, and several private schools. Appendix I further describes
our scope and methodology. A list of the studies we reviewed is provided in
the bibliography to this report. We conducted our review between February
and July 2001 in accordance with generally accepted government auditing
standards.

State laws and regulations govern the participation of students and schools
in the Cleveland and Milwaukee programs. Both programs were primarily
designed to provide educational opportunity for students of low Results in
Brief

Page 3 GAO- 01- 914 Publicly Funded Voucher Programs

income families residing within the Cleveland school district or the city of
Milwaukee and are similar in their design. In Cleveland, priority is given
to students from families whose income is less than 200 percent of federal
poverty guidelines. However, if low- income Cleveland students do not use
all available vouchers, students above the low- income guidelines may use
the remaining vouchers. For Milwaukee, all students must come from families
whose incomes do not exceed 175 percent of federal poverty guidelines. Both
programs require participating private schools to be located within the city
or the city?s school district and to adhere to state standards for private
schools, such as those covering health and safety. These schools must also
agree to use random selection processes to determine voucher student
enrollment when student applications exceed available slots at specific
schools.

Among the research reports addressing student characteristics, the most
commonly reported data (excluding race) focused on family income, the
family?s living arrangement, and parents? education. In both Cleveland and
Milwaukee, voucher students were more likely than public school students to
come from families that had less income and were headed by a parent who was
single or not married to the person he or she were living with. 2 Voucher
students? parents in Cleveland and mothers in Milwaukee were also more
likely to have completed at least high school than were public school
students? parents. Some research for Milwaukee also provided reliable
information on students? academic achievement prior to their participation
in a voucher program. Researchers found that voucher schools in Milwaukee
were attracting lower- performing public school students. Data that
addressed school characteristics showed that Cleveland voucher schools,
compared to public schools, had lessexperienced teachers and smaller class
sizes. No comparable school characteristics data were collected by the
contract researchers for the Milwaukee program. Other data indicate that the
majority of participating voucher schools were religious in Cleveland since
the program?s inception, whereas the majority have been religious in
Milwaukee since the 1998- 99 school year, when religious schools were first
admitted to the program.

Some information about the racial and ethnic composition of Cleveland?s and
Milwaukee?s public school and voucher students is available, but

2 In Cleveland, a sample of voucher and public school student families was
surveyed, whereas in Milwaukee, all voucher and a sample of public school
student families were surveyed. The survey of Cleveland families during the
1998- 1999 school year met our criteria for data reliability, but the 1991
survey of Milwaukee families did not. See our report of findings for a
discussion of why the data are included.

Page 4 GAO- 01- 914 Publicly Funded Voucher Programs

whether the composition changed as a result of the voucher programs is
unclear. During school year 1998- 99, well over two- thirds of the students
enrolled in Cleveland?s and Milwaukee?s voucher programs and public schools
were minority group members. Most of the minority students were African-
American. Although the racial and ethnic composition data for the 1998- 99
school year are reliable, developing trend data is difficult for a variety
of reasons. For example, data available from the existing research for the
first 2 years of Cleveland?s program were unreliable or did not fully
represent the voucher and public school student population. Prior to 1998-
99, reliable racial and ethnic composition data for Milwaukee?s voucher
students were available only for intermittent school years: school years
1990- 91 to 1992- 93 and the 1994- 95 school year. In addition, studies that
have analyzed changes in the racial and ethnic composition of voucher and
public schools did not develop complete explanations of the changes. They
reached conclusions about the voucher programs? effect on racial composition
within voucher schools without considering the full range of factors that
could account for changes in the composition.

Ohio and Wisconsin use different methods to provide state funds for the
Cleveland and Milwaukee voucher programs and spend less on voucher students
than on public school students. The Cleveland voucher program is funded with
Disadvantaged Pupil Impact Aid funds up to a limit established by the Ohio
Legislature. These funds are deducted from the Cleveland school district?s
share of Disadvantaged Pupil Impact Aid. In the 1999- 2000 school year, the
cost of the voucher program was $6.2 million. Wisconsin funds the Milwaukee
voucher program with general state aid funds ($ 38.9 million in 1999- 2000)
based on the number of students participating in the program in a given
year. One- half of this amount is deducted from the Milwaukee school
district?s state revenues; the remainder is deducted from all other school
districts? state revenues in proportion to the total state revenues
authorized for each district. The full impact of these funding methods on
the public schools is unknown. The per- pupil amount expended for the
voucher program in Cleveland for the 1999- 2000 school year was $1,832,
which included the cost of voucher payments and program administration; for
Milwaukee, just the cost of voucher payments amounted to $5,106. Both these
amounts were less than the total respective per- pupil state aid that was
provided to the Cleveland ($ 4,910) and Milwaukee ($ 6,011) public school
districts.

The contracted evaluations of voucher students? academic achievement in
Cleveland and Milwaukee found little or no difference in voucher and public
school students? performance, but studies by other investigators found that
voucher students did better in some of the subject areas tested. None of the
findings can be considered definitive because the researchers

Page 5 GAO- 01- 914 Publicly Funded Voucher Programs

obtained different results when they used different methods to compensate
for weaknesses in the data. All of the studies satisfied most of the basic
criteria for research quality, such as using study designs and data analysis
methods that isolate the program?s effect, but they suffered from missing
test score data, low survey response rates, and the loss of students from
program groups and comparison groups over time. The researchers? different
findings likely were due to the different study designs, comparison groups
and statistical tests they used, and the extent of missing data on student
characteristics. Additional research will be needed to develop findings that
are conclusive and generalizable.

The Cleveland voucher program is officially called the Cleveland Scholarship
and Tutoring Program and provides state funding to help primarily low-
income children in kindergarten through the eighth grade attend private
schools in Cleveland or to attend public schools in districts adjacent to
the Cleveland school district. 3 The voucher program was implemented in the
1996- 97 school year, and only private schools have participated in it.
Students new to the program generally start in kindergarten through the
third grade and may have previously attended a public or a private school or
never attended school. In June 2000, the Cleveland program had about 3,400
voucher students enrolled in 52 private schools, which received about $5.2
million in publicly funded voucher payments for the 1999- 2000 school year.
4 By comparison, the Cleveland school district in 1999- 2000 had about
76,000 students enrolled in its 121 schools supported by $712 million in
total revenues. 5

In Cleveland, actual voucher payments follow the student to the school
attended, even when he or she changes schools. Voucher checks are made out
to the student?s parent or guardian and require endorsement before the
school can use the funds. These funds are sent to the participating schools
in two payments. Prior to payment, a voucher payment report is generated for
each participating school listing all current voucher students. Each school
verifies this report as accurate or updates it before it is sent to the

3 In addition to vouchers, the Cleveland program has a tutoring component
designed to provide additional academic assistance to children who continue
to attend Cleveland public schools.

4 The $5.2 million in voucher payments to students does not include about $1
million for the cost of voucher program administration. 5 Of the $712
million in total revenues for the Cleveland school district, $371. 9 million
was from the state of Ohio. Background

Page 6 GAO- 01- 914 Publicly Funded Voucher Programs

Ohio Department of Education?s School Finance Division to be processed for
payment. For low- income voucher students, the voucher amount is limited to
90 percent of school tuition up to a maximum of $2,250. For those voucher
students who do not come from low- income families, the voucher amount is
limited to 75 percent of school tuition up to a maximum of $1,875. Any
payments sent to a voucher school are proportionately reduced if a student
is not enrolled in the school for the entire period covered by the scheduled
voucher payment.

About 90 percent of the Cleveland voucher schools are religious schools. The
constitutionality of providing state- funded vouchers for attendance at
religious schools has been challenged in the courts since the program?s
inception. In December 2000, the U. S. Court of Appeals for the Sixth
Circuit ruled that the program is unconstitutional because it has the effect
of advancing religion, and that the program constitutes an endorsement of
religion and sectarian education in violation of the first amendment.
Subsequently, the court of appeals decided that the program could continue
operating while interested parties seek U. S. Supreme Court review of the
court of appeals? ruling.

Two teams have conducted research on the academic achievement of students in
Cleveland?s voucher program. The first was the contract researcher, a team
from Indiana University, which was contracted by the Ohio Department of
Education to conduct a multiyear evaluation of the program. The second team,
supported by Harvard University?s Program on Education Policy and Governance
(Harvard researchers), conducted its own studies. The contract research team
analyzed students? academic achievement in school years 1996- 97 and 1997-
98, the first 2 years of the voucher program. The Harvard team reanalyzed
the contract researcher team?s data for the first year and 1996- 97 data
from two additional private schools participating in the voucher program.

The Milwaukee voucher program, officially called the Milwaukee Parental
Choice Program, provides state funding exclusively for low- income children
in Milwaukee to attend private schools and was first implemented in the
1990- 91 school year. Wisconsin initially limited participation to
nonsectarian private schools, but amended the program to include religious
schools in 1995. For the 1994- 95 school year, 771 full- time equivalent
voucher students attended 12 nonreligious schools. Following legal
challenges to the 1995 program revision permitting religious school
participation, the Wisconsin Supreme Court upheld the revision in 1998, and
program enrollment tripled when Milwaukee voucher students began attending
religious schools in 1998- 99 school year. Subsequently, the U. S. Supreme
Court chose not to hear an appeal of the Wisconsin Supreme

Page 7 GAO- 01- 914 Publicly Funded Voucher Programs

Court decision that the program did not violate the First Amendment of the
United States Constitution. In school year 1998- 99, nearly threequarters of
the participating schools were religious.

Currently, students new to the program may start in kindergarten through the
12th grade if, in the year prior to enrolling, they attended a Milwaukee
public school; a Milwaukee private school in kindergarten, first, second, or
third grade; or never attended school anywhere. In the 1999- 2000 school
year, the Milwaukee program had 7,621 voucher students enrolled in 91
schools, which received about $38.9 million in publicly funded voucher
payments. 6 The Milwaukee school district in 1999- 2000 had about 105,000
students enrolled in 165 schools supported by $917 million in total
revenues. 7

The Wisconsin Department of Public Instruction makes voucher payments in
four installments during the school year. Similar to Ohio?s program, the
voucher check is payable to the voucher family. The Department mails the
checks to the schools where the parent or guardian endorses them to the
schools. If the school cannot obtain a signature because, for example, the
student is no longer enrolled, it returns the check to the Department. The
school keeps the lesser of the voucher amount or an amount equal to their
per- pupil operating and debt service costs as determined by an independent
financial audit. Because a school?s actual costs may be less than the
maximum allowable payment, and because of other factors that may require
adjustments to payments- such as audited enrollment reports- the Department
makes adjustments after the completion of the school year. Schools with
lower costs must return excess payments, and schools that gain students
receive an additional amount.

Wisconsin has required the Department of Public Instruction and the
Legislative Audit Bureau to evaluate the voucher program. The Department
contracted with an independent researcher to conduct an evaluation over the
first 5 years of the program. The evaluation focused on students? academic
achievement, at a time when student and private school participation was
limited to less than one- tenth of what it was in 1999- 2000 and was limited
to nonreligious schools. The evaluation was terminated at the end of school
year 1994- 95, and data on students?

6 The Milwaukee voucher program has not reached its statutory participation
limit of 15 percent of the Milwaukee school district?s enrollment. 7 Of the
$917 million in total revenues for the Milwaukee school district, $634. 1
million was from the state of Wisconsin.

Page 8 GAO- 01- 914 Publicly Funded Voucher Programs

characteristics have not been collected for an evaluation since then, nor
has student academic achievement been evaluated. Three teams conducted
research on Milwaukee?s voucher program during its early years: (1) the
contract researchers, a group of investigators affiliated with the
Department of Political Science and the Robert M. La Follette Institute of
Public Affairs, University of Wisconsin- Madison; (2) the Harvard team that
also conducted research on the Cleveland program; and (3) a researcher
affiliated with Princeton University. All three teams used the data set on
Milwaukee voucher and public school students and parents created by the
contract researcher team. All three teams also analyzed students? academic
achievement as measured by scores on the Iowa Test of Basic Skills
administered by the Milwaukee school district in school years 1990- 91 to
1993- 94. 8

In addition to the Cleveland and Milwaukee voucher programs, statefunded
voucher programs operate in Florida, Maine, and Vermont. Although not an
integral part of this review, some information on these three additional
state programs is provided to help put the Cleveland and Milwaukee programs
in a more complete context of publicly funded voucher programs. The Florida
voucher program began operating in the 1999- 2000 school year. The program
provides a private school choice to students whose public schools have been
judged by the state as failing. The Maine and Vermont programs have been
operating for more than 100 years and provide for the private, secular
education of students whose public school districts do not have sufficient
school capacity. More detail on the Florida, Maine, and Vermont voucher
programs is provided in appendix II.

Although not a direct sponsor of voucher programs, the federal government in
the past has sponsored research into alternative educational programs,
including a voucher program operated in a public school system. The National
Institute of Education sponsored research on an education voucher
demonstration program begun during the 1972- 73 school year in six schools
of the Alum Rock Union Elementary School District of San Jose, California.
In this demonstration, parents could choose from among these six public
schools and receive a voucher equal to the cost of the child?s education at
that school. The voucher amount was paid to the chosen school when the child
enrolled. After a 5- year

8 Because the state Department of Public Instruction required that new tests
be introduced, test scores for school year 1994- 95 were not comparable to
Iowa Test of Basic Skills scores for the previous 4 years. Thus, the teams?
analyses focused on the 4 years for which scores were comparable.

Page 9 GAO- 01- 914 Publicly Funded Voucher Programs

implementation period, an evaluation found little difference in the benefits
to students of the voucher and regular school programs. The U. S. Department
of Education is considering funding a grant to study Florida?s school
accountability system, which may include the voucher program and its effect
on improving school quality.

In accordance with state laws and regulations for student and school
participation in the Cleveland and Milwaukee voucher programs, both programs
target students from low- income families residing within the city or school
district. Income eligibility is determined by comparing applicant family
income to federal poverty guidelines. Participating private schools must be
located within the city or school district, comply with state requirements
for private schools- such as those covering health and safety- and randomly
select students when applications exceed available slots.

In the Cleveland voucher program, an eligible student must reside within the
Cleveland school district. Generally, first- time program enrollees must be
in kindergarten or grades one, two, or three. Priority for a voucher award
is given to students from families whose income is less than 200 percent of
federal poverty guidelines. 9 However, the state determines the number of
new vouchers that will be awarded each year within the limitations of the
amount of annual program funds appropriated. Any student who has received a
voucher in the preceding year may continue to receive one until the student
has completed grade eight. Assuming students? residency requirements are
maintained, school admission priority is given to students who were enrolled
in the school during the preceding year and to siblings of these students,
at the school?s discretion. A student?s family income is also a key
eligibility criterion for determining the monetary size of the voucher award
offered each student. Students who meet the low- income definition qualify
for a voucher amount equal to

9 Federal poverty guidelines are issued each year in the Federal Register by
the U. S. Department of Health and Human Services and are used in
determining financial eligibility for certain federal programs. Some state
and local governments have chosen to use the federal poverty guidelines in
some of their own programs and activities. The poverty guidelines are a
simplification of the poverty thresholds, a statistical version of the
federal poverty measure issued by the Census Bureau. In 1999, for example,
the poverty guideline was $16, 700 for a four- person family. Cleveland and

Milwaukee Voucher Programs Have Similar Student and Private School
Eligibility Criteria

Voucher Programs Have Similar Residency and Income Requirements for Students

Page 10 GAO- 01- 914 Publicly Funded Voucher Programs

90 percent of school tuition, 10 not to exceed $2,250. This voucher amount
has not changed over time. Students not meeting the low- income definition
qualify for 75 percent of the tuition amount, not to exceed $1,875. 11

For the Milwaukee voucher program, all students must reside within the city
of Milwaukee and come from families whose incomes do not exceed 175 percent
of the federal poverty guidelines. In addition, in the year prior to
entering the program, the student must have been enrolled in either a
Milwaukee public school or in kindergarten, first, second, or third grade in
a Milwaukee private school; or not enrolled in any school. The number of
students allowed to participate in the voucher program cannot exceed 15
percent of the public school district?s enrollment. Voucher students may
attend a voucher school at no charge for tuition up to an amount equal to
the lesser of the school?s per- pupil operating and debt service costs or a
state- determined maximum voucher amount. 12 For 1999- 2000, the Milwaukee
maximum voucher amount was set at $5, 106.

Cleveland private schools participating in the voucher program must be
physically located within the Cleveland school district. However, the state
also allows public schools located in any school district adjacent to the
Cleveland school district to participate in the voucher program. No public
schools have chosen to participate. Participating private schools must be
registered with the Ohio State Superintendent of Public Instruction.
Registered schools must adhere to a variety of requirements such as (1) not
discriminating on the basis of race, religion, or ethnic background; (2)
agreeing not to charge tuition to low- income voucher families in excess of
10 percent of the maximum voucher amount or the established school tuition,
if lower; and (3) permitting any such tuition over the voucher amount, at
the discretion of the parent, to be satisfied by the low- income family?s
provision of in- kind contributions or services. In addition, registered
schools must generally meet all of the state of Ohio?s minimum

10 Cleveland voucher parents may pay the remaining 10 percent of school
tuition in cash or in- kind services; Milwaukee voucher parents do not have
a similar co- pay requirement for tuition.

11 The Cleveland voucher program makes provision for an increased voucher
amount for students with special education needs, whereas the Milwaukee
voucher program has no similar provision.

12 The maximum voucher amount is based upon average per- pupil state aid
provided to the Milwaukee school district in 1998- 1999 and the annual
increase provided to public school districts statewide. Voucher Programs
Require

Private Schools To Be Located Within the City or City?s School District,
Meet State Standards, and Use Random Selection

Page 11 GAO- 01- 914 Publicly Funded Voucher Programs

standards for nonpublic schools that have been chartered by the state board
of education- which are essentially the same school standards, with some
modifications, as for public schools. These standards provide guidance and
direction on such things as a school?s educational goals, curriculum and
instruction, teacher qualifications, instructional materials and equipment,
and the quantity and quality of facilities. In addition, schools?
educational programs must be evaluated at least once every 5 years in
accordance with professionally recognized criteria and procedures. For the
1999- 2000 school year, 51 of the 52 private schools participating in the
Cleveland voucher program were chartered by the state.

Random selection of voucher students can be implemented by both the state
program office or by a school that enrolls students. For example, if the
number of Cleveland vouchers to be awarded to first- time voucher applicants
in any school year is less than the number of eligible applicants, the state
program office uses a random selection process in which lowincome applicants
are given priority. The director of the Cleveland voucher program stated
that random selection has generally been used at some point during each
year?s selection process. However, if the number of available vouchers
exceeds the number of low- income applicants, applicants above the low-
income threshold may be awarded the remaining vouchers. Once first- time
voucher applicants have been awarded a voucher, they seek enrollment in a
voucher- participating school. After enrolling voucher students who attended
a voucher school during the preceding year or siblings of those students,
schools must admit lowincome, first- time voucher students by random
selection if potential enrollees exceed the number of spaces in the school.
The school is to admit such students to kindergarten, first, second, and
third grades up to 20 percent of the total number of students enrolled in
the school during the preceding year for those grades. The extent to which
schools have used random selection for their enrollments is unknown because
the voucher program office has not monitored its use.

In the Milwaukee voucher program, participating schools must be located
within the city of Milwaukee. They must also be private schools as defined
in Wisconsin statute, which requires them to provide at least 875 hours of
instruction each school year and to have a sequentially progressive
curriculum of instruction in subjects such as mathematics and reading.
Schools must also meet applicable health and safety codes; meet at least one
of the state performance standards, such as for academic progress or
attendance; and comply with federal antidiscrimination laws. Participating
schools are subject to uniform financial accounting standards and must
submit an annual independent financial audit to the state. Similar to the

Page 12 GAO- 01- 914 Publicly Funded Voucher Programs

Cleveland program, a key provision is the use of a random selection process
when the number of eligible applicants exceeds the number of spaces a school
has designated for students. Each school has discretion in setting the
number of voucher students it will accommodate in each grade and must
specify this number at the time it notifies the state of its intent to
participate in the program. Each school must also submit an annual written
plan describing its intended method for randomly selecting voucher students
when the number of applicant voucher students exceeds the number of
available spaces allocated for them. As in Cleveland, the extent to which
schools have used random selection is unknown because schools are not
required to report on its use. The school must accept all eligible
applicants if space is available. In addition, schools cannot select
students on the basis of race, religion, gender, prior achievement, or prior
behavioral records. Continuing students and their siblings are exempt from
the random selection requirement.

Compared to public school students, voucher students in both Cleveland and
Milwaukee came from families with less income and that were more likely to
be headed by parents who were single or not married to the person they were
living with. Voucher students? parents were also more likely to have
completed at least high school than were public school students? parents.
Some research for Milwaukee also provided reliable information on students?
academic achievement prior to their participation in a voucher program. The
contract research team for Wisconsin found that voucher schools in Milwaukee
were attracting lower- performing public school students as evidenced by
their prior achievement test results. 13 We used the student characteristic
data presented by the Cleveland contract research team because their data
were more reliable than that of other researchers. With the exception of
achievement test score data, data on Milwaukee student characteristics
collected by the contract research team were less reliable, but we
corroborated some of the information on public school students. Data that
addressed school characteristics showed that in Cleveland, voucher schools
had lessexperienced teachers and smaller class sizes than public schools. No
comparable school data were collected by the contract research team for the
Milwaukee program. Other data indicate that the majority of participating
voucher schools have been religious in Cleveland since the program?s
inception, whereas the majority have been religious in

13 The results reported were for the last achievement test taken while the
student was enrolled in the Milwaukee school district. Voucher Family

Characteristics, as Well as Voucher School Attributes, Differ From Those of
Public School Families and Their Schools

Page 13 GAO- 01- 914 Publicly Funded Voucher Programs

Milwaukee since the 1998- 99 school year (when religious schools were
admitted to the program).

Student characteristics most commonly reported by the contract researchers
(excluding race) were family income, the family?s living arrangement, and
parents? education. In Cleveland, these data came from a 1999 survey, while
in Milwaukee the data came from annual surveys conducted between 1990 and
1994. 14 The Milwaukee surveys went to parents of all voucher student
applicants each year but only to parents of a random sample of public school
students- the comparison group- in 1991. 15 Since Wisconsin ended such
surveys in 1995, the number of voucher students and participating schools
has grown significantly (roughly tenfold and ninefold, respectively),
thereby potentially changing the character of the program since it was
evaluated in earlier years.

In Cleveland, average family income for voucher students was $18, 750
compared to $19,814 for families of students attending public schools. These
average incomes fall within the definition of low income under the Cleveland
voucher program for a family of two or more members. For example, under
Cleveland?s criterion for a low- income family- less than 200 percent of the
federal poverty guideline- a two- person family in 1999 would have qualified
with an annual income under $22, 120. The research

14 For Cleveland, we relied on the student and school characteristics
presented by the contract research team for school year 1998- 99 because
these data met a higher quality standard than that of other researchers.
However, we did not report findings from the contract research team for the
first 2 years of the Cleveland voucher program because the data collected
did not fully represent the voucher and public student populations. The data
for voucher students were limited to third grade in school year 1996- 97 and
fourth grade in school year 1997- 98. For the same years, the data for
public school students were limited to classmates of students who applied
for tutoring grants in the third and fourth grade.

15 The Milwaukee contract research team collected survey data for public
school students only in 1991, during the first year of the program, due to
the expense that would be incurred (given the large sample size) updating
the survey each year. Although the contract research team sent surveys to
both voucher and public school parents twice, they achieved very low
response rates, ranging from 30 to 50 percent- substantially below the 70-
to75percent response rate required for data reliability. (See app. I.) We
reported the data on voucher students because the contract research team?s
surveys were the only data source available. We corroborated their data on
family composition and income for public school students with Bureau of the
Census data reported in John F. Witte and others, The Milwaukee Parental
Choice Program- Private And Public Education in Wisconsin: Implications for
the Choice Debate (Madison, Wisc.: Robert M. La Follette Institute of Public
Affairs, University of Wisconsin- Madison, 1995). We were unable to
corroborate information on parental educational attainment because available
Census data did not contain comparable measures of educational status.
Voucher Families Had Less

Income, Were Headed More Frequently by Single Parents, and Usually Had More
Parental Education Than Public School Families

Page 14 GAO- 01- 914 Publicly Funded Voucher Programs

team reported that 70 percent of voucher families were headed by a single
mother, compared to 62- 65 percent for public school families. Despite lower
incomes and a higher rate of single- mother households for voucher students,
voucher student parents had a higher level of education than did the parents
of public school students. For example, 91.6 percent of voucher student
mothers had completed high school compared to 78.1 percent for mothers of
public school students. In addition, 14.2 percent of voucher mothers had a
4- year postsecondary degree compared to 7.8 percent of public school
mothers (see table 1).

Table 1: Characteristics of Cleveland Families With Students in the Voucher
Program or in the Public Schools

Student family characteristics Voucher families a (1998- 1999)

Public school families b (1998- 1999)

Families headed by a single mother 70% 62- 65% Mean family income $18,750
$19,814 Mother completed high school 91.6% 78.1% Father completed high
school 89.2% 77.7% Mother completed 4- yr postsecondary degree 14.2% 7.8%
Father completed 4- yr postsecondary degree 12.1% 8.1% a Greene, Howell, and
Peterson also collected reliable data, in the summer and fall of 1998, for a
random sample of Cleveland voucher students for the same family
characteristics with the exception of father?s education. See Jay P. Greene,
William G. Howell, and Paul E. Peterson, An Evaluation of the Cleveland
Voucher Program After Two Years (Harvard University, 1999). b This data
comes from a random sample of families identified across grades kindergarten
through five

for public school students that had not applied for a school voucher.
Source: Kim K. Metcalf, Evaluation of the Cleveland Scholarship and Tutoring
Grant Program 1996- 1999 (Bloomington: Indiana University, 1999).

In Milwaukee, the average voucher family annual income was $11,340 in the
first 5 years of the program. The comparison group, 1991 public school
students, had a family income that averaged $22,000 in 1991. Average voucher
family incomes were less than the program?s low- income requirement for a
family of two or more members. For example, under Milwaukee?s criterion for
a low- income family, 175 percent of the federal poverty guideline, a two-
person family in 1990 would have qualified with an annual income under
$14,735. 16 Voucher families were also more likely

16 Since the contract research team obtained family income statistics for
years 1990 to1994, we provided the voucher program?s low- income threshold
for a two- person family in 1990 ($ 14, 735) as a conservative comparison to
the $11,340 average voucher family income for the 5- year period. The
program?s low- income threshold for a two- person family over the 1990 to
1994 period averaged $16, 016.

Page 15 GAO- 01- 914 Publicly Funded Voucher Programs

to be headed by a nonmarried parent (76.5 percent) than public school
families (49 percent). As shown in table 2, 84.9 percent of voucher
students? mothers reported at least a high school degree or General
Education Development (GED) diploma as compared to 75 percent of mothers of
public school students (see table 2). However, fewer voucher students?
fathers completed high school or a GED (73.1 percent) than did public school
students? fathers (76 percent).

Table 2: Characteristics of Milwaukee Families With Students in the Voucher
Program or in the Public Schools

Student family characteristics Voucher families

(1990- 91 to 1994- 95) a

Public school families (1990-

91) b

Families headed by a nonmarried parent 76.5% 49% Mean family income $11,340
$22,000 Mother completed at least high school or GED 84.9% 75% c Father
completed at least high school or GED 73.1% 76% c Mother completed 4- year
postsecondary degree 8.9% 11% c Father completed 4- year postsecondary
degree 9.4% 15% c a The voucher family data are averages for the 5 school
years 1990- 1991 to 1994- 1995.

b The data comes from a random sample of families with students in grades
kindergarten through eight that did not apply for a voucher. c Data for
these categories were only available from a survey conducted by the contract
research

team. The available Census data did not contain comparable measures of
educational status to allow us to corroborate the information.

Source: GAO analysis based on John F. Witte and others, Fifth- Year Report
Milwaukee Parental Choice Program (University of Wisconsin- Madison, 1995).

Research indicates that Milwaukee voucher students already had low academic
achievement when they entered the voucher program. During the first 5 years
of the program, voucher students had lower prior achievement test results-
as measured by the Iowa Test of Basic Skills, a standardized math and
reading test given in first through eighth grade- than the average public
school student.

Only the contract research team of the Cleveland voucher program compared
private school characteristics to those of public schools for

school enrollment, numbers of teachers employed, average number of students
per classroom, and average years of teacher classroom experience. These data
were obtained from teacher and principal surveys conducted during the 1997-
98 and the 1998- 99 school years, respectively. One of the contract
researchers for the Milwaukee program conducted case studies from 1991 to
1993 as the basis for comments on staffing and Cleveland Voucher Schools

Had Smaller Enrollments and Class Sizes and Lessoverall Experienced Teachers
Than Public Schools

Page 16 GAO- 01- 914 Publicly Funded Voucher Programs

curriculum, but the data were limited to voucher schools, thereby precluding
comparison to public school characteristics.

The Cleveland data showed that private voucher schools were smaller on
average than public schools in terms of student enrollments and numbers of
teachers employed. For example, voucher schools had average student
enrollments of 201 to 300 students compared to 401 to 500 for public
schools. The average class size was somewhat smaller for voucher schools at
20.6 students compared to 23.6 for public schools. The amount of classroom
experience reported by public school teachers was significantly higher than
the classroom experience reported by their voucher school peers (14.2 years
versus 8.6 years). (See table 3.)

Table 3: Comparison of Cleveland Private Voucher and Public School
Characteristics

School characteristics Voucher schools Public schools

Average number of students enrolled- 1998- 99 201 to 300 401 to 500 Range of
student enrollments- 1998- 99 51 to over 500 150 to over 500 Average number
of full- time teachers- 1998- 99 6 to 10 21 to 25 Average number of students
per classroom- 1997- 98 20.6 23.6 Average years of teacher classroom
experience- 1997- 98 8.6 14.2

Source: Kim K. Metcalf, Evaluation of the Cleveland Scholarship and Tutoring
Grant Program 1996- 1999 (Bloomington: Indiana University, 1999).

Other data indicate that the majority of participating voucher schools in
Cleveland since the program?s inception were religious, whereas the majority
have been religious in Milwaukee since the 1998- 99 school year (when
religious schools were admitted to the program).

Page 17 GAO- 01- 914 Publicly Funded Voucher Programs

Some information about the racial and ethnic composition of Cleveland?s and
Milwaukee?s public school and voucher student populations is available, but
whether the composition has changed as a result of the voucher programs is
unclear. During school year 1998- 99, well over twothirds of the students
enrolled in Cleveland?s and Milwaukee?s voucher programs and public schools
were minority group members. Most of the minority students were African-
American. The 1998- 99 school year data are reliable, but examining changes
in racial and ethnic composition since the voucher programs? inception is
difficult for a variety of reasons. For example, data available from
existing research for the first 2 years of Cleveland?s program were
unreliable 17 or did not fully represent the voucher and public school
student population. Further, studies that have analyzed changes in the
racial and ethnic composition of voucher and public schools in both
Cleveland and Milwaukee did not examine factors other than the voucher
program, such as birth rates, that may have influenced the changes.

Research on Cleveland?s voucher program provides information on the racial
and ethnic composition of Cleveland?s public school and voucher student
populations in school year 1998- 99, the most recent year for which reliable
information is available. As shown in table 4, of Cleveland students in
kindergarten through fifth grade, most of the public school students and
students enrolled in the voucher program in school year 1998- 99 were
minority group members.

Table 4: Racial and Ethnic Composition of Cleveland Public School and
Voucher Students, School Year 1998- 99

Racial and ethnic group Public school students

(percent) Voucher students (percent)

Minority 79 a 73.4 White 21 26.6 a Among applicants who did not receive a
voucher, 77 percent were minority group members and 23 percent were white.
Some of these students may have attended private school prior to applying
for the voucher program.

Source: Kim K. Metcalf, Evaluation of the Cleveland Scholarship and Tutoring
Grant Program 1996- 1999 (Bloomington: Indiana University, 1999).

17 We determined the reliability of data sources we considered for analysis
by assessing the response rate of survey data and the completeness of
administrative data. The criteria we used for assessing data quality are
discussed in app. I. Minority Group

Members Predominate in School Populations, but Whether Voucher Programs Have
Changed Schools? Racial Composition is Unclear

Page 18 GAO- 01- 914 Publicly Funded Voucher Programs

However, data available for the first 2 years of the Cleveland program that
would indicate whether the racial and ethnic composition of public school
and voucher students has changed over the course of the Cleveland voucher
program were unreliable or did not fully represent the voucher and public
student populations. For example, data collected by the contracted Cleveland
evaluation for voucher students were limited to third graders in school year
1996- 97 and fourth graders in school year 1997- 98. 18 For the same years,
the data collected for public school students were limited to classmates of
students who applied for tutoring grants in the third and fourth grade. 19
Although the evaluator reported the proportion of all voucher students who
were minority group members, and that, over the first 3 years, 60 percent
were African- American, he did not report the composition of other minority
groups. However, a survey conducted by another research team provided racial
and ethnic composition data for voucher students in school year 1996- 97. 20
This team reported that of the voucher students, 61.3 percent were African-
American, 4.4 percent were Hispanic, 1.4 percent were some other minority
group, 4 percent were multiracial, and 28.9 percent were white. 21

Of Milwaukee public school and voucher students, African- American students
were the majority, but the proportion of African- American students in both
the public school and voucher program student body has changed over the
course of the voucher program. Research on the Milwaukee voucher program
provided reliable data about Milwaukee public school students during two
time periods: the beginning of the

18 These data were collected to analyze the effect of the Cleveland voucher
program on students? academic achievement. According to the evaluator,
complete administrative records on students? test scores in the previous
grade, which were needed for the analysis, were available only for third-
grade students. The same group of students was evaluated in the second year.
See app. IV for a detailed description of the research.

19 The limitation of public school students to classmates of students who
applied for tutoring grants is discussed in app. IV. 20 These data were
collected for a survey that gathered information on race and other
demographic characteristics. The survey was partially funded by the Ohio
Department of Education. See Jay P. Greene, William G. Howell and Paul E.
Peterson, Lessons from the Cleveland Scholarship Program, Program on
Education Policy and Governance, Taubman Center on State and Local
Government and Center for American Political Studies, Harvard University,
October 15, 1997.

21 The Cleveland Municipal School District provided additional data on
public school students? racial and ethnic composition for school years 1996-
97 and 1999- 2000. The Cleveland Scholarship and Tutoring Program Office
provided similar data for 1999- 2000 voucher students. See app. III.

Page 19 GAO- 01- 914 Publicly Funded Voucher Programs

program and school year 1998- 99. Of public school students, about 71
percent were minority group members in school year 1990- 91, the first year
of the voucher program. African- American students represented 55 percent of
the total. By school year 1998- 99, minority group members represented
almost 80 percent of public school students and AfricanAmerican students
represented 61. 4 percent of the total. The detailed racial and ethnic
composition of Milwaukee public school students for these years, including
minority subgroup composition, is shown in table 5. Data for the intervening
years were not reported in the voucher program research.

Somewhat more information was available on the racial and ethnic composition
of Milwaukee voucher students. Table 5 shows the average racial and ethnic
composition of enrolled voucher students for school years 1990- 91 to 1992-
93, the composition in school year 1994- 95- before the program was changed
to permit religious school participation- and the composition in school year
1998- 99, the first school year the court allowed voucher students to attend
religious schools. These data, and the data on Milwaukee public school
students, describe the racial and ethnic composition of Milwaukee students
at different stages of the voucher program and indicate that some changes in
the composition have occurred. For example, of voucher students, 96.5
percent were minority group members in school year 1994- 95. By school year
1998- 99, after religious schools were admitted to the program, 79 percent
of voucher students were minority group members. However, the data do not
explain why the changes occurred. Table 5 does show that, of both public
school and voucher students, African- Americans were the largest minority
group in all time periods.

Page 20 GAO- 01- 914 Publicly Funded Voucher Programs

Table 5: Racial and Ethnic Composition of Milwaukee Public School and
Voucher Students Public school students Voucher students

Racial and ethnic group School year 1990- 91 a (percent) School year

1998- 99 (percent) School years

1991- 93 b (percent)

School year 1994- 95 (percent)

School year 1998- 99 c (percent)

African- American 55 61.4 72 72.2 62.4 Asian 4 4.1 0 Not reported 2. 4
Hispanic 10 13.3 20 23.6 13.2 Native American 1 1. 0 1 Not reported 0. 7
Other 1 Not reported 1 0.7 0. 3 Subtotal- minority groups 71 79.8 94 96.5 79
White 29 20.2 6 3.5 18.8

a The percentages for school year 1990- 1991 are for public school students
in first through eighth grades. b These data report the average racial and
ethnic composition of Milwaukee voucher students during

school years 1990- 1991 through 1992- 1993. c The 1998- 99 percentages for
voucher students were based on a head count of the total number of

students for whom ethnicity was reported in response to the Wisconsin
Legislative Audit Bureau?s survey. These computations excluded 134 students-
representing 2.2 percent of all voucher students- for whom racial and ethnic
composition was unknown because nine voucher schools did not report these
data in the survey conducted by the Bureau. Therefore, the percentages do
not sum to 100.

Sources: John F. Witte and others, table 5b, Fourth- Year Report: Milwaukee
Parental Choice Program (University of Wisconsin- Madison, Dec. 1994);
Wisconsin Legislative Audit Bureau, An Evaluation of Milwaukee Parental
Choice Program, Report 95- 3 (Madison, Wisc.: Feb. 1995); Wisconsin
Legislative Audit Bureau, An Evaluation: Milwaukee Parental Choice Program,
Report 00- 2 (Madison, Wisc.: Feb. 2000).

None of the contract research teams? studies addressed changes in the racial
and ethnic composition of voucher and public school students over the course
of the voucher program. However, three other studies of the Cleveland and
Milwaukee voucher programs have examined changes in the racial composition
of students at voucher and public schools but have not developed complete
explanations of the changes. They reached conclusions about the voucher
programs? effect on racial composition within voucher schools without
considering the full range of factors that could account for changes in the
composition. 22 These studies identified the proportion of white and
minority students in public schools and in voucher programs in terms of a
standard for racial isolation. A school was

22 See, for example, Charles T. Clotfelter, ?Are Whites Still Fleeing?
Racial Patterns and Enrollment Shifts in Urban Public Schools, 1987- 1996?,
Journal of Policy Analysis and Management (Spring 2001), pp. 199- 221, for a
discussion of additional demographic factors to be considered.

Page 21 GAO- 01- 914 Publicly Funded Voucher Programs

defined as racially isolated when 90 percent or more of the enrolled
students were members of a minority group or white.

One study of the Cleveland voucher program identified the proportion of
students attending racially isolated public schools in Cleveland and its
suburbs, and in private schools participating in the voucher program. For
example, at the beginning of the 1999- 2000 school year, two- fifths of
Cleveland public school students attended schools that had fewer than 10
percent white students and more than three- fifths of suburban public school
students attended schools in which the student body was more than 90 percent
white. When the researcher combined the public schools in these metropolitan
areas, he found that 60.5 percent of the students attended schools that
either had more than 90 percent or fewer than 10 percent white students. On
the other hand, among Cleveland?s voucher students, fewer than two- fifths
attended a private school that had fewer than 10 percent white students and
less than one- fifth attended a private school that had more than 90 percent
white students. On the basis of such comparisons, the researcher concluded
that school choice helps promote integration. However, factors other than
the Cleveland voucher program- such as all population groups? moves into and
out of the city, their birth and death rates, and students? movement among
schools and school systems- that contributed to the racial and ethnic
composition of Cleveland?s public and private schools were not identified or
isolated in the analysis.

Two studies of the Milwaukee voucher program examined the proportion of
public school students and voucher students who attended racially isolated
schools and reached conclusions about the effect of the Milwaukee voucher
program on voucher students? racial isolation. One study examined the
proportion of students attending racially isolated schools in the 1998- 99
school year and found that approximately 20 percent more Milwaukee public
school students attended racially isolated schools than did voucher students
who attended 26 Catholic elementary schools. The authors concluded that the
Milwaukee voucher program appeared to have increased racial and ethnic
enrollment balance for students participating in the program and for
students at participating private schools. However, the 26 Catholic
elementary schools examined in this study were not selected randomly and
represented only 41 percent of the 63 religious schools participating in the
voucher program in the 1998- 99 school year.

The second study, which examined the proportion of Milwaukee students who
attended racially isolated schools in the 1999- 2000 school year, found that
50.3 percent of Milwaukee public school students attended schools

Page 22 GAO- 01- 914 Publicly Funded Voucher Programs

that were racially isolated. Among the 86 private schools participating in
the voucher program that year, students attending religiously affiliated
voucher schools had a different experience than students attending voucher
schools with no religious affiliation. Among the 56 religiously affiliated
voucher schools, 30. 1 percent of the students attended racially isolated
schools. Among the 30 private voucher schools with no religious affiliation,
83.1 percent of students attended racially isolated schools. The authors
concluded that the addition of religiously affiliated schools had led to a
lower level of racial isolation in private schools participating in the
voucher program than in Milwaukee public schools. However, neither this
study, nor the first study of Milwaukee students? racial isolation, ruled
out routinely occurring demographic changes, such as births, deaths, moves
into and out of the city, or students? movement among schools and school
systems, as factors contributing to the proportion of racially isolated
schools they identified.

Ohio and Wisconsin use different methods to fund their school voucher
programs and spend less on each voucher student than on each public school
student. Ohio funds the Cleveland voucher program with Disadvantaged Pupil
Impact Aid moneys appropriated from the state?s general revenue funds and
reduces the Cleveland school district?s state revenues by the amount of the
voucher program appropriation. Wisconsin funds its voucher program with
general state aid and reduces the Milwaukee school district?s state revenues
by half the amount of the program cost. The full impact of these funding
methods on the public schools is unknown. In the 1999- 2000 school year,
Ohio spent $1,832 per voucher student compared to $4,910 for each student in
the Cleveland school district. For the same year, Wisconsin spent $5,106 per
voucher student compared to $6,011 for each student in the Milwaukee school
district. Public school students in both Cleveland and Milwaukee receive
additional support from local taxes and federal sources, which results in a
larger difference in per pupil amounts between voucher and public school
students than the states? figures indicate.

The Cleveland voucher program is funded from the Cleveland public school
district?s share of state Disadvantaged Pupil Impact Aid, based on an annual
appropriation determined by the Ohio legislature. For the 1999- 2000 school
year, the legislature appropriated $11.2 million for the Cleveland voucher
program. Based on this appropriation, the Cleveland school district?s $80. 5
million in Disadvantaged Pupil Impact Aid was reduced by $11.2 million to
$69.3 million. In the context of the school district?s revenues from all
sources for 1999- 2000, the $11.2 million States Fund Voucher

Programs Differently and Spend Less on Each Voucher Student Than on Each
Public School Student

Ohio and Wisconsin Fund Voucher Programs From Different Sources

Page 23 GAO- 01- 914 Publicly Funded Voucher Programs

amounted to nearly 1.6 percent of the district?s $712. 1 million total.
Actual voucher program expenditures were $6.2 million- only 55. 4 percent of
what was appropriated. 23 Voucher program expenditures are charged to a
designated state account and the Cleveland school district does not monitor
the program?s expenditures. School district officials stated that the
district has not obtained additional property tax levies for the purpose of
recovering state revenue deductions from the district?s Disadvantaged Pupil
Impact Aid funds. According to these officials, the last major school levy
for funding school operations was passed in 1996 and provided $67 million to
the district annually over a period of 5 years.

The state of Wisconsin funds the Milwaukee voucher program from a separate
general- purpose revenue appropriation. The state deducts the amount of the
appropriation from general school aid payments to all 426 school districts
statewide. 24 Once the state determines the total amount needed to fund the
voucher program for the year, it reduces the aid payable to the Milwaukee
public school district by half that amount. The other half of program
funding is drawn from aid authorized for the remaining 425 school districts
in proportion to the total state aid to which each district is entitled. The
school districts have the option of increasing property tax levies to offset
reductions in general state aid related to the voucher program. 25 According
to a Milwaukee school district official, the district has generally levied
taxes to the maximum extent possible under state school revenue limits. For
the 1999- 2000 school year, the Milwaukee school district absorbed half of
the voucher program?s $38.9 million cost.

23 The $6.2 million cost of the voucher program excludes $476, 243 that the
Cleveland Scholarship and Tutoring Program Office paid for tutoring grants
to students attending Cleveland public schools in 1999- 2000. According to
the Ohio Department of Education?s Director of School Finance, the
appropriated funds not spent on the voucher or tutoring programs were used
for other state purposes.

24 Until the 1999- 2000 school year, the state funded the voucher program
from equalization aid paid to the Milwaukee school district. Equalization
aid is the largest state school aid program and is intended to even out the
differences in property tax base per student among school districts.

25 A May 2001 Wisconsin Legislative Fiscal Bureau report stated that, under
current funding law for the program- which allows districts to levy to
offset the aid reduction and which also increases state aid by two- thirds
of the amount of the levy increase- could result in either increased or
decreased equalization aid, after other state aid calculations are
performed. However, the Fiscal Bureau report also explained that the overall
effect of the increases or decreases on school district revenues would be
zero because districts would find increases in state aid offset by matching
reductions in their local school property tax levies. Conversely, districts
with decreases in state aid would find them offset by matching increases in
their local school levies.

Page 24 GAO- 01- 914 Publicly Funded Voucher Programs

That amount, $19.45 million, represented about 2.1 percent of the district?s
$917 million in total revenues.

Because there are no definitive studies, state and school district officials
did not have definitive explanations as to what extent the voucher programs
negatively or positively affected the Cleveland or Milwaukee public school
districts. In Cleveland, with the exception of a public accounting firm?s
management study touching on this issue, state and school district officials
were unaware of any studies addressing the financial impact of the voucher
program. 26 Official and unofficial studies of the Milwaukee voucher program
have described possible effects ranging from slightly negative to
indeterminate. According to some of these studies, changing the assumptions
of such studies could modify the results. Assumptions, for example, could
include estimates of the number of voucher students who were formerly
enrolled in the public school districts and where they might have been
enrolled in the absence of a voucher program. In addition, the amount of
funding that the Milwaukee public school district has received from state
revenues and local property tax levies has been affected by policy decisions
that have not necessarily been driven by the voucher programs. For example,
the Milwaukee public school district has experienced an increase in total
state aid, largely because of the state?s policy of funding two- thirds of
certain school costs beginning in the 1996- 97 school year. 27 In Cleveland,
local school revenues are not based on enrollments. Consequently, when
students leave public schools to attend private schools, the public school
retains the same amount of local revenue and thus has a higher expenditure
of local funds per pupil. However, in Milwaukee, the amount that may be
contributed from the local tax levy is determined by the difference between
the school revenue cap and state school aid, which are based on the school
district?s enrollment.

26 A September 1999 KMPG LLP management study on the Cleveland voucher
program stated that the state funds allotted to the voucher program did not
affect the Cleveland school district because the state aid cap limited the
aid allotted to the Cleveland schools. The KMPG conclusion notwithstanding,
state finance and school district officials said that they could not
definitively say that any revenues lost to the school district due to the
state aid cap were amounts that had been appropriated for the Cleveland
voucher program.

27 For a more complete discussion of the pattern of state aid and property
tax revenue going to the Milwaukee school district between 1990 and 1999 see
Howard L. Fuller and George A. Mitchell, The Fiscal Impact of School Choice
on the Milwaukee Public Schools (Marquette University, Mar. 1999).

Page 25 GAO- 01- 914 Publicly Funded Voucher Programs

Ohio provides less state revenue for each voucher student than for each
public school student in the Cleveland school district. For example, on a
per- student basis, the state spent $1,832 on voucher payments for each
voucher student and program administration, compared to $4,910 for each
Cleveland school district student for the 1999- 2000 school year. The $4,910
per public school student paid by the state does not include the per student
amounts of $3,212 in local taxes and $745 in federal funds that were
received by the Cleveland school district for the same year.

Two factors may help to explain why the amount spent by the state for
voucher students was only about 37 percent of the amount the state spent for
public school students in Cleveland. First, the private schools
participating in the program generally have low tuition. For example, the
estimated average voucher amount for low- income students at 33 Catholic
schools was $1,592 in 1999- 2000, which is well below the maximum voucher
amount of $2,250. 28 Several representatives from participating religious
schools stated that their schools? missions were to provide a private-
school education to children in their communities, many of whom come from
low- income families. The schools purposely subsidize the cost of educating
all enrolled students to achieve this mission. Representatives from
nonreligious schools with higher tuition (about $4, 000) stated that they
could afford to accommodate just a few voucher students because they must
find corporate or other sponsors to subsidize the difference between the
maximum voucher amount allowed and the tuition charged. Second, the maximum
voucher amount ($ 2,250 for low- income students) established by the Ohio
legislature at the beginning of the voucher program appears to have limited
the program primarily to low- tuition religious schools. The 6th Circuit U.
S. Court of Appeals stated in December 2000 that, practically speaking, the
tuition restrictions mandated by the statute limit the ability of
nonreligious private schools to participate in the program, since religious
schools often have lower overhead costs, supplemental income from private
donations, and consequently lower tuition needs. In the 1999- 2000 school
year, 90 percent of the participating schools were religious and 97 percent
of the voucher students attended these schools.

28 The $1,592 average voucher amount is a conservative estimate, based on
the higher tuition fees charged to nonparish families. At 20 of the 33
Catholic schools in the voucher program, students of parish families pay
tuition ranging from 16 to 51 percent lower than that of nonparish families.
As a result, depending upon the number of voucher students who come from
parish families, the estimated average voucher amount would be reduced. Both
Ohio and Wisconsin

Spend Less on Each Voucher Student Than on Each Public School Student

Page 26 GAO- 01- 914 Publicly Funded Voucher Programs

Wisconsin also provides less state revenue for voucher students than for
public school students in the Milwaukee school district. For 1999- 2000, the
estimated number of voucher students was 7,621- therefore the total budgeted
amount for just the cost of voucher payments was about $38.9 million, or
$5,106 per voucher student. By comparison, this per- student voucher amount
is about 85 percent of the $6,011 per student in state aid received by the
Milwaukee school district. The $6, 011 per public school student paid by the
state does not include the per student amounts of $1,573 in local taxes and
$1,073 in federal funds that were received by the Milwaukee school district
for the 1999- 2000 school year.

The Wisconsin Department of Public Instruction, which administers the
voucher program, establishes its budget for the voucher program in two
steps. First, it computes a set amount per student- the amount paid in the
previous school year to voucher schools plus the amount of per- student
revenue increase provided to public school districts taking into account
revenue limits in the current year. For example, the 1999- 2000 per- student
payment of $5,106 was based on the 1998- 99 per- student payment of $4,894
plus $212, the statewide per- student increase. The second step is
estimating the number of students who will participate in the voucher
program. This estimate comes from participating schools? annual estimates of
the number of voucher students they intend to admit in the next school year.
The Department adjusts this estimate based on its experience with the
accuracy of schools? projections in prior years.

Page 27 GAO- 01- 914 Publicly Funded Voucher Programs

The contract researcher teams for Cleveland and Milwaukee found little or no
statistically significant differences in voucher students? achievement test
scores compared to public school students, but other investigators found
that voucher students did better in some subject areas tested. 29 None of
the findings can be considered definitive because the researchers obtained
different results when they used different methods to compensate for
weaknesses in the data. Most of the studies satisfied basic criteria for
research quality, such as using study designs and data analysis methods that
isolate the program?s effect, but suffered from missing test score data and
low survey response rates. For example, scores from incompatible tests
limited the contracted Cleveland evaluation in the first year. In Milwaukee,
the contracted evaluations had low response rates for survey data and
missing test scores due to school policy changes. In addition, a substantial
proportion of students left the voucher program or left the Milwaukee public
school system when they were not selected for a voucher. The loss of these
students made it difficult to design a rigorous evaluation. The researchers?
different findings likely were due to the different study designs,
comparison groups, and statistical tests they used to address these
limitations.

The contract research team found no statistically significant difference in
the academic achievement test scores of Cleveland voucher and public school
students at the end of the first year of the program, school year 1996- 97,
when they controlled for differences in background- but not classroom-
characteristics that might affect their performance. 30 At the end of the
second year, school year 1997- 98, the evaluator found that voucher
students? scores in language achievement- one of six subject areas tested-
were higher than those of public school students when previous academic
achievement, background, and classroom characteristics were controlled. In
contrast, the test scores of the voucher students in the two additional
private schools, which the evaluator was

29 A description of the contracted research teams and other teams who
evaluated the Cleveland and Milwaukee voucher programs is included in the
report?s background section. All three teams that evaluated the Milwaukee
program analyzed the data collected by the contract research team.

30 The contract research team did not collect data on classroom
characteristics the first year of the evaluation. Contract Researchers

Found Little or No Significant Improvement in Voucher Students? Achievement,
but Other Investigators Found Some Positive Effects

Cleveland?s Contract Research Team Found Little Improvement After 2 Years

Page 28 GAO- 01- 914 Publicly Funded Voucher Programs

able to include in the second- year analyses, were lower than those of
public school students in every area, at a statistically significant level.
31

The Harvard team?s reanalysis of the contract research team?s data for the
first year of the program did not control completely for influences on
student achievement other than the voucher program. The team used a
statistical analysis method that allowed them to isolate the effect of the
voucher program, but they did not include all potential influences in the
analysis. Two Harvard team analyses of Hope school students? achievement
test scores in the first 2 years of the voucher program also identified
changes in the scores. However, neither of these two studies ruled out any
student or classroom characteristics that may have influenced the direction
of those changes. Because these three studies did not meet our criteria for
analyses of the effect of the voucher program, their findings are not
reported here. The findings and the methodological strengths and weaknesses
of the contract research team?s and the Harvard team?s research are
described in greater detail in appendix IV.

31 The two private schools, known as the Hope schools (Hope Ohio City School
and Hope Central Academy), were established especially for the voucher
program. Approximately 15 percent of the students in the Cleveland voucher
program attended these schools. In the first year of the program, the
contract research team had analyzed the Hope school voucher students?
achievement separately because he did not consider the test scores available
for this group compatible with those of the other voucher students. The Hope
school students had taken a different test. In the second year, the contract
research team was able to administer the same achievement test to all
students being evaluated, including *students from the two Hope schools.

Page 29 GAO- 01- 914 Publicly Funded Voucher Programs

Milwaukee?s contract research team concluded that there was no consistent
evidence that Milwaukee?s voucher program had positively or negatively
affected student achievement. 32 The team used three comparison groups and
multivariate analysis methods that controlled for prior student achievement
and student and family characteristics to isolate the program?s effect. 33
They adjusted the sample survey data on students? and families? background
characteristics for low survey response rates, and estimated test scores in
the fourth year of the program- when test score data were missing for about
two- thirds of the sample- to improve the reliability of their estimates. 34
They also examined whether the substantial proportion of students who left
the voucher program or who left the Milwaukee public school system when they
were not selected for a voucher was affecting their analysis of achievement
of students who remained in the voucher program and in Milwaukee public
schools. They concluded that losing these students made it difficult to be
certain about the differences between students? scores.

32 As we discuss in the background section of the report, the Milwaukee
voucher program evaluation covered only the first 4 years of the program,
school years 1990- 91 through 1993- 94, when its size and the private
schools participating were different than today.

33 The contract research team did not collect data on classroom
characteristics, such as class size, and thus did not rule out classroom
characteristics as a contributing factor in the analysis of Milwaukee
students? achievement. Because data on classroom characteristics were not
included in the public use data base that the Harvard team and the Princeton
researcher used, they were unable to examine classroom characteristics?
contribution to Milwaukee students? achievement as well. However, in
?Schools and Student Achievement: More Evidence from the Milwaukee Parental
Choice Program?, Economic Policy Review (March 1998), the Princeton
researcher provides indirect evidence that class size may explain results.

34 Although the contract research team sent its surveys to voucher and
public school parents twice, they achieved very low response rates, ranging
from 30 to 50 percent. Because the team had independent measures of race and
qualification for free lunch (which they used as a surrogate measure for
family income) from the Milwaukee Student Record Data Base for both voucher
and public school students, they were able to assess sampling bias and
construct weights to offset that bias. The contract researcher compensated
for missing math scores by a different method. Milwaukee public school
students take a number of tests required by the federal Title I program. The
requirements include testing in every grade in reading and math using a
standardized achievement test. In school year 1993- 94, Title I regulations
changed from requiring total math, consisting of three subtests, to the
?problem- solving? test. As a result, the Milwaukee public school system
stopped using all three subtests for some students. Because the correlation
between the problem- solving component and the total math scores was high,
the contract researcher was able to estimate the total math score for
students who took only the problem- solving component. Alternative Research

Approaches Result in Three Different Sets of Findings About Milwaukee
Voucher Students? Academic Achievement

Page 30 GAO- 01- 914 Publicly Funded Voucher Programs

The Harvard team found improvements in voucher students? language and math
scores. This team was the first to use a study design and multivariate
analysis procedure that reproduced the Milwaukee voucher program assignment
process, assuming that it was random, and to use nonselected voucher
applicants as a comparison group. Under this study design, the Harvard team
isolated the effect of the voucher program by controlling for factors
related to students? assignment to schools. 35 However, this design was
unable to account precisely for departures from random assignment to the
voucher program and the team did not test their assumption of random
assignment completely by analyzing whether applicants not selected for the
voucher program who left the public school system were different from the
nonselected applicants who remained. To identify improvements in students?
scores, they used a statistical test that assumed a change in voucher
students? achievement would be more favorable than would a change in the
comparison group?s and, for some results, used confidence levels that were
less stringent than conventional standards. Moreover, the analyses of
students who left the voucher program and the Milwaukee public school system
that the contract research team conducted, and additional analyses included
in the Princeton researcher?s evaluation, cast doubt on whether the students
remaining in the study samples over the 4 years being analyzed could be
considered randomly assigned. These findings also call into question the
Harvard team?s findings of improvements in students? test scores.

The Princeton researcher found positive effects of the Milwaukee voucher
program on students? achievement in math but not in reading. Like the
Harvard team?s research, the Princeton researcher?s study design focused on
voucher program applicants, but did not assume that voucher recipients had
been randomly selected for the voucher program. 36 The researcher used a
multivariate analysis procedure that estimated differences in achievement
between voucher students and students in two comparison groups after
controlling for all observed and unobserved fixed student characteristics,
including background characteristics and prior achievement. She used both
nonselected voucher applicants and a random

35 The Harvard team used an experimental design, which required random
assignment of students to the group receiving program services and to the
control group to isolate the program?s effect. In the Harvard study, voucher
students were the group that received the program and applicants not
selected for a voucher were the control group.

36 The Princeton researcher used a quasiexperimental design that did not
require that the program or comparison group be selected randomly to isolate
the program?s effect, but did require statistical controls for factors other
than the program that may have influenced students? achievement.

Page 31 GAO- 01- 914 Publicly Funded Voucher Programs

sample of Milwaukee public school students as comparison groups. The
Princeton researcher estimated missing test scores, allowed for the
dependence of later scores on earlier ones and analyzed whether the
proportion of students who left the voucher program or who left the public
school system because they were not selected for a voucher affected her
estimates of student achievement. Her tests showed that there were
systematic differences in the students in her analysis groups but that her
statistical procedures had controlled for these differences to the extent
possible with statistical methods. Her findings were consistent using either
comparison group.

The student achievement research we reviewed for the Milwaukee voucher
program was reported in four major studies. The findings and the
methodological strengths and weaknesses of these studies are described in
greater detail in appendix V.

From a national policy perspective, school choice has become a frequent
topic of discussion as a way of delivering elementary and secondary
education to the nation?s youth and giving parents more control over their
children?s education. Although voucher programs represent a small segment of
school choice options, interest in the academic achievement of voucher
program students is likely to continue and new evaluations of voucher
program initiatives may be undertaken in the future. The studies we reviewed
offer some useful lessons on the difficulties in achieving definitive
assessments of voucher programs and of other alternative education programs
targeted to low- income or disadvantaged students.

First, reliance on administrative data for achievement test scores and
student background information can conserve time and resources in data
collection where school records are complete and the data system is
automated, as in Milwaukee. However, even when complete and automated
records are available, reliance on scores from schooladministered tests can
result in data gaps if the school district changes its testing policy, as
did the Milwaukee system. On the other hand, when the evaluation team
selects and administers the achievement tests, as in Cleveland, the
cooperation of all schools in the study population must be negotiated. The
separate analysis of Cleveland?s Hope school results in the first- year
evaluation, which the contract research team felt was required because the
schools had not yet agreed to be tested, limited the applicability of the
first- year findings.

Second, the Milwaukee team?s experience with survey data collection from the
program?s low- income families confirms that special data Concluding

Observations

Page 32 GAO- 01- 914 Publicly Funded Voucher Programs

collection and followup procedures are needed to achieve survey response
rates that meet minimum data quality standards when low- income households
are members of the study population. For example, although the Milwaukee
team sent its survey twice to voucher and public school parents, response
rates were very low- from 30 to 50 percent. Additional strategies, such as
offering respondents a cash incentive and conducting several rounds of
follow- up by telephone with nonrespondents, may increase response rates
further.

Finally, vital information about voucher program performance may be lost if
adequate funding is not provided for program evaluations. For example,
Wisconsin has not funded voucher student academic achievement evaluations
since 1995, thereby losing data on program performance during the years when
the program had grown the most. Because such school choice initiatives are
of national interest, it would be useful to have more definitive research
about their effect. Through its role as a sponsor of research on education
programs, the Department of Education can encourage state departments of
instruction and others interested in the outcomes of voucher programs to
conduct additional research of a quality that leads to conclusive findings
on emerging programs.

We obtained comments on a draft of this report from Education, the Ohio
Department of Education, the Wisconsin Department of Public Instruction, and
the Wisconsin Legislative Audit Bureau. These entities provided several
technical clarifications, which we incorporated as appropriate. In addition,
the Legislative Audit Bureau questioned our description of the use of local
tax levies to offset the cost of the voucher program in Milwaukee. We
obtained and added clarifying information from the Milwaukee school
district. We also obtained comments from the researchers whose work we
assessed. Both Education and the Harvard researchers commented that we did
not mention research studies on privately funded voucher programs. We
anticipate initiating a review of these programs shortly. The Harvard
researchers also commented that we did not mention other research on the
Cleveland and Milwaukee voucher programs covering subjects such as parental
satisfaction and the effect of voucher schools? competition on public
schools. While we recognize such research exists, we focused on those topics
of greatest concern to our requestor.

Most of the researchers also provided technical comments that we
incorporated as appropriate. The contracted researchers for Cleveland and
Milwaukee generally agreed with the findings in the report. The Princeton
researcher generally agreed with the findings, but questioned our Agency and
Other

Comments

Page 33 GAO- 01- 914 Publicly Funded Voucher Programs

summary of the differences among the studies? findings. However, her
analysis of differences focused only on the differences between her work and
each of the other researchers, whereas our assessment included the
comparisons she made as well as a comparison of the differences between the
Milwaukee contract researchers? and the Harvard researchers? findings. She
also pointed out that a published version of the working paper we originally
analyzed better met our criteria for inclusion in the report. We reviewed
and included information from this article. The Harvard researchers
disagreed with our assessment of their studies and provided additional
information to support their findings about the Cleveland and Milwaukee
programs. After reviewing this information, we determined that the
additional material they provided did not support their view of our
assessment.

As agreed with your office, unless you publicly announce its contents
earlier, we plan no further distribution of this report until 30 days after
the date of this letter. At that time we will send copies of this report to
the Secretary of Education, appropriate congressional committees,
representatives of the Ohio and Wisconsin Departments of Education, and
other interested parties. If you or your staff have any questions or wish to
discuss this material further, please call me or Diana Pietrowiak at (202)
512- 7215.

Sincerely yours, Marnie S. Shaul Director, Education, Workforce,

and Income Security Issues

Appendix I: Scope and Methodology Page 34 GAO- 01- 914 Publicly Funded
Voucher Programs

Much of the public debate about the Cleveland and Milwaukee voucher programs
has concerned research findings and the quality of the research available.
In developing our report, we addressed this aspect of the debate in two
ways. First, in our analyses of student characteristics and racial and
ethnic composition, we used background data collected for studies of the
voucher programs. These studies included assessments of the racial and
ethnic composition of public school and voucher student populations and
studies designed to evaluate Cleveland and Milwaukee students? academic
achievement. Second, we included an assessment of the research on Cleveland
and Milwaukee students? academic achievement, a major outcome of interest
for the voucher programs. We selected studies for these analyses that met
two or more of the following criteria:

 The study was performed under contract to the state in which the voucher
program was implemented.

 The study was published in a peer- reviewed journal.

 The study was issued under the auspices of a research institution that
reviews work prior to release.

 The study employed quantitative data analysis to examine student academic
achievement.

We assessed both the quality of the data we used in our analyses of
students? characteristics and racial and ethnic background and the
methodology of the student academic achievement studies.

Studies of the Cleveland and Milwaukee programs have used both
administrative data collected and maintained by the school districts and
voucher program offices and surveys conducted for the studies. Most of the
studies described the completeness of the administrative data and the
elements it contained, and the methods used to conduct the surveys. The
criteria we used for assessing the data?s quality are shown in table 6.
While we recognized that the administrative data were not collected to meet
research standards and that surveys of low- income families like those
participating in the voucher programs often obtain low response, we paid
particular attention to the administrative data?s completeness and the
surveys? response rates. 1 When 30 percent or more of the administrative or
survey data were missing, we looked for analyses showing no important
difference between individuals represented in the data and those who were
not included. If such an analysis had not been conducted, we did not select
the data for our analyses, except for the analysis of Milwaukee

1 The response rate is the number of people in the survey sample who
actually responded, compared with those who were asked to respond but did
not. Appendix I: Scope and Methodology

Appendix I: Scope and Methodology Page 35 GAO- 01- 914 Publicly Funded
Voucher Programs

voucher and public school student characteristics because other data sources
were limited.

Table 6: Data Quality Criteria Survey data Administrative data

Use of a random sample Correspondence to the entire study population Sample
size greater than 30 Sample size not applicable because data are

gathered for all students Response rate of 70 to 75 percent or greater High
percentage of the study population for

whom information was located in the data base The results of nonresponse
analysis showing no important difference between individuals or families
represented in the data and those missing from the data

Comparative analyses showing no important difference between individuals or
families represented in the data and those missing from the data if 25 to 30
percent or more of the records are missing

The research on voucher students? academic achievement included both
evaluations of the voucher program?s impact on students? performance, and
analyses and papers discussing methodological issues involved in conducting
the research. 2 We reviewed the methodological papers for contextual
understanding, but our assessment of the research focused on the impact
evaluations. Our assessment included both the quality of the data used in
the evaluation and the methodological quality of the research. The criteria
we used in the assessment are shown in table 7.

Table 7: Criteria for Assessing Studies of Academic Achievement Study
component Criteria

Design For an experimental design, selecting the group receiving the program
and the control group randomly For both experimental and quasiexperimental
designs, using a comparison group Data collection Meeting the criteria for
survey and administrative data quality

shown in table 6 Data analysis For quasiexperimental design, using a
multivariate analysis

procedure For quasiexperimental design, using controls for influences other
than the program Testing and correcting for limitations such as nonrandom
selection to the program and comparison group, and missing survey and
administrative data

2 The bibliography included at the end of this report lists both types of
research.

Appendix I: Scope and Methodology Page 36 GAO- 01- 914 Publicly Funded
Voucher Programs

An impact evaluation determines a program?s effect on its participants by
isolating a program?s contribution from the effects of other influences that
could have affected participant outcomes. To isolate the program?s
influences, an impact evaluation studies two groups: those receiving program
services and a similar group not receiving program services. Researchers
compare the relevant outcomes of these two groups, such as students?
achievement test scores, to determine the program?s effect.

The criteria for study design in table 7 apply to the two types of impact
evaluation used to analyze the effect of an educational program on its
participants: an experimental design and a quasiexperimental design. The two
designs differ primarily in the way that the comparison groups are
developed. In an experimental design, the comparison group is referred to as
the control group. This group is composed of students randomly selected from
possible program participants, such as applicants to a voucher program.
Because control group members are selected randomly, researchers can compare
outcomes to determine the program?s effect without using statistical
controls for other factors that could have influenced the outcomes. In a
quasiexperimental design, the comparison group is composed of individuals
who share characteristics with program participants, but who have not been
randomly selected and who have may or may not have sought program services.
3 For example, applicants to a voucher program who did not receive a voucher
might serve as a comparison group, because they share with voucher
recipients an interest in alternative educational services. With this
design, statistical controls, such as those provided by a multivariate
analysis procedure, are needed to isolate the program from other factors
that could influence outcomes.

The same data quality criteria we discussed above were used for assessing
the administrative and survey data used in the impact evaluations. The
criteria for data analysis in table 7 refer to the need to control for
factors other than the program when program participants and comparison
group members are not randomly selected. They also encompass additional
analyses that may be needed if the group receiving program services and the
comparison group were not randomly selected or to determine if missing data
affect the reliability of the estimates of the program?s effect.

We obtained the data for our analyses of eligibility criteria and the
funding of the voucher programs from other sources. The information on the

3 See Early Childhood Programs: The Use of Impact Evaluations to Assess
Program Effects (GAO- 01- 542, Apr. 16, 2001) for a detailed description of
experimental and quasiexperimental designs.

Appendix I: Scope and Methodology Page 37 GAO- 01- 914 Publicly Funded
Voucher Programs

eligibility criteria for schools and students participating in the Cleveland
Scholarship and Tutoring Program and the Milwaukee Parental Choice Program
came from documents issued by the program offices. We also reviewed relevant
state laws and regulations.

To describe the funding of the voucher programs and compare the amounts
spent on voucher and public school students, we used information provided by
the state departments of education and the program offices, as well as
information found in program evaluations. We also examined relevant state
and school district budget and financial reports. We conducted site visits
to Ohio and Wisconsin and interviewed officials of the program offices,
school districts, state departments of education, and several private
schools to obtain their views on the financial impact of the voucher
programs.

We also interviewed the contract researchers and key researchers from
Harvard University?s Program on Education Policy and Governance and from
Princeton University. While an official evaluation of student academic
achievement in the Cleveland voucher program continues, analyses of student
academic achievement in the Milwaukee program are based on Milwaukee data
collected before 1995, when the legislature was still funding data
collection. Since student characteristic and achievement data have not been
collected for the past 7 years, conclusions reached by both the contract
researchers and other researchers may not be applicable to the current
voucher program, which has grown tenfold in the interim.

Appendix II: Information on Publicly Funded School Voucher Programs in
Florida, Maine, and Vermont

Page 38 GAO- 01- 914 Publicly Funded Voucher Programs

Objective: This program is intended to support the state constitutional
requirement that the state provide students with the opportunity to obtain a
high- quality education. Therefore, the program provides state tuition
grants to permit students attending a failing public school (that is, ?F?-

rated) to attend an eligible higher- performing public school or a private
school of choice.

Student Eligibility: Any student who spent the last year at a Florida public
school that received an ?F? rating from the state for the second time in 4
years qualifies for an Opportunity Scholarship. Also eligible are students
who did not attend an ?F?- rated school in the previous year but are now
assigned to such a school.

Private School Eligibility: A private school must be located in Florida and
may be sectarian or nonsectarian. Other requirements for private school
participation include: (1) demonstrating fiscal soundness; (2) accepting
scholarship students on an entirely random and religious- neutral basis
without regard to the student?s past academic history; (3) being subject to
the instruction, curriculum, and attendance criteria adopted by an
appropriate nonpublic school accrediting body; (4) employing or contracting
with teachers who hold a baccalaureate or higher degree, or have at least 3
years of teaching experience in public or private schools; and (5) accepting
as full tuition and fees the amount provided by the state for each student.

Maximum Student Participation: Participation is limited to the total number
of students attending or assigned to qualifying ?F?- rated schools for a
given school year. For the 1999- 2000 school year, two such schools with an
approximate total population of 900 students were designated as failing.
There were no new scholarships for the 2000- 01 school year because, as of
July 2000, no public schools received a grade of ?F? for 2 of 4 years.

Maximum Voucher Amount: The maximum voucher amount is based on (1) a
calculated amount equivalent to what would have been provided for the
student in the district school to which he or she was assigned; or (2) the
amount of the private school?s tuition and fees, whichever is less. Eligible
private school fees may include book fees, lab fees, and other fees related
to instruction, including transportation. The voucher maximum for 1999-
2000, based on the calculated costs for the two ?F? schools, was $3,353 per
student for kindergarten through third grade and $3,178 per student in
fourth through eighth grades. Appendix II: Information on Publicly Funded

School Voucher Programs in Florida, Maine, and Vermont

Florida Opportunity Scholarship Program

Appendix II: Information on Publicly Funded School Voucher Programs in
Florida, Maine, and Vermont

Page 39 GAO- 01- 914 Publicly Funded Voucher Programs

1999- 2000 Enrollment in Private Schools: The first program year was the
1999- 2000 school year. In that year, 143 out of about 900 students chose
not to attend their assigned, failing public school. Fifty- eight enrolled
in participating private schools and 85 enrolled in other, higher-
performing public schools. Total Florida enrollments for students in
kindergarten through 12th grade totaled 2,381,860 for public schools and
288,248 for private schools in 1999- 2000.

1999- 2000 Participating Private Schools: Five schools: four religious and
one nonreligious.

Status of Legal Challenges: In October 2000, the First District Court of
Appeal for the State of Florida ruled that the Opportunity Scholarship
Program was consistent with Article IX, Section 1 of the Florida
Constitution. That provision requires the state to maintain a uniform system
of free public schools. The appellate court ruling reversed a trial court
decision holding that the Opportunity Scholarship Program violated Article
IX, Section 1. In April 2001, the Florida Supreme Court declined to review
the appellate court?s ruling. The appellate court also deferred
consideration of whether the scholarship program statute was
unconstitutional under the religion clauses in the Florida and U. S.
constitutions, concluding that the trial court must first consider these
allegations.

Objective: Districts that do not have their own schools must provide tuition
to resident families for use in other schools. Students may attend a private
school approved for tuition purposes, a public school in an adjoining
district which accepts tuition students, or a school approved for tuition
purposes in another state or country upon permission of officials of the
receiving school. The ?tuitioning? system has existed in some form for over
200 years but has excluded religious schools from receiving state funds
since 1981. It especially benefits students living in the rural part of the
state.

Student Eligibility: Children of parents residing in a district which does
not maintain elementary or secondary schools.

Private School Eligibility: To receive public funds for tuition purposes, a
private school must be nonsectarian and meet other requirements for
reporting, auditing, and student assessment.

Maximum Student Participation: The number of students receiving tuition to
attend other schools depends upon the number of students in Maine Education
Tuition

Program

Appendix II: Information on Publicly Funded School Voucher Programs in
Florida, Maine, and Vermont

Page 40 GAO- 01- 914 Publicly Funded Voucher Programs

the districts without their own schools. The district pays tuition directly
to a public school or to a private school that has accepted the child, has
been selected by the child?s parents, and has been approved for tuition
purposes.

Maximum Voucher Amount: The tuition paid to a private elementary school
cannot exceed the average per- student cost in all public elementary schools
in the state for the previous year as computed by the State Education
Commissioner. For private secondary schools, the tuition paid by the
district cannot exceed the sum of the school?s allowable expenditures,
divided by the number of students at a particular school, adjusted by
certain factors; or the adjusted state average public secondary per- student
cost, whichever is lower. In the 2000- 2001 school year, the maximum tuition
rate for public elementary students attending any private school was $4,596.
For secondary students attending private schools, the amount was $5,732.

1999- 2000 Enrollment in Private Schools: The state?s total public school
enrollment was 214,985. The number of these public school students that
attended private schools with public funding was 5,614. All but 214 of the
voucher students attended secondary schools. The number of privately funded
students attending private schools was 10,394.

Status of Legal Challenges: On April 23, 1999, the Maine Supreme Court
affirmed the judgment of the Superior Court (Bagley vs. Raymond School
Department) that the exclusion of religious schools from receiving state
funds under Maine?s education tuition program does not violate any section
of the U. S. or Maine Constitution. On October 12, 1999, the U. S. Supreme
Court declined to review the ruling, allowing the lower court?s decision to
stand.

Objective: A school district that does not operate its own school or jointly
operate a school with another district or districts (a union school) must
provide for the education of its students by paying tuition to another
Vermont public school district, an out- of- state public school district, or
an approved private school. Vermont has had an educational choice system
since 1869 but prohibited the inclusion of religiously affiliated schools in
1961.

Student Eligibility: Students in grades kindergarten through 12 from
qualified districts. Vermont Education Tuition

Program

Appendix II: Information on Publicly Funded School Voucher Programs in
Florida, Maine, and Vermont

Page 41 GAO- 01- 914 Publicly Funded Voucher Programs

Private School Eligibility: A private school may operate and provide
elementary or secondary education if it obtains state approval. It must show
that it has the resources required to meet its stated objectives, including
financial capacity, qualified faculty, and physical facilities and special
services that comply with state and federal regulations.

Maximum Student Participation: Each school district decides how it will
educate its students and thus determines the number that will attend private
school. A school district that does not maintain an elementary school may
pay tuition for elementary pupils at approved private nonresidential
elementary schools. If it does not maintain a high school, it may pay
tuition for its pupils to an approved private high school.

Maximum Voucher Amount: The tuition paid to an approved private elementary
school must not exceed the lesser of (1) the average tuition of Vermont
union elementary schools or (2) the tuition charged by the public elementary
school attended by the greatest number of the district?s pupils. For
students in grades 7 and 8, the district must not pay an amount that exceeds
the average tuition of Vermont union high schools for students in grades 7
and 8. For students in grades 9- 12, the maximum is the average tuition of
union high schools for students in grades 9- 12. For the 1999- 2000 school
year, the allowable tuition for elementary pupils was $6,257; for grades 7
and 8 pupils, $6,514; and for grades 9- 12 pupils, $7,306.

1999- 2000 Enrollment in Private Schools: Total public school enrollment was
104,559 students. A Vermont Department of Education official estimated that
2,500 publicly funded students attended five private academies (designated
high schools used by districts without public high schools) and another 900
publicly funded students are enrolled in other private schools and programs.

Status of Legal Challenges: On June 11, 1999, the Vermont Supreme Court
affirmed the judgment of the Superior Court (Chittenden Town School District
v. Vermont Department of Education) that providing state aid tuition for
children at religious schools would violate a provision of the state
constitution barring compelled support for religion. On December 13, 1999,
the U. S. Supreme Court declined to hear an appeal.

Appendix III: Additional Data on Racial and Ethnic Composition of Cleveland
Schools

Page 42 GAO- 01- 914 Publicly Funded Voucher Programs

The Cleveland Municipal School District provided detailed data on public
school students? racial and ethnic composition for school years 1996- 97 and
1999- 2000. The Cleveland Scholarship and Tutoring Program Office provided
similar data for 1999- 2000 voucher students (see table 8).

Table 8: Racial and Ethnic Composition of Public School and Voucher Students
in Cleveland

Public school students a (percent)

Voucher students (percent) Racial and ethnic group 1996- 97 1999- 2000 1999-
2000

African- American 70.1 70.4 57.4 Asian 0. 9 0.6 b Hispanic 7.6 8. 3 6.1
Multiracial 0. 2 0.8 3. 8 Native American 0.3 0. 4 b Other b b 2.8 Subtotal-
minority 79.0 80.5 70.1 White 21.0 19.5 29.9 a The public school percentages
are based on students in kindergarten through the eighth grade.

b Not available. Sources: For public school students, Cleveland Municipal
School District; for 1999- 2000 voucher students, the Cleveland Scholarship
and Tutoring Program Office.

Appendix III: Additional Data on Racial and Ethnic Composition of Cleveland
Schools

Appendix IV: Strengths and Limitations of Studies Analyzing Cleveland
Students? Academic Achievement

Page 43 GAO- 01- 914 Publicly Funded Voucher Programs

Figure 1: Strengths and Limitations of Studies Analyzing Cleveland Students?
Academic Achievement

Appendix IV: Strengths and Limitations of Studies Analyzing Cleveland
Students? Academic Achievement

Contracted Studies Other Studies Who did the studies Contract researcher

Two studies of academic achievement by voucher and nonvoucher students: a, b

First study: achievement of third- graders across the district c

Second study: achievement of same students in grade 4

First study: no significant differences in scores between voucher and
nonvoucher students in year 1

Second study: all voucher students but HOPE school students scored higher in
language- one of six subject areas tested-- but voucher students in HOPE
voucher schools scored lower than public school students d

Three studies of academic achievement by voucher students: e

First study: reanalysis of third- grade voucher and nonvoucher students
covered in contract researcher?s study, and voucher students in newly-
established HOPE voucher schools

Second study: achievement of students in HOPE voucher schools

Third study: achievement of same HOPE voucher students in the following year

First study: voucher students showed statistically significant increases in
language and science-- two of six subject areas tested

Background information for many students was missing or incomplete

28 percent of the voucher students left the program in the second year

What the studies covered What the studies found Key strengths

Key limitations

Harvard researchers Compared achievement between voucher and nonvoucher
students

Accounted for possible other reasons for differences in performance, such as
family background or prior levels of achievement

Assessed whether voucher and nonvoucher students differed consistently from
each other in socioeconomic background, and whether voucher students who
remained in the program differed consistently from those who left

First study: analysis procedure was able to account for possible other
reasons for differences in voucher students? achievement, such as family
background

First, second and third studies: used the conventional standard for tests of
statistical significance

First study: excluded prior test scores when finding increases in test
scores

Second and third studies: did not control for any possible reasons for
voucher students? achievement other than the voucher program

Appendix IV: Strengths and Limitations of Studies Analyzing Cleveland
Students? Academic Achievement

Page 44 GAO- 01- 914 Publicly Funded Voucher Programs

a Students in third- grade public school classes containing two or more
students who had applied for or were participating in the tutoring
assistance component of the Cleveland Scholarship and Tutoring Program were
selected as the comparison group for the study. These classes were selected
for comparison with the voucher students because they included public school
students whose parents, like the voucher students? parents, were motivated
to apply for a supplementary educational program. Use of this comparison
group was intended to take account of nonrandom selection for the voucher
program. b The Hope schools were established especially for the voucher
program. These two schools did not

permit the contract researchers to test the achievement of their students in
the first voucher program year (1996- 97). c The contract evaluation team
administered the Terra Nova battery of achievement tests to voucher students
at all participating voucher schools except the Hope schools. Scores from
the California Achievement Tests, which the Hope schools had administered,
were used for the analysis of Hope school students? achievement. Because the
team considered these tests incompatible with the Terra Nova, scores from
the two tests were not combined in a single analysis. d The analysis
procedures controlled for prior student achievement; family income; living
arrangement;

race; gender; and school characteristics, including class size, teachers?
highest degree and years of teaching experience. e Data from the first and
third studies also are reanalyzed and reported in a single study. See
?Lessons

from the Cleveland Scholarship Program? in Paul E. Peterson and Bryan C.
Hassel, eds., Learning from School Choice, (Washington, DC: Brookings
Institution Press, 1998), pp. 357- 92. An analysis of data from the first
study, which does include prior test scores and background characteristics,
finds increases in third- grade voucher students? test scores. While this
study employs the conventional standard for tests of significance- 95
percent confidence- the statistical test used assumes that a change in
voucher students? achievement would be more favorable than would a change in
the comparison group?s. The analysis of data from the third study on changes
in HOPE school students? test scores presents more detailed findings, but,
in the comparison of test scores, does not control for any possible reasons
for voucher student?s achievement other than the voucher program.

Appendix V: Strengths and Limitations of Studies Analyzing Milwaukee
Students? Academic Achievement

Page 45 GAO- 01- 914 Publicly Funded Voucher Programs

Figure 2: Strengths and Limitations of Studies Analyzing Milwaukee Students?
Academic Achievement

Appendix V: Strengths and Limitations of Studies Analyzing Milwaukee
Students? Academic Achievement

Contracted Studies Other Studies

Contract researcher Academic achievement of voucher and nonvoucher students
during school years 1990- 94

No consistent evidence that voucher students did better or worse than public
school students in math

Analysis using nonselected applicants as a comparison group showed voucher
students ahead in math, but results perhaps affected by very low test scores
for some nonselected voucher applicants and by low- scoring voucher students
leaving the program

Same Voucher students did better in math, but not in reading Princeton
researcher

Compared voucher students, a random sample of public school students, and
nonselected voucher applicants

Controlled for all observed and unobserved student characteristics,
including prior achievement and background characteristics

Estimated missing test scores, allowed for the dependence of later test
scores on earlier ones, and assessed whether the voucher students who left
the program affected the achievement results

Assessed whether voucher students and nonselected voucher applicants
consistently differed from each other

Estimated composite math test scores for the Milwaukee public school
students in the 4th year of the program Voucher students did

better in reading and math Harvard researchers

Compared voucher students, a random sample of public school students, and
nonselected voucher applicants

Used survey responses from those who did respond as a proxy for the
responses of those who did not

Controlled for effects of background characteristics and prior achievement

Accounted for the effects of missing achievement test scores and possibility
that a student?s score on one test could affect his or her score on another
test

Determined that nonselected voucher applicants may be different because they
were likely to be lower performing students than voucher students

Originated the idea of using nonselected voucher applicants as a comparison
group for voucher students

Used a statistical procedure designed to reproduce the Milwaukee voucher
program assignment process, assuming that it was random

Who did the studies What the studies covered What the studies found

Key strengths

Same

Appendix V: Strengths and Limitations of Studies Analyzing Milwaukee
Students? Academic Achievement

Page 46 GAO- 01- 914 Publicly Funded Voucher Programs

Contracted Studies Other Studies

Parents? response to surveys was very low

Composite math score for many public school students had to be estimated

Student departures from the voucher program and public schools were
selective in ways that might affect achievement findings

Almost 50 percent of applicants not selected for a voucher left the
Milwaukee public school system

Because data on classroom characteristics were not collected, and thus not
included in the public use data base, none of the researchers were able to
rule on classroom characteristics as a contributing factor in Milwaukee
students? achievement

The composite math score for 68 percent of public school students had to be
estimated

Student departures from the voucher program and public schools were
selective in ways that might affect achievement findings Data used to
identify

similarities and differences between voucher students and nonselected
applicants were missing a large portion of the nonselected students

Statistical tests used assumed that a change in voucher students?
achievement would be more favorable than would a change in the comparison
group?s

Used less stringent standard-- 90 percent confidence-- for tests of
statistical significance

Student departures from the voucher program and public schools were
selective in ways that might affect achievement findings

Statistical procedures were unable to account precisely for departures from
random assignment to the voucher program

Key limitations

Appendix VI: GAO Contacts and Staff Acknowledgments

Page 47 GAO- 01- 914 Publicly Funded Voucher Programs

Diana M. Pietrowiak, (202) 512- 6239 Daniel C. Jacobsen, (206) 287- 4797

In addition to those named above, Sara Edmondson, Robert Miller, and Jay
Smale made key contributions to this report. Valerie Caracelli, Arthur
James, Jr., Arthur Kendall, and Douglas Sloane provided important
consultation on methodological issues for the academic achievement analysis.
Appendix VI: GAO Contacts and Staff

Acknowledgments GAO Contacts Acknowledgments

Bibliography Page 48 GAO- 01- 914 Publicly Funded Voucher Programs

Metcalf, Kim K., William Boone, Frances K. Stage, and others. A Comparative
Evaluation of the Cleveland Scholarship and Tutoring Grant Program, Year
One: 1996- 97. Bloomington: School of Education, Indiana University, 1998.

Metcalf, Kim K. Evaluation of the Cleveland Scholarship and Tutoring Grant
Program, 1996- 1999. Bloomington: The Indiana Center for Evaluation, Indiana
University, 1999.

Metcalf, Kim K., Patricia Muller, William Boone, and others. Evaluation of
the Cleveland Scholarship Program: Second Year Report (1997- 98).

Bloomington: The Indiana Center for Evaluation, Indiana University, 1998.
Greene, Jay P., William G. Howell, and Paul E. Peterson. An Evaluation of
the Cleveland Voucher Program After Two Years. Cambridge, Mass.: Program on
Education Policy and Governance, Harvard University, 1999.

---. ?Lessons From the Cleveland Scholarship Program,? Learning from School
Choice. Washington, D. C.: Brookings Institution Press, 1998, pp. 357- 392.

---. Lessons From the Cleveland Scholarship Program. Cambridge, Mass.:
Program on Education Policy and Governance, Harvard University, 1997.

---. New Findings from the Cleveland Scholarship Program: A Reanalysis of
Data From the Indiana University School of Education Evaluation. Cambridge,
Mass.: Program on Education Policy and Governance, Harvard University, 1998.
Milwaukee Voucher ProgramRelated Research Bibliography

Cleveland Voucher Program- Related Research Ohio Contract Researchers

Other Evaluators

Bibliography Page 49 GAO- 01- 914 Publicly Funded Voucher Programs

Wisconsin Legislative Audit Bureau. An Evaluation: Milwaukee Parental Choice
Program (00- 2). Madison, Wisc.: 2000.

---. An Evaluation: Milwaukee Parental Choice Program (95- 3).

Madison, Wisc.: 1995. Witte, John F. Achievement Effects of the Milwaukee
Voucher Program.

Paper presented at the 1997 American Economics Association Annual Meeting,
New Orleans, La., Jan. 4- 6, 1997.

---, Troy D. Sterr, and Christopher A. Thorn. Fifth Year Report, Milwaukee
Parental Choice Program. University of Wisconsin- Madison, 1995.

---, and Christopher A. Thorn. Fourth Year Report, The Milwaukee Parental
Choice Program. University of Wisconsin- Madison, 1994.

---. The Market Approach to Education: An Analysis of America?s First
Voucher Program. Princeton, N. J.: Princeton University Press, 2000.

---. ?The Milwaukee Voucher Experiment,? Educational Evaluation and Policy
Analysis, Vol. 20, No. 4. Winter 1998, pp. 229- 251.

Fuller, Howard L. and George A. Mitchell. The Fiscal Impact of School Choice
on the Milwaukee Public Schools. Institute for the Transformation of
Learning, Marquette University, March 1999.

Greene, Jay P., Paul E. Peterson, and Jiangtao Du. Effectiveness of School
Choice: The Milwaukee Experiment. Cambridge, Mass.: Harvard University,
1997.

---. ?School Choice in Milwaukee: A Randomized Experiment,?

Learning From School Choice. Washington, D. C.: Brookings Institution Press,
1998, pp. 335- 356.

Greene, Jay P. and Paul E. Peterson. Methodological Issues in Evaluation
Research: The Milwaukee School Choice Plan. Cambridge, Mass.: Harvard
University, Aug. 29, 1996. Milwaukee Voucher

Program- Related Research Wisconsin Contract Researchers

Other Evaluators

Bibliography Page 50 GAO- 01- 914 Publicly Funded Voucher Programs

Rouse, Cecilia E. Private School Vouchers and Student Achievement: An
Evaluation of the Milwaukee Parental Choice Program (Working Paper 5964).
Cambridge, Mass.: National Bureau of Economic Research, 1997.

---. ?Private School Vouchers and Student Achievement: An Evaluation of the
Milwaukee Parental Choice Program.? Quarterly Journal of Economics (May
1998), pp. 553- 602.

---. ?Schools and Student Achievement: More Evidence from the Milwaukee
Parental Choice Program.? FRBNY Economic Policy Review

(March 1998), pp. 61- 76.

---. School Reform in the 21st Century: A Look at the Effect of Class Size
and School Vouchers on the Academic Achievement of Minority Students
(Working Paper 440). Cambridge, Mass.: Princeton University and National
Bureau of Economic Research, Jan. 31, 2000.

Steuerle, C. Eugene, and others, ed. Vouchers and the Provision of Public
Services. Washington, D. C.: Brookings Institution Press, 2000.

(130002)

The first copy of each GAO report is free. Additional copies of reports are
$2 each. A check or money order should be made out to the Superintendent of
Documents. VISA and MasterCard credit cards are also accepted.

Orders for 100 or more copies to be mailed to a single address are
discounted 25 percent.

Orders by mail:

U. S. General Accounting Office P. O. Box 37050 Washington, DC 20013

Orders by visiting:

Room 1100 700 4 th St., NW (corner of 4 th and G Sts. NW) Washington, DC
20013

Orders by phone:

(202) 512- 6000 fax: (202) 512- 6061 TDD (202) 512- 2537

Each day, GAO issues a list of newly available reports and testimony. To
receive facsimile copies of the daily list or any list from the past 30
days, please call (202) 512- 6000 using a touchtone phone. A recorded menu
will provide information on how to obtain these lists.

Orders by Internet

For information on how to access GAO reports on the Internet, send an email
message with ?info? in the body to:

Info@ www. gao. gov or visit GAO?s World Wide Web home page at: http:// www.
gao. gov

Contact one:

 Web site: http:// www. gao. gov/ fraudnet/ fraudnet. htm

 E- mail: fraudnet@ gao. gov

 1- 800- 424- 5454 (automated answering system) Ordering Information

To Report Fraud, Waste, and Abuse in Federal Programs
*** End of document. ***