Content uploaded by R.E. Romano
Author content
All content in this area was uploaded by R.E. Romano on May 02, 2018
Content may be subject to copyright.
Journal of Economic Literature 2017, 55(2), 441–492
https://doi.org/10.1257/jel.20150679
441
1. Introduction
In the past two decades, both research and
experience have contributed to signi-
cant progress in the understanding of edu-
cational vouchers. We review both. For the
purposes of this review, we dene a voucher
to be a government-supplied coupon that is
used to offset tuition at an eligible private
school.1 Programs that distribute such vouch-
ers exhibit variation in dimensions including
1 For other reviews and discussions see Ladd (2002),
Neal (2002), McEwan (2004), Gill et al. (2007), Levin
(2008), and Rouse and Barrow (2009).
School Vouchers: A Survey of the
Economics Literature†
D E, R E. R, M U*
We review the theoretical, computational, and empirical research on school vouchers,
with a focus on the latter. Our assessment is that the evidence to date is not sufcient
to warrant recommending that vouchers be adopted on a widespread basis; however,
multiple positive ndings support continued exploration. Specically, the empirical
research on small-scale programs does not suggest that awarding students a voucher
is a systematically reliable way to improve educational outcomes, and some detri-
mental effects have been found. Nevertheless, in some settings, or for some subgroups
or outcomes, vouchers can have a substantial positive effect on those who use them.
Studies of large-scale voucher programs nd student sorting as a result of their imple-
mentation, although of varying magnitude. Evidence on both small-scale and large-
scale programs suggests that competition induced by vouchers leads public schools
to improve. Moreover, research is making progress on understanding how vouchers
may be designed to limit adverse effects from sorting, while preserving positive effects
related to competition. Finally, our sense is that work originating in a single case (e.g.,
a given country) or in a single research approach (e.g., experimental designs) will not
provide a full understanding of voucher effects; fairly wide-ranging empirical and the-
oretical work will be necessary to make progress. ( JEL H52, H75, I21, I22, I28, O15)
*
Epple: Carnegie Mellon University and NBER.
Romano: University of Florida. Urquiola: Columbia
University and NBER. We thank Roger Gordon for his
encouragement to undertake this paper, and Janet Currie
and Steven Durlauf for their invaluable support. We are
indebted to Thomas Nechyba and four anonymous refer-
ees for their numerous comments. In addition, we received
useful comments from Chelsea Cofn and Jesse Levin. Of
course, these individuals bear no responsibility for errors
or opinions herein. We thank Abby Linn for excellent
research assistance. Epple and Romano thank the National
Science Foundation and Romano thanks the Institute for
Education Sciences for support.
†
Go to
https://doi.org/10.1257/jel.20150679
to visit the
article page and view author disclosure statement(s).
Journal of Economic Literature, Vol. LV (June 2017)
442
who is eligible to receive them, their source
of funding, and the criteria for private-school
participation. For example, vouchers are fre-
quently “targeted” to low-income students,
are sometimes funded privately rather than
through tax proceeds, and religious schools
may or may not be eligible to participate.2
The research that has analyzed these
programs seeks to answer ve fundamental
questions:
1. What effects do vouchers have on the
students who use them?
2. Do vouchers induce nonrandom migra-
tion of students from public to private
schools, possibly lowering the achieve-
ment of students that remain in the
public sector via peer effects or other
channels?
3. Do voucher programs pressure pub-
lic schools to become more efcient,
increasing the achievement of students
that remain in the public sector?
4. What is the net effect of vouchers on
aggregate educational performance?
2 This program heterogeneity poses a challenge in deter-
mining which programs to classify as voucher programs.
From the perspective of this review, the most difcult deci-
sion regards whether to include charter schools under the
voucher umbrella. Most charter-school advocates would
strenuously resist classifying charter schools as voucher
schools, pointing to state requirements that the former be
chartered as public schools, subjected to the oversight of
the state-designated charter authorizer, and bound by con-
straints on admissions and funding. The counterargument
is that, in some states, charter schools can be privately
owned and operate under constraints on admission and
nancing not markedly different from those imposed by
some of the more restrictive voucher programs. While
there is merit to both perspectives, we view the distinction
between charters and vouchers to be meaningful despite
the fuzzy boundary between the two, in part because all of
the voucher programs we review permit use of the voucher
in private schools. Moreover, there is tremendous hetero-
geneity in charter school characteristics, both across and
within states, making inclusion of charters in this review
unwieldy if not unmanageable. Hence, we have chosen
not to include research on charter schools in this review.
See Epple, Romano, and Zimmer (2017) for a review of
research on charter schools.
5. What political-economy factors deter-
mine the existence and design of
voucher programs?
Research frequently focuses on more spe-
cic questions that help to get at these
fundamentals.
Our review begins by describing the issues
and controversies that frame the research on
these questions (section 2). We set out the
“for” and “against” positions typically (and
at times informally) cited on vouchers. The
brief discussion is sufcient to show that the
answers to questions 1–5 can depend on both
the characteristics of the program analyzed
and the context into which it is introduced.
We then summarize the features of
voucher programs that have been imple-
mented throughout the world (section 3).
We make a distinction between two program
types. First, by small-scale programs, we
mean those that place signicant restrictions
on who can receive vouchers. The most com-
mon restrictions involve income or geogra-
phy; for instance, vouchers may be made
available only to low-income children in a
given municipality within a country. Second,
by large-scale programs we mean those in
which vouchers are distributed countrywide
with minimal restrictions on the type of chil-
dren who can use them.
We then present a brief synopsis of the
theoretical literature (section 4). It reveals
that even in a qualitative sense, the answers
to questions 1–5 depend on voucher design.
The main exception surrounds question 2,
where most models suggest that voucher sys-
tems will display a tendency towards strati-
cation by ability and/or income—although
this too can be mitigated by design and
depends on context.
Finally, we turn to reviewing the empirical
work—the focus of this survey (section5). In
terms of question 1, the empirical research
does not suggest that awarding students a
voucher is a systematically reliable way to
443
Epple, Romano, and Urquiola: School Vouchers
improve their educational outcomes. A per-
haps surprisingly large proportion of the
most rigorous studies suggest that being
awarded a voucher has an effect that is sta-
tistically indistinguishable from zero. At the
same time, there is also evidence that in
some settings, or for some subgroups or spe-
cic outcomes, vouchers can have a substan-
tial positive effect on those who use them.
In addition, however, some recent evidence
points to some discouragingly large negative
test score effects. In terms of question 3,
the literature generally suggests that com-
petition from vouchers leads public schools
to improve. That said, it also makes clear
that it is very difcult to isolate the effect of
competition on public-sector value added
(the object of interest); this reects that as
implied by the answer to question 2, vouchers
typically lead to sorting and can thus affect
public schools through channels other than
productivity enhancements. Taken together,
these ndings point to an ambiguous answer
to question 4 regarding the net effect of
vouchers. Finally, empirical work nds sup-
port for theoretical predictions regarding the
political economy of voucher adoption.
Our “bottom line” assessment is that those
hoping for denitive answers to questions
1–5 will not nd them in the research to
date. In our view, the available answers to
these questions are insufcient to warrant
recommending that vouchers be adopted
on a widespread basis. In that respect, the
effects of vouchers have been disappointing,
relative to early views on their promise.
That said, our view is also that the record
denitely warrants continued explora-
tion. This assessment reects three factors.
First and as stated, there is evidence that
in some cases vouchers can have signi-
cant positive effects on educational perfor-
mance, or at least produce substantial cost
savings. Second, there is some indication
that the prevalence of such results might be
increased with improved voucher design.
For instance, the accumulated research has
begun to provide guidance regarding how
voucher programs may be formulated to
limit adverse effects related to sorting while
preserving achievement-enhancing effects
related to competition. Third, there is evi-
dence that the returns to a well-functioning
education system can be large, with the asso-
ciated implication that a good understanding
of voucher design could be very useful.
A nal note is that given the evidence we
have reviewed, our sense is that work orig-
inating in a single case (e.g., a given coun-
try) or in a single research approach (e.g.,
randomized-control trials) is unlikely to fully
answer questions 1–5. The work on vouchers
suggests that educational markets are com-
plex, and that therefore fairly wide ranging
empirical and theoretical work will be neces-
sary to make progress.
2. The Issues
To provide perspective for our review, we
begin with an overview of the types of issues
theoretical and empirical research on vouch-
ers must address. These issues are complex,
in part because the effect of a voucher pro-
gram depends on both its design and the
institutional and economic setting in which
it is introduced. For instance, the effects
may depend on the size of the program and
also on the alternative: What educational-
provision regime would prevail without the
voucher program? The “effects” that are of
interest are themselves a fundamental issue.
What is the social objective?
To introduce these issues, we list some
classic claims that frame the voucher con-
troversy—and virtually everything about
vouchers remains controversial—with-
out discussing any literature. This listing
will begin to illustrate the challenges that
research on vouchers faces, and our hope is
that the subsequent review will help to clarify
what issues remain the most unsettled. We
Journal of Economic Literature, Vol. LV (June 2017)
444
begin simply by stating some arguments for
and against vouchers, essentially as they are
commonly expressed.
The arguments in favor of vouchers
include:
• Vouchers would lead to market or
quasi-market provision of education,
with competition among providers and
choice by students inducing efcient
provision.3 The alternative of public pro-
vision is characterized by weak incen-
tives, both because public providers are
politically managed and monopolized,
and because the exercise of student
choice is limited. Thus, both static and
dynamic efciency would be promoted
by vouchers, with gains coming both
from private-school advantages and a
public-school response.
• Market provision would lead to educa-
tional variety, better matching prefer-
ences to supply. Diversity would increase
with respect to aspects like curricula and
teaching methodology, an improvement
over the excessive homogeneity associ-
ated with monopolized public provision.
• While there might be concerns regard-
ing externalities from educational attain-
ment (for example, a modern democratic
society requires citizens to be literate
in a common language), restrictions on
private providers could address these.
Similarly, the level of the voucher would
address capital market failures affecting
educational investment.4
• By decoupling residence and school
choice, vouchers would increase access
to quality education, especially for stu-
dents at the low end of the socioeconomic
ladder. Stratied educational provision
3 In our discussion, we will refer to the student and the
student’s household as just the “student.”
4 If necessary, this could be supplemented with policies
supporting educational loans.
wherein quality rises with socioeco-
nomic status would be reduced. In short,
vouchers would provide both efciency
and equity gains.5
The arguments against vouchers include:
• Vouchers would lead to the sorting of
students across schools along charac-
teristics like income and ability (such
sorting is often referred to as stratica-
tion). For example, the private sector
might “cream skim” the highest income
or most motivated children away from
public schools. Teachers would sort as
well—the most advantaged students
would be taught by the best teachers
and the least advantaged by the worst.
• Such sorting would have negative con-
sequences due to peer effects. These
would arise directly, for example, if the
ability to interact with higher-achiev-
ing peers helps students to learn or to
acquire useful networks. Peer effects
could also reect indirect mechanisms,
e.g., if school oversight of wealthier par-
ents disciplines school administrators
and teachers.
• Even if peer effects do not exist, sort-
ing would adversely affect less-advan-
taged students through informational
channels. For example, being at a “bad”
school could stigmatize students in the
labor market, affecting their incentives
to study. Further, sorting might be det-
rimental if the mixing of students along
categories such as race and religion
5 Additionally, a more philosophical argument for
vouchers rests on the substitution of student choice for
public choice. As noted, the traditional economic version
of this argument emphasizes better matching of prefer-
ences to educational supply. The noneconomic version of
the argument places value on freedom of choice per se,
while rejecting the paternalistic alternative. Since the liter-
ature we review considers mainly economic outcomes and
considerations, we abstract from such issues.
445
Epple, Romano, and Urquiola: School Vouchers
promotes mutual understanding valu-
able in a diverse society.
• If it is more expensive to educate dis-
advantaged students, then stratication
would impose costs on the public sec-
tor. In as much as the public sector con-
tinues to serve a segment of students,
political support for funding of public
schools would be reduced by vouchers,
compounding the problem.
• Choice by students who are not well-
informed about educational quality
could lead to poorer choices than deci-
sions by policy makers. Poor choice
could regard the focus of education, as
well as the quality.
These “for” and “against” positions make
clear that research faces a tall order in
understanding the impact of vouchers. For
example, a small-scale voucher system might
produce no sorting response, and its eval-
uation will therefore not address concerns
related to stratication. Similarly, a scheme
that does not generate large-scale entry of
new schools may not reveal gains from cur-
ricular variety.
The usual arguments also implicitly
assume “universal vouchers,” which are avail-
able to all students. In many cases, including
programs in the United States, vouchers are
“targeted” to poorer households, meaning
available only to students whose household
income falls below a threshold. Income tar-
geting of vouchers is intended to provide
access to better educational alternatives to
students who cannot afford to buy expensive
housing in neighborhoods with good pub-
lic schools or pay tuition to attend a private
school. Such targeting would seem to weaken
both the “for” and “against” arguments about
vouchers, presenting another challenge to
research on vouchers. For example, the net
aggregate effect of a targeted voucher pro-
gram could be very different from that of a
universal voucher program.
Another crucial design feature of voucher
programs regards the extent and character of
regulations imposed on schools that accept
vouchers. For example, a school that accepts
voucher-supported students might or might
not be required to accept all applicants, or
use an equal-probability lottery when appli-
cations exceed slots. Participating schools
may or may not be allowed to charge addi-
tional tuition and, if allowed, might be regu-
lated with respect to whether they can price
discriminate.6 Any of these traits will inu-
ence the type of sorting that vouchers can
generate.7
Other issues concern not the design of
the voucher program but the environment
into which it is introduced. For instance,
the extent of “take-up” of vouchers and the
ease of entry of voucher-supported schools
will vary with the population’s density and
preferences. In addition, the initial public
provision regime may or may not already
provide a degree of choice and sorting. For
example, magnet public schools might exist
or ability tracking might be practiced in the
public sector. The baseline equilibrium may
or may not have substantial Tiebout (1956)
sorting by income and preference for educa-
tion. Private schools might already attract a
substantial number of students.
The thematic message is that it is an over-
simplication to view research on vouchers
as a simple test of market versus nonmarket
provision of education. Nonmarket provision
of education is anything but uniform, and a
voucher program need not correspond to a
pure market substitute.
6 Of course, admission and tuition restrictions are
closely intertwined: requiring a school to admit all voucher
applicants but letting it discriminate in tuition can render
the former restriction moot.
7 Sufcient regulation on voucher-supported providers,
such that providers are essentially public, would violate our
denition of a voucher program, though the threshold of
regulations is blurry.
Journal of Economic Literature, Vol. LV (June 2017)
446
3. Voucher Programs
This section describes the characteristics
of existing voucher programs in a total of
eight countries. Given space constraints, our
aim is not to do full justice to the details of
these complex programs, but rather to pro-
vide the reader with useful background and
references that will be relevant in our review
of the empirical literature.
As stated, we make a distinction between
small-scale and large-scale voucher pro-
grams. This will be relevant in our review of
the empirical literature, since each type of
program has analytic advantages and disad-
vantages for conducting research. We clas-
sify voucher programs as small scale when
voucher eligibility is restricted geograph-
ically to only part of an education market
(e.g., only the central-city school district in
a metropolitan area) or vouchers are tar-
geted based on individual characteristics
(e.g., only low-income children are eligible)
or based on school performance (e.g., only
students in underperforming schools are eli-
gible). Conversely, large-scale programs are
those in which the distribution of vouchers
is largely unrestricted within the education
market—all children in a country are eli-
gible. A large-scale program need not, in
principle, be a nationwide program (e.g., a
voucher available to all students in the New
York metropolitan area would be a large pro-
gram). In practice, however, all large-scale
programs are nationwide.
3.1 Small-Scale Programs
We begin by describing the small-scale
programs in the United States, the country
that has produced the greatest number. This
section also describes small-scale programs
in Colombia and India.
3.1.1 United States
The United States has a highly decentral-
ized education system in which states and
districts have signicant control over local
schools. This has produced a large number
of small-scale voucher programs—about
sixty-ve, by an admittedly rough count. We
do not discuss each in detail; rather we sum-
marize the characteristics of three types of
programs that vary according to how vouch-
ers are funded:
• Programs funded by tax revenues. We
summarize the characteristics of nine in
this category, providing additional detail
on the largest and oldest, which operate
in Milwaukee, Wisconsin.
• Programs funded via tax credits. We
summarize the traits of seven in this
group, elaborating briey on the largest,
operating in Florida.
• Programs funded by private founda-
tions. We summarize the typical char-
acteristics of about fty in this class,
highlighting those operated by the
Children’s Scholarship Fund (CSF).
Distinguishing US programs by their source
of funding is for convenience. Nothing sug-
gests that the source of funding per se will
inuence program effects, though some pro-
gram characteristics tend to vary by funding
source, as we discuss. We also provide fur-
ther detail on some programs in our discus-
sion of the empirical research in section 5.8
Tax- Revenue-Funded Programs.—Table 1
provides a summary of the main tax-funded
8 These three categories do not include all small-scale
voucher-type public programs. Specically, since the
1800s, Maine and Vermont have had programs that give
students in sparsely populated areas an alternative to
public provision. Students can use vouchers to attend a
nonreligious private school. In Vermont, for example, the
voucher is for the lesser of average public schooling costs
or tuition. In 2001–02 there were 90 “tuitioning towns” as
they are called, with 7,147 students participating. There
are also programs that target special-needs students that,
in the interest of space, we do not discuss here.
447
Epple, Romano, and Urquiola: School Vouchers
TABLE 1
P F US V P
Milwaukee Racine Florida Cleveland Douglas, CO Ohio Indiana New Orleans Washington, DC
Targeting Below 300%
poverty
Below 300%
poverty 2013–
Failing school Open to all,
priority to low
income
All students Underperforming
public
Below 150%
poverty or failing
school
Below 250% of
poverty
Below 300%
poverty or on
food stamps
Admission to
oversubscribed
pvt schools
Lottery Lottery Lottery Lottery Lottery Pvt school
admission
criteria
Pvt school
admission
criteria
Pvt school
admission
criteria
Pvt school
admission
criteria
Years of program 1990– 2011– 1999–2006 1995– 2013– 2005– 2011– 2008– 2004–
Enrollment 24,027 in 2012 520 in 2012 3,649 in 2012 6,300 in 2009 494 in 2011
(cap 500)
12,685 in 2009 3,919 in 2011 8,000 2013 3,105 in 2012
Funding per
student
$6,442 in
2012/13
$6,442 in
2013/14
K–8: $4,250;
HS: $5,000
K–8: $4,250;
HS: $5,000
$4,600 in
2011/12
K–8: $4,250;
HS: $5,000
$4,700 for K–8,
HS slightly
higher
$7,617 maximum
in 2011/12
K–8: $8,256;
HS: $12,385 in
2013
Transportation
provided
Some, not
routinely
Some, not
routinely
Yes if voucher
school in
student’s district
Yes Yes Up to school No Voucher can be
used to pay for
metro/bus
Number of
participating
private schools
112 in 2012 13 in 2013 32 in 2012 35 in 2012 23 in 2013 305 in 2009 280 in 2013 about 130 52 in 2012
Can tuition
supplement be
required?
K–8: No. HS:
Yes above 220%
poverty
K–8: No. HS:
Yes above 220%
poverty
K–8: No. HS:
Yes above 200%
poverty
K–8: No. HS:
Yes above 200%
poverty
Yes Yes, above 200%
poverty
Yes No Yes
Can vouchers
be used at
religious
schools?
Yes since 1998 Yes Yes Ye s Yes Yes Yes Yes Yes
Same
achievement
exams as in
public schools?
Yes beginning
2006
Yes Yes Yes Yes Yes Yes Yes Yes
Journal of Economic Literature, Vol. LV (June 2017)
448
voucher programs in the United States.9 All
of these programs apply for grades K–12,
and almost all are targeted to students in
low-income households or schools desig-
nated as underperforming. Programs vary
in age, ranging from Milwaukee, which
started in 1990, to New Orleans, which
started in 2008. Programs also vary substan-
tially in the number of students receiving
vouchers. Several programs require that
oversubscribed voucher schools choose
students by lottery. Others permit private
schools to apply the same admission crite-
ria for voucher as for non-voucher students.
Funding per student varies across programs,
but most provide sufcient resources to
attract participation of a substantial number
of private schools. Vouchers may be used in
religious schools in all of these programs.
All now require that voucher recipients take
the same standardized examinations as pub-
lic school students. While not detailed in
the table, all programs require participating
schools to meet curricular and other criteria.
Some programs (e.g., Milwaukee) require
schools to be pre-accredited by an approved
national agency; some (e.g., Ohio) require
that schools obtain a state charter, and oth-
ers require annual reporting to an oversight
body.
Milwaukee is in many respects the most
important voucher program in the United
States and has served as a model for others;
it therefore merits additional discussion. The
Milwaukee Parental Choice Program was
introduced in 1990 in the Milwaukee school
district, targeting K–12 students with house-
hold income not exceeding 175 percent of
the federal poverty level. At its inception,
the program was not available to students
attending religious schools. That changed
9 Though we discuss programs in the present tense, the
Florida program described in table 1, known as the Florida
Opportunity Scholarship Program (FOSP), was declared
unconstitutional and suspended in 2006.
in 1998, with students enrolled in religious
schools retaining the right to opt out of
religious instruction. The voucher pays the
lesser of tuition at a private school and the
standard district allocation, $6,442 in 2010.
Initially, schools could not charge additional
tuition. Beginning in 2011, high schools were
permitted to charge additional tuition to eli-
gible students above 220 percent of the pov-
erty line. Transportation is provided by the
district if the student is within a set atten-
dance area. Participating private schools set
the number of available slots for voucher stu-
dents, and must accept all students, conduct-
ing a lottery if oversubscribed. They must
also be accredited by one of several agencies.
Private schools must also meet at least one of
the following four performance standards:10
(1)at least 70 percent of voucher-supported
students must advance a grade level, (2)fre-
quency of attendance by voucher stu-
dents must be at least 90 percent, (3) at
least 80 percent of program students must
demonstrate signicant academic progress,
or (4) at least 70 percent of voucher-student
families must meet parental involvement cri-
teria set by the school.
The income threshold for eligibility has
been on an upward trend. It was raised from
the initial 175percent of the federal poverty
level to 220percent in 2005, and 300 percent
more recently. This allowed the program’s
coverage to grow. In 2004 it distributed
about 24,000 vouchers, accounting for about
23 percent of total district enrollments. By
2002, 102 private schools were participating,
including 26 schools reported as entering as
a result of the voucher program (Chudgar,
Adamson, and Carnoy 2007).
Tax-Credit-Funded Programs.—We next
turn to tax-credit-funded voucher programs,
10 Wisconsin Administrative Code, chapter PI 35,
p. 117.
449
Epple, Romano, and Urquiola: School Vouchers
summarized in table 2.11 The operation and
funding of one of the earliest, the Florida
Corporate Income Tax Credit Scholarship
Program (FTC), illustrates similarities and
differences between these programs and
the tax-funded programs summarized in
table 1. The FTC was established in 2001
and is nanced by corporate contributions,
for which donors get a 100 percent corporate
income tax credit for contributions that do
not exceed 75 percent of their tax liability.
Total contributions are capped at 88 million
dollars, currently. Vouchers are for free- or
reduced-lunch students and the program is
administered by approved nonprot agen-
cies. In 2012, the FTC awarded about 51,000
vouchers to students attending about 1,300
schools. This makes it the largest voucher
program in the United States—roughly twice
the size of Milwaukee, although relative to a
much larger population.
As shown in table 2, the average voucher per
student among tax-credit-funded programs
was $4,335 in the 2012 school year. Private
schools may impose their own admission poli-
cies, restricted only by antidiscrimination stat-
utes, and can charge tuition in addition to the
voucher, so long as this is their normal policy.
Privately Funded Programs.—Roughly
fty privately funded voucher programs
also exist in the United States.12 The largest
11 The tax-credit-funded programs detailed in table 2
are to be distinguished from state income tax credit and
deduction programs available to households for educa-
tional expenses that currently exist in Arizona, Minnesota,
Illinois, and Louisiana. Given restrictions on amounts,
eligibility, and state income taxation, these programs have
limited effects. For example, the most recently passed pro-
gram in Louisiana in 2008 allows taxpayers to deduct from
the state income tax 50 percent of educational expenses,
up to the minimum of $5,000 per child or the total tax-
able income of the individual. With the maximum marginal
income tax of 6 percent in Louisiana, the maximum sub-
sidy to educational expenditure is below $300.
12 This calculation counts separately programs in dif-
ferent municipalities that are administered by the same
organization.
sponsoring organization is the CSF, and a
brief description of its operation illustrates
key characteristics of this type of program.
The CSF received a founding contribution
from the Walton Family Foundation and
has provided vouchers to low-income stu-
dents in numerous municipalities includ-
ing New York City, Charlotte, Dayton,
Baltimore, and Washington, DC. In 1999,
it received 1.25 million applicants for
40,000 vouchers.
Its Baltimore program is typical. It tar-
gets low-income students in grades K–8. In
2008, it distributed vouchers to 490 students
attending 70 private schools, 64 of which were
religious. The average voucher was $1,759
and the maximum was $2,000. Families are
required to pay at least $500 themselves,
and their average payment was $2,711. All
privately funded programs of which we are
aware are similarly targeted, typically by
income, and some to racial/ethnic groups.
Some also target students that are identied
as having high academic potential but lim-
ited means.
3.1.2 Colombia13
Small-scale voucher programs are much
less common outside the United States, but
Colombia provided a salient, if short-lived,
example. Specically, in 1992 it began oper-
ating the PACES secondary-school voucher
program.14 This initiative was launched to
increase secondary-school enrollment—the
intent was for private participation to help
ease public-sector capacity constraints. The
vouchers, which were renewable contingent
on grade completion, were targeted at
entering students who were: (1) residing in
13 For further reference see King et al. (1997), King
et al. (1998), and Angrist et al. (2002), on which this dis-
cussion is based.
14 PACES stands for Programa de Ampliación de
Cobertura de la Educación Secundaria—program for
increasing secondary school enrollment.
Journal of Economic Literature, Vol. LV (June 2017)
450
TABLE 2
T-C-F V P
Florida Georgia Indiana Iowa Oklahoma Pennsylvania Rhode Island
Targeting Below 230% of
poverty level
All public-school
students
Below 200% of
poverty level or in
underperforming
public school.
Below 300% of
poverty
Below 300% of
poverty level or in
underperforming
public school.
Below
$75,000 and in
underperforming
public school.
Below 250% of
poverty level.
Years of program 2001– 2008– 2009– 2006– 2013– 2001– 2006–
Enrollment 51,023 in 2012–13 13,285 in 2011–12 2,890 in 2012–13 10,600 in 2011–12 N/A 45,100 in 2011–12 382 in 2012–13
Funding per
student
$4,335 average in
2012–13
$3,388 average in
2011–12
$880 average in
2012–13
$1,031 in 2011–12 Up to $5,000 $990 average in
2011–12
$2,690 average in
2012–13
Number of
participating
private schools
1,330 in 2012–13 N/A N/A 154 in 2011–12 N/A 400 in 2012–13 54 in 2012–13
Can tuition
supplement be
required?
Yes Yes Yes Yes Yes Yes Yes
Can vouchers be
used at religious
private schools?
Yes Yes Yes Yes Yes Yes Yes
Same
achievement
exams as in
public schools?
Yes No Yes No No No No
Notes: Tax credits to businesses are generally limited as is the statewide total credit and amount available to provide vouchers to eligible students. Lotteries are
then generally used to provide vouchers to applicants. The poverty level for eligibility is generally measured by the federal poverty level.
451
Epple, Romano, and Urquiola: School Vouchers
low-income neighborhoods,15 (2) attending
public school, and (3) accepted at a partic-
ipating private school.
The initiative was implemented at the
municipal level, with the national govern-
ment covering about 80 percent of its cost
and municipalities contributing the remain-
der. Resource constraints at both govern-
mental levels resulted in excess demand in
most jurisdictions. When this happened,
the vouchers were generally allocated by
lottery.16
The voucher covered registration and tui-
tion payments up to a maximum. Specically,
its value increased with schools’ fees up to
about $150 dollars per year. Angrist et al.
(2002) note this was roughly equivalent
to the cost of a low-to-mid priced private
school, and that it was common for recipients
to supplement this amount. At the program’s
inception, any private school authorized to
operate by the Ministry of Education could
take vouchers, although more expensive
schools generally did not. Starting in 1996,
and following well-publicized reports of per-
ceived low quality at specic private schools,
participation was restricted to not-for-prot
institutions.
By 1995, the year of peak activity, about
90,000 students were using vouchers to
attend roughly 1,800 private voucher
schools. These students accounted for about
1 percent of all secondary-level enrollments
in Colombia (King et al. 1997). Subsequent
declines reected funding constraints that
cut both the number of vouchers and their
maximum value. The program was discontin-
ued in 1997.
15 Colombia had a scheme by which neighborhoods
were classied into six strata depending on income;
only children in the two poorest strata were eligible for
vouchers.
16 Angrist et al. (2002) present evidence consistent with
these lotteries generating random allocation in the cities of
Bogota and Cali, and use this to evaluate PACES, as dis-
cussed herein.
3.1.3 India17
India provides an interesting example
of a small-scale voucher experiment that,
like many in the United States, is privately
funded. Specically, in 2008 the Azim Premji
Foundation began distributing vouchers in
ve districts of the state of Andhra Pradesh,
focusing on 180 villages that each contained
at least one legally operating private school.
Baseline tests were conducted at all private
and public schools in these villages.
All the test takers in public schools were
then invited to apply for vouchers, with the
knowledge that these would be allocated ran-
domly. Students and parents were informed
that the vouchers would cover all school fees
and materials (e.g., textbooks, uniforms,
shoes), but not transportation costs. Students
in public schools are typically of lower socio-
economic status, and many found this offer
attractive.
Private schools were given the option
to join the program, with the understand-
ing that the value of the voucher would be
paid directly to them and was equivalent
to about the 90th percentile of the distri-
bution of all private-school fees in the 180
villages.18 In joining, private schools had to
specify how many slots they would make
available for voucher recipients. The main
condition placed upon these schools related
to non-selection. If space permitted, they
would have to admit all voucher winners; if
they were oversubscribed, they had to enroll
those selected via a lottery run by the funder.
This program featured a unique random-
ization, which took place in two stages. First,
ninety villages were randomly selected to
17 This discussion is based on Muralidharan and
Sundararaman (2015).
18 The full voucher amount is paid directly to the school,
which is in charge of distributing uniforms, textbooks, and
other materials. Muralidharan and Sundararaman (2015)
state that this arrangement reects common practice
among private schools in India.
Journal of Economic Literature, Vol. LV (June 2017)
452
receive vouchers, and ninety remained in
a control group. Second, within the ninety
treatment villages, about 3,000 households
applied for vouchers. About 2,000 were ran-
domly selected to receive them, and about
1,200 of these actually used them. As we will
discuss below, this double randomization has
analytic advantages.
3.2. Large-Scale Programs
We now turn to describing large-scale pro-
grams. The fact that they distribute vouch-
ers without targeting implies that these have
the potential to have a greater effect on
educational markets, although this impact
ultimately depends on their design and the
context into which they are introduced.
Table3 summarizes ve cases we consider,
highlighting variation in the percentage of
enrollments at independent schools and
whether these schools can operate for prot,
implement selective admissions policies,
have a religious afliation, and charge tui-
tion above the voucher. The remainder of
this section discusses these on a case-by-case
basis.
3.2.1 Chile19
In 1981, Chile introduced a universal
voucher scheme. Prior to this reform, three
types of schools were in operation: (1) pub-
lic schools were managed by the national
Ministry of Education and accounted
for about 80 percent of enrollments;
(2) unsubsidized private schools catered to
upper-income households, and accounted
for about 6 percent of enrollments; and
(3)subsidized private schools did not charge
tuition, received limited lump-sum subsi-
dies, were often Catholic, and accounted for
roughly 14 percent of enrollments.
The 1981 reform had two main compo-
nents. First, it transferred public-school
management to municipalities, simultane-
ously awarding them a per-student subsidy
sufcient to cover their costs. Second, subsi-
dized (or “voucher”) private schools began to
receive exactly the same per-student subsidy
19 For further background see Gauri (1998), McEwan
and Carnoy (2000), Mizala and Romaguera (2000), Hsieh
and Urquiola (2006), and Mizala and Urquiola (2013).
TABLE 3
I V P
Restrictions on private/independent schools
Country
Years in
operation
Enrollments in private
or independent
voucher-funded
schools
For-prot
operation
allowed
Selective
admissions
allowed
Religious
afliation
allowed
Signicant
tuition charges
allowed
Chile 1981– 47% Yes Yes Yes Yes
Denmark 1855– 12% No Ye s Yes Yes
Holland 1917– 70% No Yes Yes No
New Zealand 1989– 15% Ye s Yes Yes No
Sweden early 1990s– 10% Yes No Yes No
453
Epple, Romano, and Urquiola: School Vouchers
as municipal schools. These changes resulted
in substantial private-school entry. By 2009,
about 57 percent of all students attended
private schools, with voucher schools alone
accounting for about 50 percent.20
Chile’s scheme imposes few restrictions on
private schools. These can receive voucher
subsidies regardless of their religious status
and operate for prot. They are allowed to
implement admissions policies subject to
few restrictions and, as of 1994, can charge
tuition add-ons.21 The latter are capped at
about four times the voucher payment, but
this constraint rarely binds.22 Public schools
operate under more restrictions. They are
not allowed to turn away students unless
oversubscribed, and cannot charge tuition
at the primary level. All schools must imple-
ment elements of a national curriculum and
participate in annual standardized exams,
the results of which have been public since
the 1990s.23
Recent years have seen further reforms.
Since 1997, schools charging tuition add-ons
are required to provide exemptions on these
for a percentage of low-income students.
In 2008, the at voucher became differen-
tiated: it was increased for low-income stu-
dents. However, not all schools receive these
additional subsidies, as they have to comply
with a number of conditions to receive them.
Even further signicant reforms to the
voucher system are under active discus-
sion, in part in reaction to persistent student
20 The “elite” unsubsidized private schools continued to
account for about 6 percent of enrollments.
21 Over the years, education-related legislation often
mentions that private schools should not select students.
The anecdotal evidence indicates that this rarely binds—
for instance no admissions lotteries are required. We
return to this issue below.
22 Most of the “elite” unsubsidized private schools
could take vouchers but choose not to; see Urquiola and
Verhoogen (2009).
23 These tests have been used for purposes of account-
ability and targeting. For instance, Chay, McEwan, and
Urquiola (2005) consider a program that targeted the 900
worst-performing schools in the country.
protests. The current Bachelet administra-
tion has submitted to congress a proposal
with three main ingredients: (1) the elimi-
nation of tuition top-ups, (2) the end of the
ability of private voucher schools to operate
for prot, and (3) the introduction of a sig-
nicant reduction of the ability of private
schools to select students. The proposals
are still under discussion, and the details of
implementation remain to be seen.
3.2.2 Denmark24
Denmark has a long tradition of subsi-
dizing independent schools. Since the Free
School Act of 1855, it has allowed parents
and organizations to set up independent
schools to which any child can apply, and
which are allowed to have religious aflia-
tions. Historically, these schools were funded
through a scheme by which the State reim-
bursed a large portion of their expenses.
In 1992, this system was replaced with one
that provides independent schools with a
grant based on the number of pupils enrolled
by a certain date. Public schools continue to
be nanced by a combination of national and
local government allocations; they do not
receive per-student payments.
The voucher-type payment for indepen-
dent schools is indexed to expenditures in
public schools and varies with two factors:
school size (with higher payments for smaller
schools, to account for economies of scale)
and the age distribution of students and
teachers. These payments cover only about
80 percent of average educational costs, and
independent schools are therefore allowed
to charge tuition (low-income households
can apply for waivers) or seek external grants
to cover the remainder. Despite this, total
per-pupil expenditures are slightly lower in
24 For further discussion see Justesen (2002), on which
this discussion is based.
Journal of Economic Literature, Vol. LV (June 2017)
454
the independent-school sector.25 By 2005
independent schools accounted for about
12 percent of enrollments. They operate
subject to nationwide collective agree-
ments with teachers, and to basic curricu-
lar requirements that leave them relatively
broad pedagogical autonomy.
3.2.3 Holland26
The Dutch 1848 constitution allows for
churches, foundations, and parental associ-
ations to set up independent school boards
that operate schools to which any child can
apply. The 1917 constitution further includes
commitments of equal nancial support for
public and independent schools. While both
types of schools receive funds for infrastruc-
ture, a substantial part of schools’ support is
in the form of a per-pupil grant, with greater
payments when they enroll children of low
socioeconomic status.
While this system was initially set up to
allow for transfers to Catholic and Protestant
schools, at present it also covers schools with
other religious afliations. Nevertheless, a
majority of independent schools still identify
as Protestant or Catholic, with enrollment
shares of 27 and 29 percent, respectively.
The public sector’s share is 35 percent,
with the remaining 9 percent of children in
schools of other types. Independent schools
must be run on a not-for-prot basis and
“top-up” tuition charges are not allowed.
In addition, these schools must implement
at least parts of a core national curriculum,
participate in national standardized exams,
and comply with regulations regarding
aspects like class size, teacher qualications,
and minimum enrollments. Private schools
implement selection policies and may deny
25 Justesen (2002) indicates that in the aggregate, 77
percent of independent schools’ resources come from
voucher-type payments, 18 percent from user fees, and the
remaining 5 percent from other external sources.
26 For further reference, see Justesen (2002), Patrinos
(2002), and Levin (2004), on which this discussion is based.
admission for various reasons, including reli-
gious afliation.
3.2.4 New Zealand
27
In 1989, New Zealand implemented a
decentralization initiative transferring con-
trol of each public school from a national
department of education to a “Board of
Trustees”—largely consisting of parents—
elected locally. In 1991, this system was
extended by granting per-pupil funding to
all schools, including independent and “inte-
grated” institutions. The latter are schools
which, while being institutionally indepen-
dent, had been afliated with the public sys-
tem since the 1970s; most, though not all,
have a religious afliation. At present, the
enrollment shares of public, integrated, and
independent schools are 85, 11, and 4 per-
cent respectively.
These arrangements put in place a key
ingredient of a voucher system—schools that
attract more students receive greater fund-
ing. That said, they depart from the textbook
voucher in some ways. First, not all schools
receive the same per-student funding. Public
schools receive subsidies for teacher sala-
ries, operational costs, and capital expenses;
integrated schools are only compensated for
teacher salaries and operational costs; and
independent schools receive only a portion
of the per-student payments awarded to
integrated schools (the percentage has uc-
tuated around 30 percent over the years).
Second, public and integrated schools do not
have control over teacher pay; pay scales are
centrally determined for all but the indepen-
dent schools.
In addition, while public schools may
supplement their central subsidies via fund-
raising activities and donations, they are not
allowed to charge mandatory fees. Integrated
schools are allowed to collect donations and
27 This description is based on Ladd and Fiske (2001);
Adams (2009); and Lubienski, Lee, and Gordon (2013).
455
Epple, Romano, and Urquiola: School Vouchers
charge compulsory “attendance dues” to
meet capital costs. Independent schools can
charge fees.
The 1991 legislation allowed all schools
wide latitude in setting up admissions pro-
cedures. For example, admissions policies
could specify a catchment area, sibling
preferences, and the use of parental inter-
views. Lubienski, Lee, and Gordon (2013)
point out that some restrictions on selec-
tion were implemented in the 1990s. These
mainly related to transparency in stating the
selection criteria and the specication of
catchment areas (there had been objections
around the fact that children living very close
to a given school might not gain access to it).
Nevertheless, schools retain wide latitude in
selecting students.
3.2.5 Sweden28
Prior to the early 1990s, almost all Swedish
children (about 99 percent) attended munic-
ipal schools. While controlled by local juris-
dictions, municipal schools were funded
by earmarked transfers originating in the
national government, which also directly
hired teachers. Beginning in 1991, these
arrangements underwent reforms that had
four main components. First, the govern-
ment turned the earmarked funds into
largely lump-sum subsidies, with munici-
palities gaining latitude in nancial man-
agement. Second, municipalities became
teachers’ ofcial employers, obtaining
the ability to negotiate pay and terminate
employment. Through 1996, however, the
national government largely xed teacher
pay as a function of credentials and expe-
rience (Hensvik 2012). Beginning in 1996,
salaries were determined by negotiation at
the local level. Although these negotiations
allowed for greater pay differentials, their
28 For further reference see Sandstrom and Bergstrom
(2005) and Bohlmark and Lindahl (2007), on which this
discussion is based.
outcomes continued to be constrained by
agreements at the national level.29 Third,
“open enrollment” plans were instituted at
the municipal level, such that, in principle,
students could attend any school in their
jurisdiction; in practice, distance to school
continued to be a criterion for admission.
Fourth, independent schools were given the
right to receive municipal funding as well—
the government mandated that municipali-
ties fund them with a per-student payment
equivalent to the resources they would have
spent themselves. In practice, these pay-
ments equal about 80 percent of per-student
costs at municipal schools.30
Independent schools must be approved by
the National Agency for Education. While
municipalities can raise objections regarding
specic institutions that apply for approval,
they do not have veto power. Independent
schools may be operated on a for-prot or
nonprot basis, they can be religious or sec-
ular, and they can focus on specic ethnic
groups or languages. In all cases, however,
independent schools must be open to all
students—regardless of their municipality
of origin, ethnicity, or religion—and they
cannot charge tuition beyond the voucher.
Further, grades cannot be used as admissions
criteria at the compulsory level. Instead,
proximity to the school, wait list ( rst-come,
rst-served), and sibling presence at a school
determine priority. Ability-based admissions
are allowed at the secondary level. Top-up
funding is not permitted. Ownership of a
school is unrestricted, and hence can be reli-
gious, for-prot, or nonprot.
29 There were two ve-year agreements between the
central government and the teachers’ union. The rst
raised teacher pay by 10 percent over ve years and the
second, beginning in 2000, by 20 percent. Hence, local
negotiations were constrained by minimum-pay require-
ments set at the national level.
30 Further, if a student crosses municipal lines, the
locality losing him has to make a similar transfer to the
municipality that accommodates the student.
Journal of Economic Literature, Vol. LV (June 2017)
456
Bohlmark and Lindahl (2012) point
out that there was relatively little voucher
school entry through 1998; at that point, the
independent-school share began to grow. By
2004, the independent-school enrollment
rate reached 10 percent for high schools and
6 percent for primary and lower secondary.
By 2009, independent schools accounted for
roughly 10 percent of all students. The rea-
sons for the relative lack of voucher school
entry for the rst several years are unclear.
These could include anything from a lack
of information or risk aversion on the part
of parents, to the fact that the 1996 relax-
ation of centralized wage setting might have
allowed independent schools to compete
more effectively.
4. Theory
In this review, our primary emphasis is
on empirical research (a comprehensive
review of theoretical and computational
research is provided by Epple and Romano
2012). Nonetheless, this section provides a
brief summary of the theoretical literature
with an emphasis on empirical and policy
implications.
The case for a market-based educational
voucher was laid out by Milton Friedman
(1962), who provided a vision for voucher
design and an enumeration of the benets
that he foresaw from voucher adoption. He
supported public funding of education on
the grounds that such funding was warranted
by the social externalities owing from an
educated population and due to borrowing
constraints, but argued that public fund-
ing need not imply public provision. He
envisioned a system in which parents could
choose a school for their child with public
funding going to the chosen school. The role
of government would be to provide funding
while “… insuring that schools meet certain
minimum standards, such as inclusion of
minimum common content in programs…”
(p. 89). Implicit in this government role
would be assurance that voucher funds be
spent on education. Friedman argued that
competition for students would induce
schools to operate efciently and reward
quality teaching, with effective schools
establishing good reputations. The poor
would have educational choices not bound
by the residence restrictions embodied in
neighborhood public school systems. In
Friedman’s view, the education environment
was not sufciently different from other
market settings to interfere signicantly with
effective functioning of such a marketplace
for education.
More recent research has modeled edu-
cational vouchers taking account of distinc-
tive features of the education environment.
Table 4 provides a summary of the charac-
teristics of models that we discuss in this
section. The delineation of model charac-
teristics in table 4 is imperfect and does not,
of course, fully describe differences across
papers. For example, the table indicates
whether the vouchers that are studied are
targeted, but does not indicate the type of
targeting, which varies in important ways
across studies. Likewise, there are import-
ant differences in what makes public schools
heterogeneous in the models that have such
differentiation. In the discussion below,
these modelling differences are highlighted.
In reviewing this recent literature, we make
reference to how it helps to address the ve
fundamental questions on vouchers that we
set out in section 1.
A central theme that emerges is that
the answers to these questions depend on
voucher design. Regarding question 2, vir-
tually all theoretical analyses predict that
a laissez-faire design will induce “cream
skimming,” with the associated implication
for question 1 that some students will gain
more than others; and some will be made
worse off unless the effects on public-school
performance (question 3) are substantial.
457
Epple, Romano, and Urquiola: School Vouchers
TABLE 4
C T M V
Public schools Student differentiation Vouchers
Private
price
discriminate
Housing
market
Public choice
Voucher models
Peers affect
quality
Homoge-
neous
Heteroge-
neous
Rent
seeking Income Aptitude
Imperfectly
observed Universal Targeted
No topping
up
Limit
admissions
Public
expenditure Voucher
Manski (1991) X X X X X X
Epple and Romano
(1998)
X X X X X X
Nechyba (1999) X X X X X X X
Nechyba (2000) X X X X X X X
Nechyba (2003) X X X X X X X X
McMillan (2005) indirectly X X X X X
Ferrerya (2007) X X X X X X
Epple and Romano
(2008)
X X X X X X X X
Macleod and Urquiola
(2009)
X X X X X
Ferrerya and Liang
(2012)
X X X X X X X
Chakrabarti (2013c) X X X X
Nielson (2013) X X X X X
Chakrabarti (2013d) X X X X X X
Public-choice models
Ireland (1990) X X X
Epple and Romano
(1996)
X X X X
Hoyt and Lee (1998) X X X X X
Chen and West (2000) X X X X X X
Fernandez and
Rogerson (2003)
X X X X X X
Bearse et al. (2009) X X X X
Epple and Romano
(2014)
X X X X X
Epple, Romano, and
Sarpca (2014)
X X X X X
Journal of Economic Literature, Vol. LV (June 2017)
458
Theoretical models are often paired with
a computational counterpart to quantify
magnitudes, distributional effects, and,
with respect to question 4, net impacts. As
problematic implications of the laissez-faire
design have become better understood,
research has increasingly emphasized ways
in which benets from voucher-induced
competition can be obtained without adverse
distributional effects. While the natural
focus with regard to question 1 is on educa-
tional outcomes, theory has also developed
interesting implications regarding impacts
on residential choice and housing values, and
the connection to voucher design. Regarding
question 3, theoretical research has iden-
tied potential sources of efciency gains
from educational competition, as well as
ways that public-school performance might
be adversely affected. Failures of voucher
proposals at the ballot box have motivated
research addressing question 5.
4.1 The Effects of Vouchers
The theoretical and computational litera-
ture typically begins from a characterization
of the educational environment, while taking
the existence and characteristics of vouch-
ers as exogenous. The question then is how
the introduction of a given voucher program
into a school “market” affects school effec-
tiveness, the distribution of outcomes and
welfare across students, the distribution of
students across schools, tax revenues, public-
school expenditures, residences, and prop-
erty values. In our discussion below, we draw
out the predictions of theoretical models,
while also noting those that have not yet
received empirical testing.
Manski (1992) pioneered this type of
approach developing a theoretical and com-
putational model that captures features of the
educational environment including: public
and private sectors between which students
can choose; students differing by house-
hold income and motivation, with demand
for educational quality rising with income
and motivation; a positive peer externality
from highly motivated students; educational
quality determined by expenditure per stu-
dent and peer quality; analysis of alternative
public-sector objectives including rent seek-
ing; and zero-prot private schools that set
tuition to maximize enrollments with tuition
and a voucher spent on educational inputs.31
Manski uses this setup to assess if vouchers
would induce changes that equalized edu-
cational opportunity. The simulations and
outcomes he considers are numerous, but
overall the conclusion is that vouchers are
not a “panacea.” A key prediction is that, as
the voucher level rises, the fraction of highly
motivated students in the public schools
tends to fall, especially in poor communities.
He states that “even in the most favorable
case, a systemic choice system would not
come close to equalizing educational oppor-
tunity across income groups.” Thus, Manski’s
analysis predicts that cream skimming of the
sort raised in question 2 will adversely affect
less-motivated students.
Epple and Romano (1998) study private-
and public-school competition when stu-
dents vary in ability and household income,
and school quality increases with peer abil-
ity. Private schools maximize prots and
can price discriminate, i.e., charge tuition
that varies with student ability and income.
Schools have xed costs, as well as variable
costs that are an increasing convex function of
enrollment, i.e., cost per student is U-shaped
in enrollment. The model gives rise to the
following predictions. First, because school
quality increases with peer ability, private
schools charge lower tuitions (i.e., provide
31 While Manski describes students as varying in motiva-
tion, we label such variation as “aptitude” in table 4. Many
authors describe students as varying in ability, and we have
just chosen one term in the table to describe student vari-
ation along these lines. Note, too, that Manski considers
rent-seeking public schools as we indicate in table 4, but
also other objectives.
459
Epple, Romano, and Urquiola: School Vouchers
more nancial aid) for high-ability students.
Second, higher-income households with
low-ability students pay a tuition premium to
enable their children to attend high-quality
schools. Thus, the school system will have
a general tendency towards stratication
in two respects. There will be stratication
across schools within the private-school sec-
tor and there will be stratication between
the public and private sectors. Moreover,
private schools will be differentiated in
quality and will exhibit “diagonal strati-
cation,” with each private school having a
student body that is a “diagonal slice” in the
income–ability plane. The lowest-income
and lowest-ability students will attend public
schools. Thus, the model predicts that there
will be a higher correlation between income
and ability in public than in private schools.32
The model predicts that introduction of
a universal ( at-rate) voucher will induce
additional private schools to enter, with each
entering school being of lower quality than
the preceding entrant, and each exhibiting
the diagonal-stratication pattern. As the
amount of the voucher increases, average
peer quality in the public schools is pre-
dicted to decline as private entrants “cream
skim” higher-income and higher-ability
students from the public schools. If a com-
paratively low voucher is introduced, those
switching to private schools will attend a
school with higher peer quality than the
public-sector school they depart. As the
voucher level is increased, however, stu-
dents who are induced to switch to private
school exit a public-school sector whose
peer quality has been diminished by cream
skimming to a private school with compar-
atively low peer quality. Thus, regarding
questions 1 and 3, the model predicts that
there will be high-achieving voucher schools
32 These and other predictions of the model are tested
in Epple, Figlio, and Romano (2004) and are found to be
supported by the data.
serving the relatively more able and afu-
ent, lower-achieving voucher schools serving
the relatively less able and afuent, and a
public-school sector with lower achievement
still. Epple and Romano (2008) extend this
setup to show that these properties persist
when school quality depends on expenditure
per student in addition to peer quality.33
Could all students nonetheless have
improved educational outcomes with the
voucher? If private schools have an educa-
tional approach that is sufciently superior to
that of the public schools they supplant, and
if the remaining public schools are induced
by competition to adopt a superior delivery
approach, all students might have higher
achievement than in the no-voucher equilib-
rium. Computational analysis calibrated to
the US context suggests, however, that some
students will benet from the voucher—the
comparatively more able and afuent—while
others—the comparatively less able and
afuent—will be hurt. Regarding question 4,
the effect on normed aggregate achievement
(equal to future earnings) may be positive or
negative depending on the extent to which
private-school education delivery is more
effective than preexisting public schools, and
the extent to which public schools upgrade
delivery in response to competition. In
summary, the model yields unambiguous pre-
dictions about stratication, private-school
pricing, and relative achievement across the
predicted school hierarchy, while predicted
aggregate effects depend on the impact of
vouchers on educational effectiveness. It
33 One additional nding is that low-quality “bottom
feeder” schools may enter when vouchers are available,
providing nancial aid “kickbacks” to induce low-income
households to choose low-quality schools. It is shown that
this can be prevented by a mandate that the voucher be
spent on education. Theoretical models have generally
assumed that kickbacks are not allowed. The incentive to
kickback monies to poor students raises the related ques-
tion as to whether vouchers would lead schools to pro-
vide noneducational goods to students as a way around a
requirement to spend all of a voucher on education.
Journal of Economic Literature, Vol. LV (June 2017)
460
bears emphasis that these predictions are
for a universal ( at-rate) voucher design, the
Chilean voucher (at least in roughly its rst
two and a half decades) being perhaps the
closest observed counterpart.
Epple and Romano (2008) also investigate
the implications of voucher design for cream
skimming, showing that an ability-targeted
and tuition-constrained voucher can preserve
efciency benets from competition while
eliminating cream skimming and providing
relatively uniform benets across the distri-
bution of student income and ability types.
The tuition constraint disallows “top-ups”
and “kickbacks.” Such an ability-targeted
design has not been implemented in prac-
tice.34 Voucher designs requiring that
oversubscribed schools select by lottery
and that all school funds be spent on edu-
cation, coupled with a prohibition against
topping up, may be the nearest operational
counterpart. Chakrabarti (2013b) provides a
model of such a voucher and tests the sort-
ing predictions, as discussed in section 5.3.35
34 In related work, Eden (1994) examines efcient
voucher policy in a model with peer effects within schools
and an achievement externality to society. Education is
a pure investment good and capital markets are (implic-
itly) perfect. He shows that an achievement subsidy
aligns school and social incentives, and combined with a
type-dependent voucher equal to the efcient expenditure
plus the student’s peer externality cost induces an efcient
( zero-prot) equilibrium in which students pay nothing out
of pocket.
35 Chakrabarti (2013b) assumes students differ continu-
ously in income and ability, with demand for school quality
increasing in both. School quality is equated to expendi-
ture per student, with a maximum quality. She considers a
voucher that, for simplicity, covers the highest-quality cost
of education, effectively implying no topping up. Neither
can private schools kick back any of the voucher. Private
schools have no incentive to base admissions on ability
due to an absence of peer effects, as with a voucher that
requires equal probability of admission. Private-school
slots are, however, limited. While the voucher covers all
tuition costs, students face utility costs of applying for a
voucher that they may not get, and may face a monetary
cost of attendance if, for example, transportation costs
are not covered. Chakrabarti shows that there is sorting
by ability at the application stage, but there may not be
sorting by income. The former is because higher-ability
Relative to the design of Epple and Romano
(2008), the absence of enhanced voucher
funding for low-ability students reduces
incentives for schools to seek out and retain
less able students.
Nechyba (1999, 2000, 2003) analyzes the
effects of voucher programs in multi-district
local economies. He develops a rich theoret-
ical and computational model to investigate
the effect of several voucher programs under
alternative public-school nancing schemes.
He demonstrates the importance of house-
hold mobility and general equilibrium effects
in predicting outcomes from large-scale
voucher programs. In his 1999 framework,
there are multiple local school districts, a
xed stock of heterogeneous housing units,
neighborhoods within districts differentiated
by housing quality, district-wide homoge-
neous public schools, perceived education
quality that varies with expenditure per stu-
dent and average peer quality, and peer qual-
ity that is correlated with household income.
Tuition varies across private schools, but, in
contrast to Epple and Romano (1998, 2008),
price discrimination is not permitted, imply-
ing each private school is specialized to serve
one student type. This and the willingness
of higher-income households to pay a pre-
mium for quality results in stratication by
peer ability and income in the private sector.
Households simultaneously choose where to
live (district and neighborhood), whether to
send their child to public or private school,
and vote for a district-wide property tax
used to nance public schools. Nechyba
conducts policy analysis in a computational
types value quality by more and there are utility costs of
applying. The latter is because tuition is fully covered, util-
ity costs of applying are independent of income, and the
potential monetary cost may not be enough to deter appli-
cation. In contrast, in the enrollment stage, there will likely
not be ability sorting but there will be sorting by income.
The former is because ability sorting has already occurred,
and the latter is because monetary-attendance costs that
arise for some will deter take up by some lower-income
students.
461
Epple, Romano, and Urquiola: School Vouchers
counterpart calibrated to data for New Jersey.
His model predicts that private schools will
emerge in poor communities as high-income
households take vouchers and relocate to
occupy higher-quality housing in such com-
munities. Hence, stratication of income
and property values across communities is
reduced. Poorer households do not experi-
ence improved peer quality in their (pub-
lic) schools, however, because incumbent
somewhat higher-income households opt for
private schools or relocate to communities
with better public schools. Expenditure per
student rises in public schools as long as the
voucher is not high enough to induce more
than half the population to attend private
schools. Hence, public-school quality could
increase if this increased spending offsets
the decline in public-school peer quality. By
allowing mobility and expenditure effects in
public schools, Nechyba’s analysis predicts
more favorable effects of universal vouch-
ers on poor students, relative to Epple and
Romano (1998, 2008). On the other hand,
by not allowing price discrimination, bene-
ts from vouchers to high-ability students,
whether rich or poor, are curtailed.
Nechyba (2000, 2003) extends this frame-
work by studying vouchers targeted to poor
individuals and poor districts, as compared
to universal vouchers. This is of particular
relevance for the US context.36 He concludes
that a small non-means-tested voucher tar-
geted to residents of low-income districts is
largely equivalent to a universal voucher that
is not targeted, due to household mobility.
36 As reported in tables 1 and 2, targeting vouchers
to the poor usually characterizes US voucher programs.
Targeting to poor districts would be similar to the prac-
tice of targeting to poorly performing schools if households
need only reside where their designated public school
is so labeled to get a voucher, though prior attendance
requirements limit this. In addition to differences in the
underlying models, the ability targeting analyzed in Epple
and Romano (2008) has a normative focus, while targeting
forms studied in Nechyba (2000, 2003) are better moti-
vated empirically.
Most households taking such vouchers would
reside in or move to low-income districts,
whether or not targeted to these districts.
More generally, for realistic values, vouchers
targeted to the poor district will have small
effects. Similarly, income-targeted vouchers
will have modest effects unless school qual-
ity depends largely on child ability. In that
case, low-income parents of high-ability chil-
dren will choose private schools in districts
with poor-quality public schools. Milwaukee
would appear to be fertile ground for empir-
ical investigation of Nechyba’s predictions of
the effects of vouchers on household location,
but such testing has not been undertaken.
Ferreyra (2007) builds on Nechyba’s
model introducing both preferences for reli-
gious schools and idiosyncratic (randomly
drawn) preferences for school types (pub-
lic, private, secular, religious) and location.
She estimates the parameters of the model
using data from seven metropolitan areas.
She then uses these estimates to simulate
the effect of several voucher programs. In
particular, Ferreyra examines the differential
effects of vouchers depending on whether
these can be used at religious schools.37 She
nds that both types of voucher programs
increase private-school enrollment and give
rise to mobility effects of the type identied
by Nechyba. She also nds that a prohibition
on the use of vouchers at religious schools
results in less private-school enrollment and
can shrink religious enrollments as some stu-
dents take a voucher and switch from reli-
gious to secular private schools. Milwaukee’s
1998 shift to allowing the use of vouchers
at religious schools provides a potentially
promising environment for testing these pre-
dictions, although, to date such testing has
not been undertaken.
37 We describe the vouchers in Ferreyra (2007) as
universal and nontargeted in table 4, but the variation in
whether vouchers can be used at religious schools or not is
central to her paper.
Journal of Economic Literature, Vol. LV (June 2017)
462
Neilson (2013) develops and tests a model
with geographically differentiated schools
that compete for vouchers that are higher for
poorer households. The empirical applica-
tion is to Chile, as we discuss in section 5.3.
Private schools are differentiated by their
endogenously chosen quality, as well as by
their location. There are no peer effects.
Households differ demographically (e.g.,
in income) and in residence, with idiosyn-
cratic preferences over schools, as well as
(estimated) systematic differences in prefer-
ences. Locational differences among private
schools and idiosyncratic preferences imply
market power.38 Prot-maximizing private
schools choose quality below the competitive
( zero-prot) level, with the quality reduction
increasing in their market power. The quality
markdown is greater in poorer areas, where
households are estimated to be more price
sensitive. As such, vouchers that are higher
for poorer households have a greater posi-
tive effect on quality. This paper speaks to
question 1, with gains to voucher students
coming largely from reduction in market
power among private providers; but it is
also relevant to question 3 on public-sector
responses.
In contrast to the research discussed so
far, McMillan (2004) is squarely focused
on question 3. This paper endogenizes how
public schools adjust their effectiveness in
response to competition from more effec-
tive private schools. In McMillan’s frame-
work, households are of two income types,
with high-income households willing to
pay more than low-income households for
school quality. Schools exert effort, which
raises their effectiveness but comes at a cost
to them. Competition constrains private
schools to provide efcient effort. Private
38 Nielson’s schools are then differentiated “vertically”
by quality and “horizontally” by location and idiosyncratic
appeal. Epple et al. (2013) also provide a model of vertical
and horizontal school differentiation, applied to colleges.
schools serving low-income students charge
lower tuition and provide lower effort than
private schools serving high-income stu-
dents. The rent-seeking public-school sec-
tor will provide one of two quality levels, a
high level sufcient to attract both high-
and low-income students, or a low level
that attracts only low-income students.
If the former, high-income students pre-
fer public schools over paying tuition to
attend private school; hence the public sec-
tor attracts all students. McMillan consid-
ers the effect of a universal voucher in the
high public-school effort case. The voucher
lowers the cost of private-school education,
and may induce high-income households to
switch to private schools. If this happens,
public schools choose to lower effort to the
level required to retain only low income stu-
dents. A voucher could, alternatively, induce
public schools to increase effort to retain
high-income students. Hence, McMillan
provides a mechanism such that, instead
of improving public-school effectiveness,
voucher-induced private-school competi-
tion, and associated income stratication,
may have an adverse effect on public-school
effectiveness. McMillan’s framework thus
captures an endogenous peer effect associ-
ated with variation in how parents of differ-
ent income levels are able to induce school
effectiveness.
Building on McMillan (2005), Ferreyra
and Liang (2012) model imperfect paren-
tal and policy maker monitoring of schools’
effort choices. Households vary in ability
and income, and higher-ability households
are more efcient at monitoring their chil-
dren’s learning. Competitive private schools
are sufciently small that no free-rider prob-
lem arises in parental monitoring, while free
riding prevails in the public sector. They
demonstrate that combining vouchers with
increased public monitoring of the public
sector has the potential to increase every-
one’s achievement and aggregate welfare.
463
Epple, Romano, and Urquiola: School Vouchers
Chakrabarti (2013c) also develops a model
where vouchers could increase or decrease
effort of rent-seeking public schools. In
her model, students differ continuously in
income and ability, with demand for qual-
ity increasing in both and school quality
depending on school effort and mean ability.
Private schools cannot price discriminate as
in Nechyba’s (1999) model. Universal vouch-
ers induce higher-ability students to exit the
public sector, implying students at the margin
of attending public school are of lower ability
in the voucher regime. Increasing effort and
public-school quality has a smaller marginal
effect on increasing their attendance. If this
is the dominant effect, then public schools
worsen as in McMillan (2004).
Motivated by voucher programs like the
former Florida program (FOSP), targeted
to failing schools, Chakrabarti then contrasts
such a voucher with a policy that awards
vouchers only if the public school fails to
meet a quality standard. She shows that
with appropriate setting of the quality stan-
dard, such a program will induce increased
public-school effort and quality improve-
ment. This is because public schools have
a stronger incentive to improve to meet the
standard and avert the voucher and loss of
students, while students at the margin of
attending private schools would always exit
with a universal voucher. She goes on to test
the model as discussed in section 5.3.
In exploring why sorting might adversely
affect students left in the public sector,
the above models focus on peer effects.39
MacLeod and Urquiola (2009, 2012) depart
from this by studying informational mecha-
nisms instead. Specically, they model the
combination of educational and labor mar-
kets. In a rst period, each individual attends
school and accumulates skill as a function of
39 See Sacerdote (2011) and Epple and Romano (2011)
for recent surveys of the literature on educational peer
effects.
her innate ability, her effort, and her school’s
value added. In the second period, the indi-
vidual is employed in a competitive labor
market. MacLeod and Urquiola assume
that innate ability and effort are not directly
observed; employers infer ability from all the
available information. A key assumption is
that individual innate ability is more accu-
rately assessed by schools than by employers.
For example, schools might be better able
to administer admissions exams or conduct
parental interviews. As a result, employers
rationally use an individual’s school of ori-
gin as a signal of her skill. In turn, students
seek to attend schools with good reputations,
where a school’s reputation is the expected
skill of its graduates.
Two key sets of empirical implications
emerge. First, “laissez-faire” school sys-
tems have a tendency towards stratication
by ability. Students in nonselective schools
(e.g., the public sector) will be hurt by this
stratication; their low ability is revealed to
employers by their failure to gain admission
to a selective school. Second, the effects of
school competition induced by vouchers will
depend on design. For example, schemes
that restrict schools’ ability to select students
will maximize effort on the part of students
and their willingness to choose schools with
the highest value added. Schools, in turn,
will be forced to build their reputations on
their advantage in value added, as opposed
to just their ability to select high-ability stu-
dents. In contrast, systems that facilitate
selection will tend to lower students’ study
effort and raise the probability that they
choose schools based on peer quality rather
than value added.
To summarize, MacLeod and Urquiola
(2009) show that even in the absence of peer
effects, the reputational mechanisms empha-
sized by Friedman (1962) do not ensure that
vouchers will increase the production of
skill. The intuition behind this result is two-
fold. First, the fact that school membership
Journal of Economic Literature, Vol. LV (June 2017)
464
allows students to convey their innate ability
reduces the incentive they face to work hard
and do well in school. Second, under some
conditions, rational parents will not always
prefer the highest value-added schools,
and rational schools will not always choose
to compete on value added. These implica-
tions are consistent, for example, with the
well-identied empirical evidence that selec-
tive schools only sometimes produce higher
learning (e.g., Clark 2010; Abdulkadiroglu,
Angrist, and Pathak 2014; Pop-Eleches and
Urquiola 2013).
4.2 Vouchers and Political Economy
The research reviewed thus far studies
implications of vouchers but does not analyze
the endogenous public choice of voucher
policy, a subject of obvious importance given
the poor performance of voucher proposals
in referenda in California and Michigan.
Ireland (1990) provided the foundation for
research on this issue. In his framework,
households obtain utility from the education
of their children and from the consump-
tion of other goods. Households’ demand
for educational expenditure is increasing in
income. Expenditures on public schools and
on a voucher, if any, are funded by a propor-
tional income tax. Ireland investigates how
public-school spending is impacted by the
introduction of a universal voucher smaller
than per-student public-school expendi-
ture. The effect may be either an increase
or decrease, depending on whether the
reduction in outlay for students who switch
from public to private school in response to
the voucher is sufcient to offset the cost of
providing a voucher to students who would
attend private school anyway.
Ireland treats the voucher and tax rate
as exogenous; subsequent work has sought
to model these as chosen by majority rule.
This effort encounters two challenges. First,
the policy vector has three variables (tax
rate, public-school expenditure per pupil,
voucher). Invoking the public-sector bud-
get constraint eliminates one of these vari-
ables, leaving a two-dimensional choice
set and the accompanying challenges for
analyzing majority rule set forth by Plott
(1967). Second, even if one variable, say the
voucher, is exogenous, preferences over the
tax rate are not single peaked. Epple and
Romano (1996) investigate the second of
these issues, considering voting over the tax
rate that funds public educational expen-
diture and the voucher, taking the voucher
amount as given.40 They show that, despite
the non-single-peaked preferences, equilib-
rium under majority rule is likely to exist for
realistic parameter values. The equilibrium
is of an ends-against-the-middle form, with
a coalition of poor and wealthy households,
comprising half the population, favoring a
reduction in the tax rate and middle-income
households, comprising the other half, pre-
ferring an increase.41
Work to endogenize voucher choice has
followed two avenues. One is to consider vot-
ing one issue at a time. The other is to limit
the choice set in other ways, such as requiring
that the voucher equal public expenditure
or by considering voucher-only economies.
Hoyt and Lee (1998) endogenize both the
voucher and tax rate by considering sequen-
tial voting with the voucher determined rst
and then the tax rate second. Employing
information on the income distribution in
each state, they nd that there are some states
in which introduction of a $1,000 voucher
would permit lowering the tax rate without
lowering public expenditure per student.
40 See also Glomm and Ravikumar (1998).
41 In a very similar model, assuming existence of majority-
choice equilibrium, Rangazas (1995) identied the trade-
offs in the public choice of expenditure in the public school
for a given voucher. His computational analysis predicts
that a voucher equal to 1.25 percent of a teacher’s annual
salary would cause per-student public expenditure to
increase. Investigation of “ ends-against” voting is under-
taken by Brunner and Ross (2010).
465
Epple, Romano, and Urquiola: School Vouchers
Chen and West (2000) take the voucher
as equal to public-school expenditure per
student in their examination of the political
economy of income targeting. If vouchers
produce some efciency gains, they nd that
the targeted regime is likely to be majority
preferred both to the no-voucher status quo
and to a nontargeted voucher regime.
Another approach, voucher-only econ-
omies, is employed by Fernandez and
Rogerson (2003) to study vouchers in a
dynamic setting in which education spend-
ing impacts adult earnings. They consider
three alternative voucher programs: a uni-
versal at-rate voucher, a means-tested
voucher, and a “ means-equalizing” voucher
that depends on household income and the
amount of income devoted to education.
They nd that all three alternatives increase
utilitarian welfare substantially, relative to
the purely private system, and all tend to cor-
rect the inefciency from low investment on
the part of poor households.42
Bearse et al. (2013) continue the study
of income-targeted programs by consider-
ing a voucher that is positive for the lowest-
income household and declines linearly with
income to zero. A sequential voting equilib-
rium, with the tax rate chosen rst, followed
by the parameters of the voucher program,
is shown to exist. Their computational model
shows that, compared to the no-voucher pub-
lic equilibrium, the means-tested voucher
chosen by majority rule benets the poor via
higher education spending and a lower tax
rate, while also benetting wealthy house-
holds who prefer private schooling coupled
with a low tax rate.
An alternative approach to overcome the
Plott existence issue is adopted by Epple and
42 Fernandez and Rogerson (2003) note that their three
voucher systems can be viewed as analogous to three dif-
ferent systems of state grants to local districts (foundation,
means tested, and power equalizing). Hence, their analysis
can alternatively be viewed as informing the political econ-
omy of public-school nance.
Romano (2014). They analyze simultaneous
voting over the tax rate, public-school expen-
diture per student, and the voucher, exploit-
ing the citizen-candidate model of Besley
and Coate (1997). They provide necessary
and sufcient conditions for equilibrium,
and show computationally that equilibrium
exists for realistic parameter values. They
also show that a voucher is likely to gar-
ner greater political support when income
inequality is low. Intuitively, when inequality
is low, a relatively small number of house-
holds choose private school. A marginal
increase in the voucher induces a relatively
large exodus from the public schools, permit-
ting an increase in public-school expenditure
per student with a lower tax rate, a change
that garners unanimous support.
Epple, Romano, and Sarpca (2014)
extend this model to include income target-
ing via simultaneous voting over the tax rate,
expenditure per student in public schools,
the voucher amount, and the maximum
income of households eligible for vouchers.
They nd that income targeting increases
political support for vouchers by limiting
their use by high-income households that
would use private school even in absence
of vouchers, and that a targeted voucher
always garners political support. The pref-
erence for targeted vouchers conforms to
observation, but the nding that a targeted
voucher would always garner political sup-
port does not. This latter nding brings
to the fore limitations of the workhorse
Ireland (1990) framework, particularly the
assumptions that all households have the
same preference function and differ only
in income. Evidence on voting by legisla-
tors for voucher proposals discussed in sec-
tion 5, along with differences in opinions
evoked by vouchers, point to consideration
of ideological differences in preferences as
an avenue for extending the Ireland frame-
work. This is being pursued in ongoing
research of Epple, Romano, and Sarpca.
Journal of Economic Literature, Vol. LV (June 2017)
466
To summarize the research on the polit-
ical economy of vouchers, two prominent
themes emerge. One is that the majority
of voters, those who do not choose private
school in the absence of vouchers, will ben-
et by targeting vouchers so as to prevent
take-up by those who would attend private
school anyway. The other is that a voucher
offering less than per-student expenditure in
public school will generally be preferred by
those who would continue to attend public
school. Such a partial voucher induces some
households to switch to private school, and
this yields a net tax savings to those attend-
ing public school equal to the differential
between per-student public spending and
the voucher.
5. The Empirical Evidence
This section reviews the empirical evi-
dence on each of the ve questions raised
above. For each, we highlight the method-
ological challenges that arise and we focus
the review on the papers that have most suc-
cessfully dealt with these challenges. This
implies that we discuss some voucher pro-
grams more than others, depending on the
question at hand.
5.1 Question 1: What effects do vouchers
have on the students who use them?—The
key challenge in answering this question is
establishing credible counterfactuals; e.g.,
what would the outcomes of voucher win-
ners have been had they not received a
voucher? While different types of research
attempt to do this, the papers on small-scale
voucher programs are the most focused on it
and have tackled it with the highest degree
of credibility.
This reects that, in many cases, their
setup at least emulates a randomized con-
trolled trial in which subjects are randomly
assigned to treatment (voucher) or control
( no-voucher) groups. Specically, some
publicly and privately funded voucher pro-
grams have been explicitly designed as exper-
iments. In other instances, nonexperimental
programs are oversubscribed, and random
assignment arises from the use of lotteries to
allocate scarce slots. These cases, at least in
principle, establish a clear counterfactual—
on average the observed and unobserved
characteristics of treated and untreated
groups should be identical, and therefore
simple comparisons of their achievement
can reveal the causal effect of vouchers.
One aspect to bear in mind is that in all
the programs we discuss, those who are
offered a voucher are not required to use
it. Hence, a distinction arises between the
effects calculated by focusing the compar-
ison on those who have been offered the
voucher and those who actually take it up.
A comparison of the average outcomes of
those offered and not offered the voucher
yields an “Intent to Treat” (ITT) estimate.
A “Treatment on the Treated” (TOT) esti-
mate adjusts for the proportion of students
who take up the voucher—thus providing
an approximation to the effect of the treat-
ment on those who received it. Both types
of estimates have analytical advantages. For
instance, the ITT estimate might provide a
reasonable approximation to the effect of
implementing a small-scale voucher scheme,
since it is always the case that the proportion
of students who take up the voucher is less
than one.
Finally, in some cases we also review
the results of papers that, facing a lack of a
(quasi-) experimental counterfactual, seek
to establish one by using matching tech-
niques or otherwise attempting to control
for observable characteristics. In such cases,
one must bear in mind that estimates can still
be biased if unobserved student or parental
characteristics are correlated with treatment.
To preview the bottom line on question 1,
the evidence does not suggest that awarding
students a voucher is a systematically reliable
467
Epple, Romano, and Urquiola: School Vouchers
way to improve their educational outcomes.
A perhaps surprisingly large proportion of
the best-identied studies suggest that win-
ning a voucher has an effect on achieve-
ment that is statistically indistinguishable
from zero. Moreover, three recent studies
nd large negative effects on test scores of
voucher recipients. This is contrary to what
one would expect, for example, if private
or independent schools had systematically
higher value added. There is, however, recent
evidence from two randomized-control stud-
ies that point to more favorable effects on
attainment. There is also evidence that in
some settings, or for some subgroups or spe-
cic outcomes, vouchers can have substan-
tial positive effects on those who use them.
A question is, therefore, what accounts for
the variation in estimated impacts? The liter-
ature offers some tentative and useful clues,
but no denitive guidance. This reects two
aspects we will be explicit about. First, the
best evidence on question 1 comes from very
different settings—in this section we review
studies on the United States, Colombia, and
India—and while all these provide useful
evidence, extrapolating is difcult, as these
settings vary along multiple dimensions.
Second, the experimental studies can provide
clear counterfactuals and credibly answer
question 1, but they deliver a “reduced-
form” answer that does not fully reveal what
mechanisms account for the effects—a fur-
ther reason for why extrapolation is difcult.
5.1.1 The United States
Wolf et al. (2010a, 2010b) report on the
Washington DC Opportunity Scholarship
Program, which used an experimental design.
Their sample consists of roughly 2,300 stu-
dents, of whom about 60 percent were
offered a voucher, with the rest serving as a
control group. Of those offered a voucher,
77 percent made use of it. The authors nd
no signicant impact on test scores after
one, two, or four years (a signicant effect
emerges for reading after three years, but
none for math). Overall, there is little evi-
dence that the Washington DC Opportunity
Scholarship Program resulted in a sustained
improvement on test scores.
In contrast, Wolf et al. (2010a) report
that the program had a large and statisti-
cally signicant impact on graduation rates.
After four years, students who were offered
a voucher (ITT) were 12 percentage points
more likely to graduate than those who were
not, with a corresponding TOT effect of
21percentage points. Exploring heterogene-
ity in impacts, Wolf et al. nd similar effects
among students originally in schools desig-
nated as “in need of improvement.”
The School Choice Scholarship Foundation
created three voucher programs—New
York City, Dayton, and Washington, DC—
that also conform to experimental design.
The most intensively studied of these is
the one in New York, where in 1997 a lot-
tery was conducted among approximately
11,000 applicants (Peterson et al. 2003).
None of these experiments yield signicant
test score effects for non–African American
students. Nonetheless, Mayer et al. (2002)
nd the program increased the test scores
of African American students in New York
by about 6 percentile points (ITT). Krueger
and Zhu (2004) show, however, that this nd-
ing is sensitive to how ethnicity is coded, as
well as to how one handles students with
missing baseline scores.43 In Washington,
Peterson et al. (2003) nd signicant effects
for African American students at the end of
two years of treatment, but these vanish by
the third. In Dayton, they nd that African
Americans had a 4 percentile point advan-
tage at the end of two years (signicant at
the 10 percent level). Overall, this group of
experiments suggests some—albeit not very
robust—indication of test score effects for
43 On this issue see also Barnard et al. (2003), Krueger
and Zhu (2003), and Peterson and Howell (2004a, 2004b).
Journal of Economic Literature, Vol. LV (June 2017)
468
African American students, and none for the
rest.
As in the DC Opportunity Scholarship
Program, however, these experiments pro-
duced better results in terms of graduation
rates, at least for African American students.
Specically, Chingos and Peterson (2015)
study college enrollment as another outcome
in New York. At the time of the experiment
(late 1990s), participants were in grades K–5.
Chingos and Peterson were able to obtain
follow-up information on college enrollment
for a remarkable 99 percent of the roughly
2,700 students in the original study. They
nd no signicant differences between the
treatment and control groups as a whole, but
a signicant difference for African American
students—for those offered a voucher (ITT),
the estimated increase in part- and full-time
enrollment is 7 percentage points, a 20 per-
cent increase.
The above papers focus on voucher pro-
grams that were designed with an explicit
experimental setup in mind. This was not the
case in Milwaukee, which displays variation
in its ability to deliver clear counterfactuals
over time. Specically, in its early years the
Milwaukee program featured randomization
due to oversubscription, but starting in 1994,
increases in the income cap and the incor-
poration of Catholic schools had the effect
of making vouchers generally available to
most eligible students without recourse to
lotteries.
Rouse (1998) analyzes impacts during
both periods.44 First she exploits random-
ization, nding little evidence of effects for
reading. Her estimates (both ITT and TOT)
for mathematics are substantial—statistically
signicant effects of 0.3 to 0.5 standard devi-
ations over a four-year period. She also fol-
lows Witte (1997) in considering a random
44 See also Witte, Sterr, and Thorn (1995) and Greene,
Peterson, and Du (1996).
sample of public-school students as the com-
parison group, obtaining similar ndings.
Witte et al. (2008, 2009, 2010, 2011, 2012)
conduct a ve-year study of the effects on
voucher recipients (TOT) using a match-
ing strategy. They nd no signicant effects
on test scores in the rst, second, and third
year of the program. A statistically signicant
fourth-year gain of 0.15 standard deviations
emerges in reading; gains in mathematics
are also evident, but are signicant only for
students in grade seven. Overall, this analy-
sis suggests that during this post-expansion
phase, the Milwaukee voucher program had
little, if any, effect on test scores. In con-
trast Cowen et al. (2011) and Cowen et al.
(2013) use a matching strategy and nd pos-
itive impacts on years of schooling, although
the results are not entirely robust to differ-
ent specications. Beyond potential biases
from unobserved characteristics, a concern
with these longer-term matching-based esti-
mates emerges from the possible existence
of contemporaneous non-voucher-related
policies. For instance, starting in 2002 the
public reporting of schools’ test scores began
to be required, and this might have affected
schools’ performance quite aside from
vouchers.
While the studies discussed above nd
little systematic evidence of positive effects
of vouchers on achievement, they do not
nd signicant negative effects. By contrast,
Abdulkadiroglu, Pathak, and Walters (2015)
(APW henceforth) nd large negative test
score effects for the Louisiana Scholarship
Program (LSP). APW study test scores in
2012–13, the rst year following statewide
adoption of the Louisiana voucher program,
a program that previously had been available
only in New Orleans.
APW exploit lottery selection of students
to oversubscribed private schools for LSP
voucher applicants in third through eighth
grades, studying effects on scores in stan-
dardized tests of math, science, social studies,
469
Epple, Romano, and Urquiola: School Vouchers
and English. APW nd that 1,412students
who applied to oversubscribed schools were
selected by lottery, with subsequent test
scores available for upwards of 85 percent
of these students. Causal local average treat-
ment effects on math, science, and social
studies are −0.413σ, −0.263σ, and −0.331σ
respectively, all signicant at the 1 percent
level. A negative but insignicant effect is
found for English.
APW conduct extensive analyses to inves-
tigate robustness and possible differences in
impacts of the voucher across student income
groups and geographic areas. Their ndings
of negative impacts of vouchers are robust,
e.g., with estimation methods accounting for
sample attrition, and are found for all income
groups and geographic areas. Participating
private schools are found to have falling
enrollments prior to their participation in
LSP, suggesting that the voucher program is
attracting schools that are struggling. It will
be important to investigate whether these
negative effects persist for subsequent years
of the LSP, but as APW conclude, “These
results suggest caution in the design of
voucher systems aimed at expanding school
choice for disadvantaged students.”
A rst indication regarding persistence
of negative effects of the LSP emerges
from work by Mills and Wolf (2016). These
authors use a slightly larger sample of LSP
students, and are able to expand the analysis
to a second year. Mills and Wolf state that
their rst-year results largely match those
in APW. For the second year, they still nd
negative effects, although somewhat smaller
in magnitude.
Figlio and Karbownik (2016) also nd
large negative and signicant effects on
both math and reading scores of voucher
recipients in the statewide Ohio “EdChoice
Scholarship Program,” notably persisting
for students through the three years of the
study. Voucher-eligible students have their
designated public school deemed to be low
performing, but must also be accepted by
a participating private school. Mathematics
scores decline each year by around −0.5σ
and reading by around −0.3σ, estimates that
are very robust to alternative specications.
Because participating private schools can
select students, propensity score matching
is employed for identication. Specically,
voucher takers in barely eligible schools are
matched on observables to ineligible stu-
dents in public schools that barely exceeded
low performance. As the authors discuss,
this sharp identication limits the treated
students to 445, implying that effects on
voucher takers in the worst public schools
might differ.
Why students in these studies elect to
attend private school where they perform
relatively poorly on tests is an important and
open research question. Students may have
alternative goals, e.g., religious study, or may
be making mistakes. It is notable that expen-
diture per student in the attended private
schools is generally lower than in their public
alternatives.
Overall, the evidence on the United States
nds not very robust effects on test scores,
most frequently nonexistent, some positive
effects on blacks, but also the just-discussed
large negative effects. More robust evidence
has accumulated regarding positive impacts
on graduation probabilities, particularly for
black students.
5.1.2 Colombia
Colombia yields perhaps the most positive
evidence that emerges from (quasi-)exper-
imental work on question 1. Specically,
Angrist et al. (2002) exploit voucher lot-
teries implemented in the cities of Bogotá
and Cali. In terms of test scores, their key
nding is that three years after the alloca-
tion, voucher winners scored 0.2 standard
deviations higher on achievement tests.
Using a similar design, Angrist, Bettinger,
and Kremer (2006) nd that, correcting for
Journal of Economic Literature, Vol. LV (June 2017)
470
differences in test taking between lottery
winners and losers, the program increased
high-stakes college admissions test scores
also by 0.2 standard deviations. They fur-
ther nd an effect on longer-term outcomes:
voucher winners were 15–20 percent more
likely to complete secondary school, less
likely to work while in school, and less likely
to marry or cohabit as teenagers.
In short, the experimental results from
Colombia are more positive than those from
the United States in every measured dimen-
sion. A relevant question is: why are voucher
winners in Bogotá and Cali beneting more
consistently than those in New York City or
Washington DC? None of the experimental
studies reveal the precise channels through
which its effects work—the clear identica-
tion comes at this cost, to some extent—so it
is not possible to answer this question den-
itively. Nevertheless, the institutional differ-
ences between the US and Colombia voucher
experiments render some differences more
or less likely as potential candidates.
Is it possible that public schools in
Colombia are much weaker than in the
United States, and so the opportunity to use
a private school has a large effect? This may
play a role but is unlikely to provide a full
explanation, as—somewhat unusually—both
voucher winners and losers in Colombia
tended to enroll in private schools. For
example, Angrist et al. (2002) point out that
while about 94 percent of lottery winners
attended private school in the rst year, so
did 88 percent of the losers. This partially
reects (section 3.1.2) that a requirement for
application was to have been accepted at a
private school (the stated goals of the pro-
gram were related to raising enrollment rates
by increasing private participation). Thus,
one might reasonably expect the program to
be more attractive to students strongly inter-
ested in private school anyway.
Is it possible that the positive ndings
reect greater resources at the receiving
schools? This is a possibility, since winners
on average used the voucher to “upgrade”
to more expensive institutions—Angrist
et al. (2002) report that vouchers “crowded
in” educational expenditure to some extent.
There is less data on the resource changes
that vouchers imply in the United States.
A nal possible channel relates to student
incentives. In Colombia the vouchers were
renewable contingent on grade completion,
and thus the program included an incentive
component—voucher winners faced a stron-
ger reward for doing well at school. Thus,
superior performance could have been due
to incentives, rather than to the voucher pro-
vision itself.45
To summarize, the Colombian experiment
suggests that vouchers had a positive effect
on tests scores and several other outcomes,
but the difculty in attributing effects to a
precise channel makes it difcult to draw
precise policy implications.
5.1.3 India
Turning to India, Muralidharan and
Sundararaman (2015) analyze an experiment
in villages of the state of Andhra Pradesh, as
described in section 3.1.3. They nd that four
years after treatment, lottery winners did not
have higher test scores than losers in Telugu
(the local language), Math, English, Science,
and Social Studies; in contrast, there was a
positive and signicant effect in Hindi. The
authors interpret this as an overall positive
effect, since private schools achieve higher
Hindi results with no disadvantage in the
other tests.
45 Digging even deeper into mechanisms, the gains in
Colombia could have also reected peer effects. However,
Bettinger, Kremer, and Saavedra (2010) suggest that
at least for a subgroup of PACES beneciaries, at least
observable peer quality does not seem to account for the
results. Specically, they show that the advantage found in
Angrist et al. (2002) persists, even when one considers stu-
dents who chose to attend vocational schools with typically
lower observable peer quality.
471
Epple, Romano, and Urquiola: School Vouchers
Interpreting the effect in Hindi is com-
plicated by the fact that public schools in
Andhra Pradesh essentially do not teach
Hindi at all. More specically, Muralidharan
and Sundararaman use interesting comple-
mentary data to show that the private schools
spend much more time teaching Hindi
(essentially relative to a base of zero in public
schools) and substantially less on the remain-
ing subjects, except for English.46
The setting this paper explores also
raises interesting contrasts with the US and
Colombia cases described above. Importantly,
in contrast to Colombia, where voucher win-
ners beneted from greater expenditure,
Muralidharan and Sundararaman (2015)
point out that private schools in the villages
analyzed have expenditures that are only
about one-third of those in public schools.
This reects that many private schools in
India operate without subsidies, even as they
cater to very low-income individuals. This is
observed in other settings, including Pakistan
and parts of Africa (e.g., Andrabi, Das, and
Khwaja 2013); in these areas, private expan-
sion has been observed on a magnitude that
in middle- or high-income countries would
seem to require signicant public subsidies.
This might reect parental willingness to
escape a deeply dysfunctional public sector
in which there is evidence of rampant absen-
teeism by teachers; see, for example, Duo,
46 The authors further note that not all private schools
use Telugu or Hindu as the language of instruction, with
some using English instead. While acknowledging that this
is an endogenous choice (and hence not a feature of the
experimental design), they present suggestive evidence
that the choice of English as a medium of instruction dis-
rupts learning. If parents do not realize this is the case,
then further regulation of private schools may be war-
ranted. But another possibility is that parents are aware
of this but willing to make the sacrice if, for example,
English has high labor-market returns. A broader point the
ndings around Hindi and English illustrate is that choice
is likely to produce more of what parents want that may
or may not be skills along the precise dimensions policy
makers prefer.
Hanna, and Ryan (2012) and Muralidharan
and Sundararaman (2011).
Muralidharan and Sundararaman (2015)
emphasize that, coupled with the above
ndings on achievement, the fact that the
private schools spend so much less implies
a substantial private productivity advantage.
They also show that the crux of this cost dif-
ference is in teacher salaries—private-school
teachers are younger, typically less trained,
more often female, and on average make
only one-sixth the salary of their public-
sector counterparts. In short, private-school
students have instructors who are paid much
less. On the other hand, in some dimensions
private-school students have access to more
educational resources. For instance, they
enjoy lower class sizes, and the probability
that they are in multi-grade classrooms is
lower by about 50 percentage points.47
In short, private and public schools in
Andhra Pradesh are organized very differ-
ently, but on net have no testing performance
differences except in Hindi, which the pub-
lic schools do not teach. It is again the case
that the myriad of differences in setting pro-
vide interesting implications, but complicate
extrapolation. Just to cite one example, the
voucher schools in Andhra Pradesh, India,
on average have lower class sizes than public
schools; the opposite is the case in present-
day Chile.
To summarize, question 1 is perhaps the
most straightforward among those we ask,
and yet the above review makes clear that
the answer to it is complex. The results are
not clear-cut—in many cases and for many
outcomes transferring students to private
school does little to their achievement; in
others, it improves or lowers it substantially.
In addition, the most rigorous research on
this question typically delivers reduced-form
47 The research in Duo, Dupas, and Kremer (2011)
suggests this might have a large favorable impact on
performance.
Journal of Economic Literature, Vol. LV (June 2017)
472
results. These do not reveal the mechanisms
that account for the differences, making it
hard to draw clear implications. It is relevant
to point out that the evidence—particularly
that from the United States—is consistent
with vouchers improving some types of skills
more than others. For example, it is possible
that the positive effects on graduation rates
in the United States stem from improve-
ments in noncognitive skills, while the lack
of a consistent impact on test scores points
to weaker impacts in a cognitive dimension.48
5.2 Question 2: Do vouchers induce
nonrandom migration from public to private
schools, possibly lowering the achievement of
students that remain in the public sector?—
Another fundamental question on vouchers
is whether these lead to sorting. As discussed
above, this is the common prediction of
theoretical work. However, that work also
makes clear that the type of sorting observed
will be a function of the rules governing the
voucher system.
We begin by considering the evidence
on large-scale voucher programs. These are
analytically suited to addressing question 2
in that they provide a chance to study situa-
tions in which large numbers of students of
all types are given a chance to exercise school
choice, and in which schools get a chance to
enter and exit the market in response.
Yet the evidence from large-scale pro-
grams has disadvantages too. Most impor-
tantly, almost by denition, very clear
identication is difcult to obtain from these
programs. They involve the distribution of
vouchers to anyone who wishes to use them,
and are typically implemented countrywide.
Thus, it is very difcult to establish clear
counterfactuals regarding what would have
happened in the absence of these programs,
48 We thank an anonymous referee for this observation.
and of course randomizing at the country
level is not feasible.
5.2.1 Large-Scale Voucher Programs:
Chile and Sweden
The original design of Chile’s voucher
scheme allowed private schools ease in set-
ting up admissions policies. These could
include features such as admissions exams
and parental and student interviews.49 In
addition, since the mid-1990s, schools have
been able to charge tuition add-ons. Under
such circumstances, models like Epple and
Romano (1998) and MacLeod and Urquiola
(2009, 2015) suggest that the introduction
of vouchers would lead to cream skimming
from the public sector, and stratication by
income and/or ability within the private
sector.
Hsieh and Urquiola (2003, 2006) analyze
this by looking at the growth of the private
sector across municipalities. They essen-
tially implement a difference-in-differences
analysis that asks if stratication measures
increased more in markets with greater
growth in the private-voucher sector. Again,
this is not equivalent to a randomized exper-
iment and biases could arise, for instance,
from preexisting trends. Nevertheless,
Hsieh and Urquiola nd evidence that the
voucher-induced growth in the private sec-
tor was associated with a “middle class”
exodus from public schools consistent with
cream skimming; this is robust to the use of
candidate instrumental variables for private
growth.
McEwan, Urquiola, and Vegas (2008)
additionally present a descriptive analysis
of sorting in small markets. The idea—one
related to an approach used in industrial
49 Over the years, more restrictions on selection by
private schools have been written into law. In our under-
standing, these include generalized statements regarding
prohibitions on selection rather than specic, legally bind-
ing mechanisms (e.g., centralized lotteries).
473
Epple, Romano, and Urquiola: School Vouchers
organization—is that xed costs determine
that private voucher schools must be of at
least a minimum size in order to break even.
This implies that very small markets—say
very small towns—will tend be served by
public schools only. Larger towns will have
private participation. McEwan, Urquiola,
and Vegas (2008) focus on the range of
towns (by population size) where private
entry is rst observed, comparing the degree
of sorting observed just below the approxi-
mate size threshold that determines private
entry to that observed just above. This is akin
to a “fuzzy” regression discontinuity design,
albeit one with limited statistical power.
They nd that private entry is indeed related
to stratication, consistent with the rst pri-
vate-school cream skimming the highest
ability/income kids from the public sector.
Finally, a large number of studies point to
cross-sectional evidence of high stratication
in Chile. For instance, Valenzuela, Bellei,
and de los Rios (2010) suggest that Chile dis-
plays one of the highest levels of school-level
stratication by socioeconomic status in the
OECD. In addition, Mizala, Romaguera,
and Urquiola (2007) suggest stratication is
particularly extensive in the private sector.50
In closing the discussion on Chile, it is
worth mentioning that there is widespread
consensus among observers and policy mak-
ers there around the claim that vouchers
have facilitated sorting. In addition and as we
discuss below, the high degree of stratica-
tion in the school sector seems to be a signif-
icant contributing factor to student protests
that have persisted over a number of years.
Stratication has been less of a focus on the
research in Sweden. Speculating, this might
reect that Sweden’s voucher design likely
promotes sorting less than Chile’s. As stated,
50 For other examples of large-scale school market lib-
eralization leading to stratication, see Lucas and Mbiti
(2012) for the case of Kenya. For related evidence in the
United States, see Urquiola (2005).
independent schools must partially apply a
rst-come-rst-served criterion if oversub-
scribed, and tuition add-ons are not allowed.
Nevertheless, research points to some sort-
ing effects. For instance, Sandstrom and
Bergstrom (2005) report that indepen-
dent-school students are more likely to be
immigrants and to have parents with higher
income and education (see also Bjorklund
et al. 2006). Further, Bohlmark and Lindahl
(2007) nd that public schools tend to lose
students who are second-generation immi-
grants or whose parents are more highly
educated, but nd no evidence of sorting by
income.
In more recent work that considers effects
through 2009, Bohlmark, Holmlund, and
Lindahl (2015) nd that after accounting
for residential sorting, school segregation
increased more in municipalities in which
school choice became more prevalent (as
measured by the number of independent
schools in operation). Specically, segrega-
tion in such areas has grown between immi-
grants and natives, and between children of
parents with high/low levels of education.
Nevertheless, they conclude that the magni-
tude of the effect of choice on segregation
is relatively low, and that Sweden still ranks
as a country with relatively low levels of
across-school sorting along such dimensions.
5.2.2 Small-Scale Voucher Programs
The literature on small-scale programs has
also explored question 2. These programs
can be expected to have a different impact
on sorting (relative to large-scale programs)
due to the fact that they tend to display
four features: (1) targeting of low-income
students, (2) lotteries in cases of oversub-
scription, (3) requirements that voucher pro-
ceeds be used for educational expenses (i.e.,
not to provide nancial aid), and (4) rules
against tuition “ top-ups.” In addition, in
the case of the United States the potential
for voucher-induced sorting is set against a
Journal of Economic Literature, Vol. LV (June 2017)
474
backdrop of a public-school system in which
parents can sort into school districts or catch-
ment areas (Tiebout 1956). Nevertheless,
some scope for cream skimming remains
since the more afuent and/or more able
among the eligible population may be more
likely to apply for or use a voucher.
Turning to the evidence, we begin with
the largest voucher system in the United
States: Milwaukee’s. Chakrabarti (2013b)
investigates sorting during its rst ve years,
1990–94. She nds signicant evidence that
the probability of applying to the program
rises with individual ability, but not with
income. Among winners, there is some evi-
dence that the probability of take-up rises
with income but not with ability. Given the
income targeting of the Milwaukee program
during this period, any income stratication
with respect to voucher use is limited to vari-
ation within the population of low-income
students. Hence, the nding of stratication
by ability is more salient from the perspec-
tive of concerns about cream skimming.
As it has grown, Milwaukee’s program has
changed in ways that increase the poten-
tial for sorting. First, the income eligibility
limit was raised to 300 percent of the pov-
erty level. Second, a tuition premium for
high-school students with household income
more than 220 percent above the poverty
limit was allowed.51 Fleming et al. (2013)
analyze sorting in Milwaukee during this
more recent period. They compare voucher
students to a random sample of Milwaukee
51 That these two changes would go together is in some
sense not surprising—raising the income eligibility limit
results in a concurrent increase in political support for
permitting private schools to charge a tuition premium to
voucher students. When only low-income households are
eligible for vouchers, few could afford a tuition premium,
making the issue moot. As higher-income households
become eligible, more are willing to pay a premium for
more costly private schools. The conditions giving rise to
political support for permitting tuition premia are thus the
same as the conditions likely to give rise to cream skim-
ming, if such premia are permitted.
public-school students in school year 2006.
They nd that voucher students are signi-
cantly more likely to be black or Hispanic,
English language learners, and female. They
also use a matched panel to consider paren-
tal characteristics. They nd that voucher
parents report signicantly lower incomes.
Their overall education levels are also lower,
but there is a somewhat higher college par-
ticipation rate among them. Perhaps the
most salient difference in the matched panel
is that a much higher proportion of voucher
recipients report being Catholic (29 percent
versus 19 percent) and attending religious
service at least once a week (62 percent ver-
sus 48 percent).
Wolf et al. (2009) investigate the take-up
of vouchers in Washington, DC. Of about
1,400 students offered a voucher, 41 per-
cent used it for the full ( three-year) period,
34 percent made partial use of it, and 25 per-
cent never used it. Comparing ever-users to
never-users, the most prominent difference
was that ever-users were only one-third as
likely to have a learning or physical disabil-
ity.52 There were no signicant differences
in baseline test scores, family income, or
mothers’ education. Ever-users were some-
what more likely to be African American,
less likely to be Hispanic, and somewhat less
likely to be male.
There is less evidence on the effects
of tax-credit-funded vouchers on sorting.
An exception is work by Figlio, Hart, and
Metzger (2010), who study the FTC program.
They nd that, compared to nonparticipants,
voucher participants attended lower-per-
forming public schools. Moreover, voucher
students were among the lower-performing
students in the public schools they had
attended. Hart (2014) presents results
52 Similarly Wolf, Witte, and Fleming (2012) estimate
that between 7.5 to 14.6 percent of Milwaukee voucher
students have disabilities, as compared with 19 percent for
Milwaukee public schools.
475
Epple, Romano, and Urquiola: School Vouchers
consistent with this, suggesting that voucher
participants tend to come from schools
with worse academic performance and
higher rates of violence. In addition, they
are more likely to exit public schools when
they encounter more convenient (e.g., by
distance) and varied private-school options.
Finally, they are more likely to exit schools
with high concentrations of African American
students, and the latter holds regardless of
the student’s own race.
To summarize, the evidence suggests
that, perhaps not surprisingly, vouchers can
result in the nonrandom reallocation of stu-
dents across and within sectors. That said,
the details of program design clearly matter.
For one example from among large-scale
programs, Chile’s design generally facilitates
sorting more than Sweden’s.
5.3 Question 3: Do voucher programs
pressure public schools to become more ef-
cient?—A key reason to introduce competi-
tion into any industry is the possibility that
it will lead to productivity improvements.
Question 3 asks whether vouchers induce
these in the public sector, perhaps as its
schools attempt to fend off gains on the part
of private competitors.
To illustrate the methodological chal-
lenges that this question raises, Hsieh and
Urquiola (2003) point out that a rst-pass
answer to it is provided by calculating the
difference in average public-school perfor-
mance before and after the introduction of
vouchers. They point out however, that one
would ideally want to decompose this change
into three effects:
(1) The public sector’s change in value
added—essentially the object of inter-
est in question 3
(2) A composition effect—for example,
vouchers may worsen public-sector
performance if they result in higher
ability/income children leaving to pri-
vate schools, and
(3) A peer effect—for instance, if there is
nonrandom sorting the performance
of those “left behind” may be affected
by no longer interacting with departing
students.
While (1) is the object of interest in this case,
the presence of (2) and (3) make it very dif-
cult to isolate this effect. Specically, in
order to isolate (1) it would be necessary to
control for children’s characteristics, some of
which may be unobserved. Further, even if
one observed all student characteristics, ade-
quately controlling for peer effects is difcult,
given that the literature has not produced a
consensus on the functional form of such
effects, or even on whether a stable functional
form exists (Carrell, Sacerdote, and West
2013). A clear prior on the direction of at least
some of the effects in (1)–(3) could provide
analytical leverage to empirically get a sense
of the direction of the others. However, as
noted in section 4, theory does not provide
unambiguous guidance on any of these.
Hsieh and Urquiola (2003, 2006) present
evidence of these difculties for the case of
Chile. For example, they show that public
performance worsened in municipalities that
experienced greater private growth after the
introduction of vouchers. They point out that
this could be driven purely by composition
effects—(2) in the listing above. They there-
fore call for a focus on how the introduction
of vouchers affects aggregate performance.
This does not solve the challenge posed by
(3), but it does control for (2) and poten-
tially comes closer to identifying the effect
of vouchers on overall school productivity.
We will return to that issue in the context
of discussing question 4 (which focuses on
aggregate, net effects); for now we review
the evidence on question 3, keeping in mind
the above challenges.
Journal of Economic Literature, Vol. LV (June 2017)
476
5.3.1 The United States
Hoxby (2003a) studies question 3 in the
context of Milwaukee. She measures the
intensity with which public schools face
competition by the proportion of their stu-
dents that are eligible for vouchers. In
Milwaukee, this measure varies because
eligibility for a voucher is dependent on stu-
dents’ income. In this approach, schools with
a low proportion of poor students serve as
the control group. As an additional control,
Hoxby chooses a set of low-income public
schools outside of Milwaukee. She nds that
more intensively treated public schools have
higher rates of productivity growth measured
as achievement on a standardized exam, rela-
tive to expenditure.
Chakrabarti (2008) extends Hoxby’s
approach, considering two periods in the
Milwaukee voucher system. She designates
1990–97 as Phase I; Phase II consists of years
after 1998, roughly when the program expe-
rienced changes including the expansion of
the proportion of eligible students and reli-
gious school eligibility. She uses school-level
data, with the proportion of students on free
or reduced lunch (FRL) as a measure of the
intensity of treatment. As controls, she uses
the thirty-three Wisconsin public schools
outside Milwaukee that had at least 25 per-
cent of their population eligible for FRL,
had a black population of at least 15 percent,
and were in locales similar to Milwaukee in
1990. Chakrabarti nds little evidence that
the voucher program had effects on public
schools in Phase I. By contrast, for PhaseII,
she nds signicant gains on the order of
0. 1–0.15 standard deviations; these are sta-
tistically signicant in reading and language
arts, but mostly not in mathematics and
science.
Figlio and Hart (2014) study the effect
of the FTC on public-school performance.
This program is available for students from
families with incomes below 185 percent of
the poverty line. Their logic is that the com-
petitive impact of voucher availability on
public schools will be greatest when public
schools have nearby private competitors.
They create four measures of private-school
proximity, and nd that public-school
achievement is related to each of these mea-
sures signicantly, albeit modestly: a one
standard-deviation increase in the number
of nearby private schools raises achievement
by 0.02 to 0.03 standard deviations.
A special feature of the (now canceled)
FOSP facilitates identication of program
effects, but potentially confounds their inter-
pretation for purposes of analyses related
to vouchers. Specically, in this program
Florida schools were graded based on stu-
dent performance on a series of standardized
exams. If a school received an F grade twice
within four years, its students became eligi-
ble for a voucher. Schools that received one
F faced a threat of a voucher, and the perfor-
mance of students in these schools can then,
for example, be compared to that of their
peers in schools that received a D grade.
Figlio and Rouse (2006) examine test-
score gains in schools receiving one F grade
using several strategies including regression
discontinuity, with D schools serving as a con-
trol group. They nd effects on high-stakes
mathematics tests (i.e., tests relevant to grad-
ing of the schools) of about 0.2σ, larger than
effects on low-stakes exams, suggesting non-
trivial gains but also some strategic focusing
of resources. However, they emphasize and
present some evidence suggesting that the
gains may be due to the stigma associated
with receiving another F grade, rather than
to the threat of loss of funding associated
with a voucher. In short, the program may
confound the effects of vouchers with those
of accountability.
Chakrabarti (2013a, 2013d) continues
the study of the FOSP program. Using
difference-in-differences and regression
discontinuity designs, Chakrabarti (2013a)
477
Epple, Romano, and Urquiola: School Vouchers
provides evidence that schools receiving
one F focused resources on improving the
scores of students predicted to be near the
boundary of the threshold of failure, and
on preparing for the writing exam, where
performance is believed to be more eas-
ily improved. She provides some evidence
that the voucher threat was important, for
example, nding larger effects of receiv-
ing an F grade on schools that faced more
private-school competition, as in Figlio and
Hart (2014). Chakrabarti (2013a) focuses on
pure gaming effects of the program by inves-
tigating whether schools classify students
with an eye to preventing their scores from
counting toward the school grade. This anal-
ysis again compares F to D schools, and nds
signicantly increased classication of stu-
dents as having limited English prociency.
While F schools might have tried to do the
same by classifying more students as special
education, their test scores also excluded
from the school’s grade, the gains from this
strategic reclassication would come at the
cost of these students then becoming eligi-
ble for another Florida voucher program.
These schools did not increase such clas-
sication. It bears repeating that studies
of the FOSP program provide interesting
avenues for identication of the program
effects, but the challenge of disentangling
the accountability and voucher effects weak-
ens implications that can be drawn for the
more common accountability-independent
voucher programs.
Figlio and Karbownik (2016, FK hence-
forth) study the competitive impact of the
Ohio voucher program, EdChoice, using a
regression discontinuity analysis. The idea
is to compare performance of students
assigned to public schools barely eligible
for vouchers to ineligible students in pub-
lic schools that barely avoided their stu-
dents being voucher eligible. The estimated
impacts are sensitive to functional form.
With linear functions on either side of the
voucher eligibility threshold, they nd large,
signicant three-year achievement impacts
of voucher eligibility ranging between 0.05σ
to 0.1σ, and four-year impacts of similar
magnitude. These estimates are robust to
including demographic controls and allow-
ing different slopes on the two sides of the
eligibility threshold. Analysis of impacts by
demographic groups suggests that these
estimated gains accrued primarily to white,
non-disadvantaged students. Insignicant
effects are found with polynomial functions.
Hence, while the results with linear func-
tions point to positive competitive impacts,
lack of robustness to more exible functional
forms argues for caution.
5.3.2 Canada
Chan and McMillan (2009) study the
effect on public-school performance from
a private-school tax credit in Ontario. This
was passed into law on short notice in June
2001, with the credit becoming available
in January 2002. The plan provided for a
private-school tax credit that was scheduled
to grow in increments over ve years, but the
program was canceled in December 2003,
retroactively to January 2003. Using the
2002–03 private-school enrollment share in a
public-school attendance zone as their mea-
sure of private-school competition, Chan and
McMillan nd that a 1 standard-deviation
increase in competition is associated with a
statistically signicant 0.1 standard-deviation
increase in the percentage of public-school
students achieving the provincial perfor-
mance standard for grade 3.
5.3.3 Sweden
Sandstrom and Bergstrom (2005) consider
whether students in Swedish public schools
perform better if they live in municipali-
ties that have a larger share of independent
schools. As stated, such an evaluation con-
fronts difculties that originate in the non-
random sorting that follows private entry, as
Journal of Economic Literature, Vol. LV (June 2017)
478
well as the endogeneity of this entry. The rst
of these might be somewhat mitigated, since
Swedish independent schools are not allowed
to select on ability. Nevertheless, applying to
an independent school is still endogenous.
Sandstrom and Bergstrom (2005) deal with
this by way of a Heckman correction. In order
to address the endogeneity of private entry,
they use variables approximating whether
local authorities are “hostile to independent
schools.” Specically, they proxy for these atti-
tudes using measures of the extent to which
municipalities contract out responsibilities to
the private sector. The identifying assumption
is that this attitude will only affect educational
outcomes through the channel that munici-
palities with less hostility will be less likely
to block independent-school entry. The key
nding is that the presence of independent
schools results in better public performance
in a GPA-type measure, as well as in stan-
dardized mathematics exams and an indicator
for whether students passed all three exams
necessary for high-school admission.
To summarize, several studies of
public-school response to voucher-school
competition have measured intensity of
competition either by the proportion of a
public school’s students who are potentially
eligible for vouchers, or by the proximity of
private-school alternatives. Virtually all of
these studies nd that public-school achieve-
ment increases with the intensity of treat-
ment. That said, most of these analyses do
not have an iron-clad strategy to deal with
potential biases from composition effects
(which section 5.2 suggests could be signif-
icant) or with potentially confounding poli-
cies such as accountability.53
5.4 Question 4: What is the net effect of
vouchers on educational outcomes?—As
53 Figlio and Hart (2014) study the effects of Florida’s
FTC program before implementation, and therefore avoid
sorting-related concerns.
stated above, some studies focus not on
questions 1–3, but rather on the net effect
of vouchers—question 4. The reason to do
this can be stated by summarizing some of
the difculties that arise in answering ques-
tions 1–3, and why even ideal answers to
these three questions may only give a partial
sense of the overall effect of vouchers.
Specically, as stated, question 1 can be
credibly addressed with experiments, but
doing so provides only a partial assessment of
the impact of vouchers. For instance, to the
extent that a private advantage is at least par-
tially due to a peer effect, then this gain will
not be independent of the size of the private
sector and/or the sorting its growth induces.
In other words, the advantage conferred by
transferring to a private school may dissipate
as the private sector grows and incorporates
weaker children. In some scenarios—e.g.,
if private schools are not more productive
and peer effects are linear in means, pri-
vate expansion may be zero sum (Hsieh and
Urquiola 2003).
Further, even a solid answer to ques-
tion 1 does not provide an assessment of
the consequences on the children not using
vouchers. Studying questions 2 and 3 begins
to provide a sense of this, but immediately
raises signicant empirical challenges. For
instance, if vouchers induce sorting, then it
is very hard to empirically isolate their effect
on public-school value added.
One alternative in the face of these dif-
culties—particularly when looking at
large-scale programs—is to simply analyze
market-level net effects (Hsieh and Urquiola
2003). This does not allow one to isolate the
specic channels through which vouchers
work, but can provide a sense of their aggre-
gate effect.
5.4.1 Chile
One of the more common approaches to
addressing question 4 has been to use panel
data for multiple local school markets. The
479
Epple, Romano, and Urquiola: School Vouchers
changes in the private sector’s reach within
these are then compared to the change
in average performance. This controls for
market-specic xed characteristics. In this
spirit, Hsieh and Urquiola (2006) apply
a difference-in-differences approach to
municipal-level data for 1982 to 1996, sug-
gesting that while municipalities with greater
private growth display clear signs of greater
sorting, they display no relative advantage in
terms of the evolution of achievement on stan-
dardized tests and years of schooling. As with
the analyses of the effects of large-scale pro-
grams on sorting (question2), the key source
of concern with these estimates—despite the
use of some candidate instrumental variables
(e.g., population density)—is that private
entry into school markets is endogenous. For
instance, if outcomes had been declining in
areas where the private sector grew more,
these effects would underestimate the salu-
tary effects of competition.
Other work implements a cross-sectional
variant of this idea by looking for an
instrument for the prevalence of voucher
schools and evaluating its effect on aggregate
performance. Here again, the challenge is
nding credible instrumental variables for
private enrollment. Auguste and Valenzuela
(2006) use distance to a nearby city and nd
evidence of cream skimming and signicant
positive effects on achievement. Gallego
(2006) uses the density of priests per dio-
cese and nds substantial effects on average
achievement.
Another way to consider aggregate
effects—the one that takes this logic the fur-
thest—is to simply look at aggregate country
performance, particularly in international
tests. Hsieh and Urquiola (2003) point out
that Chile’s performance did not improve
in the rst twenty years after the voucher
reform. Recent experience in this area has
been more favorable. After dropping by ve
and seven points respectively from 1999 to
2003, Chile’s 8th grade math and science
scores increased substantially—by twenty-
nine and forty-eight points respectively
from 2003 to 2011. Hanushek, Peterson,
and Woessmann (2012) estimate that Chile
had the second-highest growth rate among
forty-nine countries they studied for this
period. Looking at national tests, Neilson
(2013) shows evidence of improvement
among voucher schools in fourth-grade test
scores during 2008–12—this contrasts with
stagnant performance between 2003 and
2007. At the same time, this progress seems
to have signicantly decelerated, depending
on the subject tested, as measured by the last
PISA tests in 2013.
The literature features another approach
to look at the net aggregate effects of choice:
structural estimation. Papers that take this
route often achieve interesting analytical
richness, but also require strong assump-
tions. Beginning with test scores as an
outcome, Neilson (2013) argues that the
2008–12 improvement cited in the previ-
ous paragraph is due to a 2008 reform that
increased the voucher for low-income stu-
dents. Specically, the reform increased
the voucher by about 40 percent for the
poorest 40 percent of the population, with
schools having higher concentrations of
low-income students receiving even higher
payments. In exchange for this higher sub-
sidy, voucher schools were required to elim-
inate tuition top-ups for these students, and
to refrain from selective admissions. About
three-fourths of voucher schools eventually
chose to participate.
The paper uses rich data (that includes dis-
tance to school) to estimate school-specic
quality measures and a random utility model
of school choice by heterogeneous house-
holds. The results suggest that the targeted
voucher: (1) increased average school qual-
ity by 0.21σ, (2) increased average vouch-
er-school quality by 0.16σ, and (3) did not
affect quality at the non-voucher (elite) pri-
vate schools. These estimates are then used
Journal of Economic Literature, Vol. LV (June 2017)
480
to parcel achievement gains between those
that result as low-income students switch to
higher-quality schools (as tuition drops to
zero) and to increases in school quality. The
counterfactual with the targeted voucher
introduced but with no change in school
qualities is simulated to estimate the former
effect. The analysis suggests that one-third
of the achievement gain is from changes in
school choice, and two-thirds is due to school
quality improvements. Further research
here might consider further whether gains
can be explained “simply” by increased
school expenditure from the voucher.
This approach addresses very interesting
issues, but also requires strong assumptions.
For instance, period-specic school quality
is measured by regressing test scores on a
school xed effect and a number of controls.
This assumes there is no selection on unob-
servable characteristics; e.g., parental tastes
for educational achievement or religious
instruction.54 In addition, the framework
assumes that students can attend any school
they are willing to travel to and pay for. While
consistent with the Chilean legislation, these
assumptions would seem at odds with the
extensive stratication and heated current
discussion over the implementation of more
binding mechanisms to stop selection, such
as centralized lotteries.
Finally, Bravo, Mukhopadhyay, and Todd
(2010) present an evaluation of the attain-
ment and labor market—as opposed to test
score—effects of the Chilean voucher regime
based on estimation of a structural model of
dynamic school and labor-market choices.
They use the 2002 and 2004 waves of a large
survey of working-aged individuals that
includes educational and work history, and
54 This is a highly restrictive way of measuring school
quality/value added, relative to other recent efforts in
the literature, e.g., Abdulkadiroglu, Angrist, and Pathak
(2014), Kane and Staiger (2008), and Chetty, Friedman,
and Rockoff (2014a, 2014b).
family background characteristics. The iden-
tication originates in that individuals vary
in the number of years they were exposed
to the voucher regime. The parameters of
the structural model are estimated using
about 100,000 person–year observations.
The results are then used to simulate choices
and outcomes for those fully exposed to the
voucher regime and to compare these to the
counterfactual with the voucher program
shut down.
The results suggest that the introduction
of vouchers led to an increase in earnings
from years attending municipal and subsi-
dized private primary schools, but a decrease
from years attending secondary schools.
The authors note that the secondary-school
effects might be explained by entry of less-
efcient schools induced by the voucher,
and by reduced per-student expenditures
noted in section 3. Educational attainment
is estimated to be substantially increased by
the voucher regime. For example, full expo-
sure to the regime is estimated to increase
high-school graduation by 3.6 percent, and
completion of at least two years of college
by 2.6 percent. Average lifetime earnings are
estimated to be unchanged by the voucher.
The increased attainment is offset by the
lower return to secondary education and
delayed entry into the labor force.55 Earnings
variation is reduced by the voucher, as those
at the bottom end of the earnings distribu-
tion benet from improved primary educa-
tion, while those at the top end suffer from
weakened secondary schools. The authors
nd substantial increases in average dis-
counted lifetime utility, approximately
10 percent. These gains arise from the utility
of time spent attending school and not work-
ing. Gains are found at all percentiles, with
55 Earning returns to college education are estimated to
be relatively low. The model does not allow the voucher
regime to directly affect the returns to college education.
481
Epple, Romano, and Urquiola: School Vouchers
larger increases at lower percentiles than
higher percentiles.
The paper is ambitious and the ndings of
large attainment effects, no (average) earn-
ings effects, and substantial lifetime utility
gains add new ndings on the net effects of
vouchers. Identifying the effects of vouchers
at different school levels and on different
school types is also notable, relative to other
approaches. Again, however, the structural
approach comes at a cost. First, the identi-
cation challenges inherent in a large-scale
program are still present. Second, to make
the model tractable, individuals are of “just”
three types. Related to this, the model has
limited potential to provide insight into sort-
ing effects on schooling and labor-market
outcomes. For example, might the nding
of lower earnings from attending second-
ary school be a result of sorting (with both
municipal and subsidized schools having
worse students on average in the voucher
regime), whether through composition or
peer effects?
5.4.2 Sweden
Bohlmark and Lindahl (2008) present
an analysis analogous to that of Hsieh and
Urquiola (2006)—they ask if outcomes
improved by more in municipalities that expe-
rienced more extensive independent-school
entry. They focus on three types of student
outcomes: (1) GPA for the ninth and twelfth
grades, (2) participation in higher education
(a dummy for having completed at least one
year of education within six years of leav-
ing compulsory schooling), and (3) years of
schooling eight years after compulsory edu-
cation. On outcome (1), the results point to a
small positive effect on average ninth-grade
GPA that does not persist until grade12; on
(2) and (3) there is no evidence of an effect.
Bohlmark and Lindahl (2012) extend this
analysis to several additional cohorts, nding
signicantly more positive conclusions.
Specically, measures (1)–(3) are found to
increase with the independent-school enroll-
ment share. For instance, a 10 percentage-
point increase in this share is associated with
a 0.08σ increase in language and math scores
at the end of ninth grade, and a 0.04σ rise in
the fraction of individuals completing at least
one semester of university. As stated, these
estimates are obtained using data at the
municipality level. When observations are
aggregated further the results, with respect
to test score and grade gains, are robust,
while those with respect to college atten-
dance and years of schooling are somewhat
less so.
Bohlmark and Lindahl (2012) conclude
by discussing the contrast of ndings for
Sweden relative to Hsieh and Urquiola’s
(2006) ndings for Chile. They observe
that the more favorable results for Sweden
with respect both to cream skimming and
educational outcomes are consistent with
the predictions of the reputational model
of MacLeod and Urquiola (2009). Namely,
the fact that it is harder for independent
schools to cream skim may imply that they
may seek to build reputations for quality on
value added, rather than peer composition;
parental school choice may in turn be driven
by value added as well.
However, recent work suggests a poten-
tial source of caution with respect to the
test-score-related results from Sweden.
Specically, while as stated the content of
the tests is nationally standardized, they are
graded at each school. A concern that had
been mentioned in Sweden is that indepen-
dent schools might grade more leniently;
indeed this is something that Bohlmark
and Lindahl (2008) themselves mention.
This was recently analyzed in a regrading
exercise described by Tyrefors, Hinnerich,
and Vlachos (2013). Independent graders
reexamined exams from different schools.
The authors point out that independent
schools were more likely to have their grades
Journal of Economic Literature, Vol. LV (June 2017)
482
lowered upon a second examination.56 It is
possible that the independent schools—per-
haps under greater pressure to please par-
ents—engaged in more grade ination.
Consistent with this, Wondratschek,
Edmark, and Frolich (2013) look at other
outcomes and nd modest to zero effects.
Two aspects distinguish their study. First,
rather than look at variation in school choice
at the municipality level, they use a more
nely grained measure calculated using geo-
graphical information. Specically, they sup-
pose that children who lived closer to more
(public or private) schools were impacted
more by the voucher reform, which allowed
them greater choice among nearby schools.
Second, they consider both short- and
long-term outcomes. They nd small effects
on test scores from exams graded at school
(again, grade ination may be a concern,
although Wondratschek, Edmark, and
Frolich (2013) suggest it is not a major one)
and military exam scores for men. They nd
zero effects on longer-term outcomes such as
university educational attainment, employ-
ment, criminal activity, and health.
Finally, as in Chile, one can undertake the
most aggregate analysis by simply looking
at the evolution of Sweden’s performance
in international test scores (with analogous
threats to identication). Here the picture is
distinctly more negative, as Sweden has seen
signicantly deteriorating performance in
the years since vouchers were implemented.
As in Chile, there have been calls for reform,
although the details remain under debate.
5.4.3 Canada
The aggregate effect of a voucher-like sys-
tem is also explored by Card, Dooley, and
Payne (2010) in Canada. Specically, they
consider the case of Ontario, where two
56 This brief synopsis is based on correspondence with
the authors, as the article cited is in Swedish. In addition,
see the reporting in Fisman (2014).
parallel systems coexist. First, non-Catholics
are allowed to attend public schools—for this
segment the public sector essentially oper-
ates a monopoly. Second, Catholics (who
account for roughly 40 percent of the pop-
ulation) are allowed to attend a “separate”
Catholic school system. If they do so these
schools receive the resources that would have
otherwise gone to the public sector, emu-
lating a voucher system. In short, Catholics
have greater choice and their presence can
generate competition between schools.
In particular, the competitive pressure will
be greater in a given area to the extent that:
(1) Catholics are prevalent in it, and (2)they
are willing to switch between sectors. Card,
Dooley, and Payne (2010) measure the
former simply by the population share of
Catholics; they proxy the latter by measur-
ing how willing the eligible population is to
switch between systems when a school of a
system that previously did not exist opens in
a given area. They nd that this is more likely
to be the case in areas that are growing. A key
assumption is that the fraction of Catholics
has no direct effect on average test score
gains. A further concern is that the entry of
the Catholic system into a given area may be
endogenous.
Card, Dooley, and Payne (2010) nd sta-
tistically signicant average gains in achieve-
ment (in terms of test-score improvement
between the third and sixth grades) that are
greater in more competitive markets—those
that have more Catholics and are grow-
ing faster. They describe these as modest;
for instance, markets with 60 as opposed
to 20 percent of children with choice
would have achievement higher by 0.03 to
0.05standard deviations.
5.4.4 India
Finally, as noted in section 3.1.3, the
experimental design in Andhra Pradesh
was unique and notable in that random-
ization involved not only students, but also
483
Epple, Romano, and Urquiola: School Vouchers
towns/markets. Specically, rst some towns
were selected for distribution of vouch-
ers; second, within the towns selected for
treatment, some children were randomly
selected to receive the vouchers. This allows
Muralidharan and Sundararaman (2015) to
go beyond the usual comparison (lottery win-
ners versus lottery losers) and address poten-
tial externalities on children who remain in
public school. For example, by comparing
non-applicants in towns that did not receive
vouchers to non-applicants in towns that did,
they can get a sense of negative effects on
children “left behind” in the public sector.
The authors nd little if any evidence of such
externalities. In short, the gains in perfor-
mance found in the tests for Hindi may rep-
resent true aggregate-level gains.
5.5 Question 5: What political-economy
factors determine the existence and design
of voucher programs?—We close our review
of the empirical literature by looking at evi-
dence related to political economy. Brunner,
Sonstelie, and Thayer (2001) study voting on
the 1993 California voucher proposal using
data from 3,786 precincts in Los Angeles
County. They take Nechyba (1999) as their
point of departure, noting that this model
predicts that a voucher would cause house
prices to fall (rise) in neighborhoods with
high (low) quality schools. The authors use a
hedonic equation to estimate the relationship
of housing prices and school quality—mea-
suring the latter by scores on standardized
tests—to estimate the housing price pre-
mium for schooling in each district. This
price premium is then included in a regres-
sion in which the dependent variable is the
proportion of voters supporting the voucher.
They nd strong support for predictions with
respect to housing owners. Holding other
variables constant, their estimates imply that
the difference in the vote favoring vouch-
ers between districts with a housing pre-
mium 15 percent above the average would
be 8 percent lower than in a district with a
housing price premium 15 percent below
the average.
Brunner and Sonstelie (2003) analyze a
2000 voucher proposal in California. They
study opinions on the proposal using data
from a statewide opinion poll conducted
three months prior to the vote. The poll asked
respondents whether they supported the
voucher, their perception of quality at their
local public school, and their income and
demographic characteristics. The authors
investigate responses among three groups of
home owners: those without children, those
with children in public school, and those
with children in private school. For owners
without children, support for the voucher
was inversely related to local public-school
quality. This is in keeping with a concern for
property values. Households with children in
private schools are much more supportive of
the voucher, but also at the margin have a
larger reduction in support as public-school
quality increases. For those with children in
public school, voucher support was positively
related to public-school quality, though sig-
nicant at only the 10 percent level. Brunner
and Sonstelie note that this nding might
emerge if (as Nechyba’s analysis predicts)
some households with children in public
school expect to move to a neighborhood
with low-quality public schools (low housing
prices), take up the voucher, and send their
children to private school. Households with
children already in private school who choose
nonetheless to live in a neighborhood with
a high-quality public school would presum-
ably not relocate in response to the voucher.
Hence, results that at rst glance seem con-
tradictory can in principle be reconciled.
Brunner and Imazeki (2008) extend the
analysis of California voting, arguing that
higher-income voters’ support for a voucher
depends on the extent of choice in local
education markets. In markets with many
districts, higher-income households can be
Journal of Economic Literature, Vol. LV (June 2017)
484
expected to have paid a substantial housing
price premium to locate in a high-quality dis-
trict. For such households, a private-school
voucher program could adversely affect
home values. By contrast, in low-choice
markets, high-income households would
likely benet from being able to use vouch-
ers to pay for private schooling. They use
block-level voting data from the 2000 ballot
initiative (Proposition 38) that would have
offered a at voucher of $4,000 per student
in California. They create an index for choice
among public schools and estimate regres-
sions in which the dependent variable is a
logistic transformation of the fraction of yes
votes. The regressions include income, an
interaction of income and the choice index,
and demographic controls. They predict that
income will have a positive coefcient, the
interaction of income and the choice index
will have a negative coefcient, and the sum
of the coefcients will be negative. Their pre-
dictions are supported. Moreover, the effects
are quite large and are robust to a variety of
specication checks, providing evidence that
the extent of Tiebout choice impacts voucher
support.
Brunner, Imazeki, and Ross (2010)
exploit the idea that votes on the California
voucher initiative may signal intent to use
the voucher. They nd that support for the
voucher by white households with children
increased with the proportion of nonwhite
students in their children’s schools. No com-
parable phenomenon was present either
for nonwhite households or for households
without children. They provide some evi-
dence that households may be responding to
correlates of race/ethnicity rather than race/
ethnicity per se. For example, voucher sup-
port among nonwhite households with chil-
dren increased with the share of nonwhite
students with limited English prociency.
Kenny (2005) studies support for vouch-
ers, both in referenda and state legislatures.
He identies ten referenda initiatives that,
in some way, provide support for private
schools. These include proposals for trans-
portation for private-school students, in
addition to a variety of voucher models. All
ten referenda were defeated. Kenny then
turns to a descriptive assessment of factors
that inuence whether a voucher proposal is
considered by a state legislature, a number
of which have been passed or at least sup-
ported by one chamber, and factors that raise
its probability of success. He concludes that
ideology plays a central role, noting that vir-
tually all proposals were in states where the
Republican Party controlled the legislature,
and among those, the successful proposals
were in states with relatively more conserva-
tive Republicans. Kenny also nds evidence
that voucher proposals focused on big-city
districts tend to garner more support.
Kenny (2010) investigates voting on two
voucher proposals that have come before the
US Congress. One, an amendment to the No
Child Left Behind (NCLB) Act, would have
allowed federal funds to help children in
poorly performing or unsafe public schools
attend private schools. The other was to
authorize a voucher for Washington, DC.
Only four Democrats, found by Kenny to be
highly conservative, voted for either. Hence,
Kenny again focuses on an empirical descrip-
tion of the determinants of votes by House
Republicans. He considers the degree of
conservatism of the legislator—based on
Americans for Democratic Action (ADA)
scores—and characteristics of the legislator’s
district: percent urban, percent low income,
percent black, percent teachers unionized,
and percent of private-school attendees. Of
these, the ADA score is highly signicant
in the expected direction. The variation in
percent teachers unionized is negatively
correlated with voting on the NCLB amend-
ment, but essentially uncorrelated with the
position on the Washington, DC vote. Kenny
notes this is consistent with teachers’ unions
exerting more effort to defeat the NCLB
485
Epple, Romano, and Urquiola: School Vouchers
legislation, which would have had impact
nationwide. Variation in percent black across
districts has a relatively signicant posi-
tive impact on the NCLB vote, but not on
the Washington, DC vote. This may also be
consistent with legislators voting in constitu-
ents’ interest,57 since constituents would be
directly impacted by NCLB legislation but
not by the Washington, DC voucher.
We saw in our review of theoretical
research on the political economy of vouch-
ers an emphasis on variation in preferences
for education spending across the income
distribution, and associated nancial incen-
tives for support or opposition to vouchers.
Kenny’s work suggests a greater role for ide-
ology in theoretical modeling of preferences
over vouchers. Theoretical analysis of the
political economy of vouchers also empha-
sizes the scal incentives for adoption. In
particular, that research highlights the poten-
tial for reducing school taxes if a voucher less
than per-student public-school expenditure
induces students to exit to private schools.58
For example, Brunner and Sonstelie (2003)
report that the Legislative Analyst’s Ofce
estimates overall scal cost of the voucher to
be in the range from an increase of $500mil-
lion to a decrease of $2.5 billion. Thus far,
these scal incentives have not been investi-
gated empirically, making this an interesting
open issue for research.
57 Opinion polls regarding vouchers generate consid-
erable debate about the sensitivity of responses to the
phrasing of questions. Overall, however, polls on voucher
programs tend to nd greater support among African
Americans than among the overall population. See, for
example, responses to questions (16a) through (16c) in
http://educationnext.org/les/2014ednextpoll.pdf. Eight
surveys of African Americans were conducted between
1996 and 2008 by Bositis (2008). The proportion support-
ing vouchers exceeded the proportion opposed in the eight
surveys, with a majority supporting vouchers in ve of the
eight.
58 We noted above that such scal gains arise from the
Milwaukee voucher, although peculiarities of the nancing
result in the gains accruing to Wisconsin taxpayers outside
of Milwaukee.
6. Conclusion
Vouchers have been neither the rousing
success imagined by proponents nor the
abject failure predicted by opponents. While
the evidence does not make a case for whole-
sale adoption of vouchers, recent theoretical
and empirical results suggests a need for—
and reasons for cautious optimism about—
potential gains from improving voucher
design.
In high-income countries, research on
the impact of small-scale programs on test
scores exhibits no consistent, robust pattern.
While the effects are sometimes adverse and
sometimes favorable, it is frequently the case
that no signicant impact is found. The most
robust nding is that voucher threats induce
public schools to improve. While signicant
identication challenges arise in this type
of analysis, the estimated effects seem to us
reliable and large enough to be educationally
meaningful and warrant further research. In
addition, recent evidence from small-scale
experiments in the United States nds sub-
stantial gains in years of school for recipients
who had not experienced gains in test scores.
While potentially due to differential peer
composition between public and private
schools, the effects are large by the standards
of the peer effects literature, and therefore
encouraging with respect to the impact of
vouchers. Nonetheless, the evidence is from
a few programs, and hence still too limited to
permit generalizations.
More encouraging results on the effect of
small-scale programs come from developing
countries. First, there are positive reduced-
form ndings from Colombia, although
questions remain as to whether the cen-
tral mechanisms that account for these are
really due to vouchers. Further interesting
evidence comes from India. While vouch-
ers there delivered modest test-score gains,
they did so at one-third the cost per stu-
dent of public schools and with no adverse
Journal of Economic Literature, Vol. LV (June 2017)
486
distributional effects. Such results have not
been mirrored elsewhere, but may have rel-
evance in other developing countries that,
like India, have dysfunctional public-school
systems (e.g., with severe teacher attendance
shortfalls). Other educational reforms are
also likely to be effective in such countries,
and at some level, interventions that increase
achievement might have higher priority than
innovations that merely reduce cost. Further,
the extent to which cost savings—which
come largely from lower pay to untrained
(e.g., individuals who may be high-school
graduates only), entry-level teachers—can
be sustained in the longer term and at higher
grade levels remains an open question. On
the other hand, the observed results might
improve as new teachers gain experience.
The evidence on large-scale programs
raises methodological challenges, and very
much highlights the importance of voucher
design. For instance, analysis of the rst two
decades of the Chilean voucher provided
strong evidence of cream skimming and,
at best, mixed evidence of impacts on test
scores. These adverse ndings—and discon-
tent with the education system more gener-
ally—brought and is likely to bring further
reforms. Notably, in 2008 the government
introduced greater voucher payments tar-
geting lower-income students, and prohib-
ited tuition “top-ups” (charging more than
the voucher) for these students at participat-
ing schools. Recent research suggests favor-
able effects from this reform and highlights
the desirability of further analysis in this
dimension.
Finally, the Chilean case also shows the
importance of getting voucher design right.
As stated, large protests surrounding the
school system have been prevalent over the
past years. Michelle Bachelet, who returned
to the presidency in 2014, made a salient
campaign promise to remove the ability of
private voucher schools to operate for prot.
The existing research does not speak directly
or denitively to whether this would be pro-
ductive, but aside from this, these events
highlight the desirability of getting design
elements right, before popular discontent
and political considerations direct voucher
design.
In the case of Sweden’s large-scale voucher
program, early research with respect to
effects on test scores likewise found little
effect. More recent work features evidence
of signicant gains, although there are mixed
results and concern related to grade ination
among private schools. Recent research also
tends to support the nding that voucher
competition has improved the performance
of public schools, and that the program
design has contributed to limiting cream
skimming. Like in Chile, however, there
is signicant discontent with vouchers in
Sweden, although perhaps not as widespread.
Interestingly, while Chile historically has had
low performance in international test scores,
it has been improving; in contrast, Sweden
historically had high performance and has
been declining. This juxtaposition is to
some extent arbitrary—for example, Chile’s
gains may be larger and more relevant than
Sweden’s decline. The more general point is
that aggregate educational performance is
the product of complicated processes with
signicant lags, and methodologically it is
difcult to isolate how vouchers affect it.
To summarize, the evidence does not make
a case for wholesale adoption of vouchers,
but does strongly suggest the desirability of
continued experimentation and evaluation.
Recent evidence from the United States
highlights the attendant challenges. On the
one hand, some encouraging positive evi-
dence on graduation and college attendance
has emerged along with some additional
evidence of competition-induced improve-
ments in public-school performance. On
the other hand, some discouragingly large
negative achievement effects for voucher
recipients have been found. The evidence
487
Epple, Romano, and Urquiola: School Vouchers
also suggests that work originating in a sin-
gle country or in a single research approach
is unlikely to completely answer questions
regarding vouchers. Small-scale experiments
are appealing in providing strong statisti-
cal identication, but do not always isolate
mechanisms (e.g., peer effects, differences
in expenditure per pupil) and leave open
the issue of scalability. Large-scale programs
provide scope for assessment of the effects
of vouchers in practice, but identication is
a greater challenge due to potential selec-
tion effects and associated differential peer
effects (and, sometimes, to potential con-
founding effects of contemporaneous policy
changes). For research, the ongoing tasks
include continuing renement of identi-
cation strategies, investigating longer-term
impacts, providing a better understanding
of why effects emerge or fail to emerge, and
marshaling theory and evidence to improve
voucher design.
R
Abdulkadiroglu, Atila, Joshua D. Angrist, Susan M.
Dynarski, Thomas J. Kane, and Parag A. Pathak.
2011. “Accountability and Flexibility in Public
Schools: Evidence from Boston’s Charters and Pilots.”
Quarterly Journal of Economics 126 (2): 699–748.
Abdulkadiroglu, Atila, Joshua D. Angrist, and Parag
A. Pathak. 2014. “The Elite Illusion: Achievement
Effects at Boston and New York Exam Schools.”
Econometrica 82 (1): 137–96.
Abdulkadiroglu, Atila, Parag A. Pathak, and Christo-
pher R. Walters. 2015. “School Vouchers and Student
Achievement: First-Year Evidence from the Louisi-
ana Scholarship Program.” School Effectiveness and
Inequality Initiative Working Paper 2015.06.
Adams, Mark. 2009. “Tomorrow’s Schools Today: New
Zealand’s Experiment 20 Years On.” George Mason
University Mercatus Center Working Paper 09-01.
Aedo, Cristian, and Osvaldo Larrañaga. 1994. “The
Chilean Experience.” In Social Service Delivery Sys-
tems: An Agenda for Reform, edited by Cristian Aedo
and Osvaldo Larrañaga, 13–49. Washington, DC:
Inter-American Development Bank.
Ahlin, Asa. 2003. “Does School Competition Matter?
Effects of a Large-Scale School Choice Reform on
Student Performance.” Uppsala University Depart-
ment of Economics Working Paper 2003:2.
Altonji, Joseph G., Todd E. Elder, and Christopher
R. Taber. 2005. “Selection on Observed and Unob-
served Variables: Assessing the Effectiveness of
Catholic Schools.” Journal of Political Economy 113
(1): 151–84.
Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja.
2013. “Students Today, Teachers Tomorrow: Iden-
tifying Constraints on the Provision of Education.”
Journal of Public Economics 100: 1–14.
Angrist, Joshua D., Eric Bettinger, Erik Bloom,
Elizabeth King, and Michael Kremer. 2002. “Vouch-
ers for Private Schooling in Colombia: Evidence
from a Randomized Natural Experiment.” American
Economic Review 92 (5): 1535–58.
Angrist, Joshua D. Eric Bettinger, and Michael Kremer.
2006. “Long-Term Educational Consequences of
Secondary School Vouchers: Evidence from Admin-
istrative Records in Colombia.” American Economic
Review 96 (3): 847–62.
Auguste, S., and J. Valenzuela. 2006. “Is It Just Cream
Skimming? School Vouchers in Chile.” Unpublished.
Barnard, John, Constantine E. Frangakis, Jennifer L.
Hill, and Donald B. Rubin. 2003. “Principal Strati-
cation Approach to Broken Randomized Experi-
ments: A Case Study of School Choice Vouchers in
New York City.” Journal of the American Statistical
Association 98 (462): 299–323.
Bearse, Peter, Buly A. Cardak, Gerhard Glomm, and
B. Ravikumar. 2013. “Why Do Education Vouchers
Fail at the Ballot Box?” European Journal of Political
Economy 32: 26–37.
Beleld, Clive R. 2006. “The Evidence on Education
Vouchers: An Application to the Cleveland Scholar-
ship and Tutoring Program.” Unpublished.
Besley, Timothy, and Stephen Coate. 1997. “An Eco-
nomic Model of Representative Democracy.” Quar-
terly Journal of Economics 112 (1): 85–114.
Bettinger, Eric, Michael Kremer, and Juan E. Saavedra.
2010. “Are Educational Vouchers Only Redistribu-
tive?” Economic Journal 120 (546): F204–28.
Bjorklund, Anders, Melissa A. Clark, Per-Anders Edin,
Peter Fredriksson, and Alan B. Krueger. 2006. The
Market Comes to Education in Sweden: An Evalu-
ation of Sweden’s Surprising School Reforms. New
York: Russell Sage Foundation.
Bohlmark, Anders, Helena Holmlund, and Mikael
Lindahl. 2015. “School Choice and Segregation:
Evidence from Sweden.” Institute for Evaluation of
Labour Market and Education Policy Working Paper
2015:8.
Bohlmark, Anders, and Mikael Lindahl. 2007. “The
Impact of School Choice on Pupil Achievement, Seg-
regation and Costs: Swedish Evidence.” Institute for
the Study of Labor Discussion Paper 2786.
Bohlmark, Anders, and Mikael Lindahl. 2008. “Does
School Privatization Improve Educational Achieve-
ment? Evidence from Sweden’s Voucher Reform.”
Institute for the Study of Labor Discussion Paper
3691.
Bohlmark, Anders, and Mikael Lindahl. 2012. “Inde-
pendent Schools and Long-Run Educational Out-
comes: Evidence from Sweden’s Large Scale Voucher
Reform.” CESifo Working Paper 3866.
Bositis, David A. 2008. “National Survey of African
Journal of Economic Literature, Vol. LV (June 2017)
488
American Families’ Views on Education.” http://
jointcenter.org/sites/default/files/Education_0.pdf.
Bravo, David, D. Contreras, and C. Sanhueza. 1999.
“Educational Achievement, Inequalities and Private/
Public Gap, Chile 1982–1997.” University of Chile
Department of Economics Working Paper 163.
Bravo, David, Sankar Mukhopadhyay, and Petra E.
Todd. 2010. “Effects of School Reform on Educa-
tion and Labor Market Performance: Evidence from
Chile’s Universal Voucher System.” Quantitative
Economics 1 (1): 47–95.
Brunner, Eric J., and Jennifer Imazeki. 2008. “Tiebout
Choice and Universal School Vouchers.” Journal of
Urban Economics 63 (1): 253–79.
Brunner, Eric J., Jennifer Imazeki, and Stephen L.
Ross. 2010.” Universal Vouchers and Racial and Eth-
nic Segregation.” Review of Economics and Statistics
92 (4): 912–27.
Brunner, Eric J., and Stephen L. Ross. 2010. “Is
the Median Voter Decisive? Evidence from Refer-
enda Voting Patterns.” Journal of Public Economics
94 (11–12): 898–910.
Brunner, Eric J., and Jon Sonstelie. 2003. “Homeown-
ers, Property Values, and the Political Economy of
the School Voucher.” Journal of Urban Economics 54
(2): 239–57.
Brunner, Eric J., Jon Sonstelie, and Mark Thayer. 2001.
“Capitalization and the Voucher: An Analysis of Pre-
cinct Returns from California’s Proposition 174.”
Journal of Urban Economics 50 (3): 517–36.
Card, David, Martin D. Dooley, and A. Abigail Payne.
2010. “School Competition and Efciency with Pub-
licly Funded Catholic Schools.” American Economic
Journal: Applied Economics 2 (4): 150–76.
Carnoy, Martin. 1997. “Is Privatization through Edu-
cation Vouchers Really the Answer? A Comment
on West.” World Bank Research Observer 12 (1):
105–16.
Carrell, Scott E., Bruce I. Sacerdote, and James E.
West. 2013. “From Natural Variation to Optimal
Policy? The Importance of Endogenous Peer Group
Formation.” Econometrica 81 (3): 855–82.
Chakrabarti, Rajashri. 2008. “Can Increasing Private
School Participation and Monetary Loss in a Voucher
Program Affect Public School Performance? Evi-
dence from Milwaukee.” Journal of Public Econom-
ics 92 (5–6): 1371–93.
Chakrabarti, Rajashri. 2013a. “Accountability with
Voucher Threats, Responses, and the Test-Taking
Population: Regression Discontinuity Evidence
from Florida.” Education Finance and Policy 8 (2):
121–67.
Chakrabarti, Rajashri. 2013b. “Do Vouchers Lead to
Sorting under Random Private School Selection?
Evidence from the Milwaukee Voucher Program.”
Economics of Education Review 34: 191–218.
Chakrabarti, Rajashri. 2013c. “Impact of Voucher
Design on Public School Performance: Evidence
from Florida and Milwaukee Voucher Programs.”
BE Journal of Economic Analysis and Policy 13 (1):
349–94.
Chakrabarti, Rajashri. 2013d. “Vouchers, Public School
Response, and the Role of Incentives: Evidence from
Florida.” Economic Inquiry 51 (1): 500–526.
Chan, Ping Ching Winnie, and Robert McMillan.
2009. “School Choice and Public School Perfor-
mance: Evidence from Ontario’s Tuition Tax Credit.”
Unpublished.
Chay, Kenneth Y., Patrick J. McEwan, and Miguel
Urquiola. 2005. “The Central Role of Noise in
Evaluating Interventions That Use Test Scores to
Rank Schools.” American Economic Review 95 (4):
1237–58.
Chen, Zhiqi, and Edwin G. West. 2000. “Selective
versus Universal Vouchers: Modeling Median Voter
Preferences in Education.” American Economic
Review 90 (5): 1520–34.
Chetty, Raj, John N. Friedman, and Jonah E. Rockoff.
2014a. “Measuring the Impacts of Teachers I: Eval-
uating Bias in Teacher Value-Added Estimates.”
American Economic Review 104 (9): 2593–632.
Chetty, Raj, John N. Friedman, and Jonah E. Rockoff.
2014b. “Measuring the Impacts of Teachers II:
Teacher Value-Added and Student Outcomes in
Adulthood.” American Economic Review 104 (9):
2633–79.
Chingos, Matthew M., and Paul E. Peterson. 2015.
“Experimentally Estimated Impacts of School
Vouchers on College Enrollment and Degree Attain-
ment.” Journal of Public Economics 122: 1–12.
Chubb, John E., and Terry M. Moe. 1990. Politics,
Markets, and America’s Schools. Washington, DC:
Brookings Institution Press.
Chudgar, Amita, Frank Adamson, and Martin
Carnoy. 2007. “Vouchers and Public School Perfor-
mance: A Case Study of the Milwaukee Parental
Choice Program.” http://www.epi.org/publication/
book_vouchers/.
Clark, Damon. 2009. “The Performance and Competi-
tive Effects of School Autonomy.” Journal of Political
Economy 117 (4): 745–83.
Clark, Damon. 2010. “Selective Schools and Academic
Achievement.” BE Journal of Economic Analysis and
Policy 10 (1).
Coons, John E., and Stephen D. Sugarman. 1978.
Education by Choice: The Case for Family Control.
Berkeley and Los Angeles: University of California
Press.
Costrell, Robert M. 2010. “The Fiscal Impact of the
Milwaukee Parental Choice Program: 2010–2011
Update and Policy Options.” School Choice Demon-
stration Project Milwaukee Evaluation Report 22.
Cowen, Joshua M., David J. Fleming, John F. Witte,
and Patrick J. Wolf. 2011. “Student Attainment and
the Milwaukee Parental Choice Program.” School
Choice Demonstration Project Milwaukee Evalua-
tion Report 24.
Cowen, Joshua M., David J. Fleming, John F. Witte,
Patrick J. Wolf, and Brian Kisida. 2013. “School
Vouchers and Student Attainment: Evidence from
a State-Mandated Study of Milwaukee’s Parental
Choice Program.” Policy Studies Journal 41 (1):
489
Epple, Romano, and Urquiola: School Vouchers
147–68.
Cullen, Julie Berry, Brian A. Jacob, and Steven Levitt.
2006. “The Effect of School Choice on Participants:
Evidence from Randomized Lotteries.” Economet-
rica 74 (5): 1191–230.
Duo, Esther, Pascaline Dupas, and Michael Kremer.
2011. “Peer Effects, Teacher Incentives, and the
Impact of Tracking: Evidence from a Randomized
Evaluation in Kenya.” American Economic Review
101 (5): 1739–74.
Duo, Esther, Rema Hanna, and Stephen P. Ryan.
2012. “Incentives Work: Getting Teachers to Come
to School.” American Economic Review 102 (4):
1241–78.
Eden, Benjamin. 1994. “How to Subsidize Education:
An Analysis of Voucher Systems.” Unpublished.
Elacqua, Gregory. 2006. “Enrollment Practices in
Response to Vouchers: Evidence from Chile.” http://
ncspe.tc.columbia.edu/working-papers/OP125.pdf.
Engberg, John, Dennis Epple, Jason Imbrogno, Holger
Sieg, and Ron Zimmer. 2014. “Evaluating Education
Programs That Have Lotteried Admission and Selec-
tive Attrition.” Journal of Labor Economics 32 (1):
27–63.
Epple, Dennis, David Figlio, and Richard E. Romano.
2004. “Competition between Private and Public
Schools: Testing Stratication and Pricing Pre-
dictions.” Journal of Public Economics 88 (7–8):
1215–45.
Epple, Dennis, and Richard E. Romano. 1996. “Ends
against the Middle: Determining Public Service Pro-
vision When There Are Private Alternatives.” Journal
of Public Economics 62 (3): 297–325.
Epple, Dennis, and Richard E. Romano. 1998. “Com-
petition between Private and Public Schools, Vouch-
ers, and Peer-Group Effects.” American Economic
Review 88 (1): 33–62.
Epple, Dennis, and Richard E. Romano. 2008. “Edu-
cational Vouchers and Cream Skimming.” Interna-
tional Economic Review 49 (4): 1395–435.
Epple, Dennis, and Richard E. Romano. 2011. “Peer
Effects in Education: A Survey of the Theory and
Evidence.” In Handbook of Social Economics, Vol-
ume 1B, edited by Jess Benhabib, Alberto Bisin, and
Matthew O. Jackson, 1053–163. Amsterdam and
Boston: Elsevier, North-Holland.
Epple, Dennis, and Richard E. Romano. 2012.
“Economic Modeling and Analysis of Educational
Vouchers.” Annual Review of Economics 4: 159–83.
Epple, Dennis, and Richard E. Romano. 2014. “On the
Political Economy of Educational Vouchers.” Journal
of Public Economics 120: 62–73.
Epple, Dennis, Richard E. Romano, and Sinan Sarpca.
2014. “Majority Choice of an Income Targeted Edu-
cational Voucher.” Unpublished.
Epple, Dennis, Richard E. Romano, Sinan Sarpca, and
Holger Sieg. 2013. “The U.S. Market for Higher
Education: A General Equilibrium Analysis of State
and Private Colleges and Public Funding Policies.”
National Bureau of Economic Research Working
Paper 19298.
Epple, Dennis, Richard E. Romano, and Ron Zimmer.
2017. “Charter Schools: A Survey of Research on
Their Characteristics and Effectiveness.” In Hand-
book of the Economics of Education, Volume 5,
edited by Eric A. Hanushek, Stephen J. Machin,
and Ludger Woessmann, 139–208. Amsterdam and
Oxford: Elsevier, North-Holland.
Fernandez, Raquel, and Richard Rogerson. 2003.
“School Vouchers as a Redistributive Device: An
Analysis of Three Alternative Systems.” In The Eco-
nomics of School Choice, edited by Caroline M.
Hoxby, 195–226. Chicago and London: University of
Chicago Press.
Ferreyra, Maria Marta. 2007. “Estimating the Effects
of Private School Vouchers in Multidistrict Econo-
mies.” American Economic Review 97 (3): 789–817.
Ferreyra, Maria Marta, and Pierre Jinghong Liang.
2012. “Information Asymmetry and Equilibrium
Monitoring in Education.” Journal of Public Eco-
nomics 96 (1–2): 237–54.
Figlio, David, and Cassandra M. D. Hart. 2014. “Com-
petitive Effects of Means-Tested School Vouchers.”
American Economic Journal: Applied Economics 6
(1): 133–56.
Figlio, David, Cassandra M. D. Hart, and Molly
Metzger. 2010. “Who Uses a Means-Tested Schol-
arship, and What Do They Choose?” Economics of
Education Review 29 (2): 301–17.
Figlio, David, and Krzysztof Karbownik. 2016. “Eval-
uation of Ohio’s EdChoice Scholarship Program:
Selection, Competition, and Performance Effects.”
Thomas B. Fordham Institute https://edexcellence.
net/publications/evaluation-of-ohio’s-edchoice-
scholarship-program-selection-competition-and-
performance.
Figlio, David N., and Cecilia Elena Rouse. 2006.
“Do Accountability and Voucher Threats Improve
Low-Performing Schools?” Journal of Public Eco-
nomics 90 (1–2): 239–55.
Figlio, David N., and Joe A. Stone. 2000. “Are Private
Schools Really Better?” In Research in Labor Eco-
nomics, Volume 18, edited by Solomon Polachek,
115–40. Bingley, UK: Emerald Group Publishing.
Fiske, Edward B., and Helen F. Ladd. 2000. When
Schools Compete: A Cautionary Tale. Washington,
DC: Brookings Institution Press.
Fisman, Ray. 2014. “Sweden’s School Choice Disaster.”
http://www.slate.com/articles/news_and_politics/
the_dismal_science/2014/07/sweden_school_
choice_the_country_s_disastrous_experiment_with_
milton_friedman.html.
Fleming, David J., Joshua M. Cowen, John F. Witte,
and Patrick J. Wolf. 2013. “Similar Students, Dif-
ferent Choices: Who Uses a School Voucher in an
Otherwise Similar Population of Students?” Educa-
tion and Urban Society 47 (7): 785–812.
Friedman, Milton. 1962. “The Role of Government in
Education.” In Capitalism and Freedom. Chicago
and London: University of Chicago Press.
Fuller, Howard. 2011. “Keep Intact the Mission of
Choice Program.” Milwaukee Journal Sentinel,
Journal of Economic Literature, Vol. LV (June 2017)
490
April23.
Gallego, Francisco A. 2006. “Voucher-School Com-
petition, Incentives, and Outcomes: Evidence from
Chile.” Unpublished.
Gauri, Varun. 1998. School Choice in Chile: Two
Decades of Educational Reform. Pittsburgh: Univer-
sity of Pittsburgh Press.
Gill, Brian, P. Mike Timpane, Karen E. Ross, Dominic
J. Brewer, and Kevin Booker. 2007. Rhetoric versus
Reality: What We Know and What We Need to Know
about Vouchers and Charter Schools. Santa Monica
and Pittsburgh: RAND Corporation.
Glomm, Gerhard, and B. Ravikumar. 1998. “Opting
Out of Publicly Provided Services: A Majority Voting
Result.” Social Choice and Welfare 15 (2): 187–99.
Greene, Jay P., Jonathan N. Mills, and Stuart Buck.
2010. “The Milwaukee Parental Choice Program’s
Effect on School Integration.” School Choice
Demonstration Project Milwaukee Evaluation
Report 20.
Greene, Jay P., Paul E. Peterson, and Jiangtao Du.
1996. “Effectiveness of School Choice: The Milwau-
kee Experiment.” Harvard University Education Pol-
icy and Governance Occasional Paper 97-1.
Hanushek, Eric A., Paul E. Peterson, and Ludger
Woessmann. 2012. “Achievement Growth: Inter-
national and U.S. State Trends in Student Per-
formance.” Harvard Kennedy School Program on
Education Policy and Governance Report 12-03.
Hart, Cassandra M. D. 2014. “Contexts Matter: Selec-
tion in Means-Tested School Voucher Programs.”
Educational Evaluation and Policy Analysis 36 (2):
186–206.
Hensvik, Lena. 2012. “Competition, Wages and
Teacher Sorting: Lessons Learned from a Voucher
Reform.” Economic Journal 122 (561): 799–824.
Hoxby, Caroline M. 2000. “Does Competition among
Public Schools Benet Students and Taxpayers?”
American Economic Review 90 (5): 1209–38.
Hoxby, Caroline M. 2003a. “School Choice and School
Competition: Evidence from the United States.”
Swedish Economic Policy Review 10 (2): 9–65.
Hoxby, Caroline M. 2003b. “School Choice and School
Productivity: Could School Choice Be a Tide That
Lifts All Boats?” In The Economics of School Choice,
edited by Caroline M. Hoxby, 287–342. Chicago and
London: University of Chicago Press.
Hoyt, William H., and Kangoh Lee. 1998. “Educational
Vouchers, Welfare Effects, and Voting.” Journal of
Public Economics 69 (2): 211–28.
Hsieh, Chang-Tai, and Miguel Urquiola. 2003. “When
Schools Compete, How Do They Compete? An
Assessment of Chile’s Nationwide School Voucher
Program.” National Bureau of Economic Research
Working Paper 10008.
Hsieh, Chang-Tai, and Miguel Urquiola. 2006. “The
Effects of Generalized School Choice on Achievement
and Stratication: Evidence from Chile’s Voucher
Program.” Journal of Public Economics 90 (8–9):
1477–503.
Ireland, Norman J. 1990. “The Mix of Social and
Private Provision of Goods and Services.” Journal of
Public Economics 43 (2): 201–19.
Justesen, Mogens Kamp. 2002. Learning from Europe:
The Dutch and Danish School Systems. London:
Adam Smith Institute.
Kane, Thomas J., and Douglas O. Staiger. 2008. “Esti-
mating Teacher Impacts on Student Achievement:
An Experimental Evaluation.” National Bureau of
Economic Research Working Paper 14607.
Kava, Russ. 2013. “Milwaukee and Racine Paren-
tal Choice Program.” Wisconsin Legislative Fiscal
Bureau Informational Paper 26.
Kenny, Lawrence W. 2005. “The Public Choice of Edu-
cational Choice.” Public Choice 124 (1–2): 205–22.
Kenny, Lawrence W. 2010. “The Appeal of Vouchers
for Failing Large City School Districts: Voting in
Congress on Two Very Different Voucher Proposals.”
Journal of School Choice 4 (1): 5–22.
King, Elizabeth M., Peter F. Orazem, and Darin Wohl-
gemuth. 1998. “Central Mandates and Local Incen-
tives: The Colombia Education Voucher Program.”
World Bank Development Research Group Working
Paper 6.
King, Elizabeth M., Laura Rawlings, Marybell Guti-
errez, Carlos Pardo, and Carlos Torres. 1997.
“Colombia’s Targeted Education Voucher Program:
Features, Coverage, and Participation.” World Bank
Development Research Group Working Paper 3.
Krueger, Alan B., and Pei Zhu. 2003. “Comment: Prin-
cipal Stratication Approach to Broken Randomized
Experiments: A Case Study of School Choice Vouch-
ers in New York City.” Journal of the American Sta-
tistical Association 98 (462): 314–18.
Krueger, Alan B., and Pei Zhu. 2004. “Another Look
at the New York City School Voucher Experiment.”
American Behavioral Scientist 47 (5): 658–98.
Ladd, Helen F. 2002. “School Vouchers: A Critical
View.” Journal of Economic Perspectives 16 (4):
3–24.
Ladd, Helen F., and Edward B. Fiske. 2001. “The
Uneven Playing Field of School Choice: Evidence
from New Zealand.” Journal of Policy Analysis and
Management 20 (1): 43–64.
Levin, Henry M. 1991. “The Economics of Educational
Choice.” Economics of Education Review 10 (2):
137–58.
Levin, Henry M. 2008. “Issues in Educational Pri-
vatization.” In Handbook of Research in Education
Finance and Policy, edited by Helen F. Ladd and
Edward B. Fiske, 391–401. New York: Routledge.
Levin, Jesse. 2004. “Differences in Educational Produc-
tion between Dutch Public and Religious Schools.”
Columbia University National Center for the Study
of Privatization in Education Occasional Paper 93.
Lubienski, Christopher, Jin Lee, and Liz Gordon. 2013.
“School Autonomy and Equity in the New Zealand
Education Market: A Geographic Analysis of Organi-
zational Behavior and Access.” Unpublished.
Lucas, Adrienne M., and Isaac M. Mbiti. 2012. “Access,
Sorting, and Achievement: The Short-Run Effects
of Free Primary Education in Kenya.” American
491
Epple, Romano, and Urquiola: School Vouchers
Economic Journal: Applied Economics 4 (4): 226–53.
MacLeod, W. Bentley, and Miguel Urquiola. 2009.
“Anti-lemons: School Reputation and Educational
Quality.” National Bureau of Economic Research
Working Paper 15112.
MacLeod, W. Bentley, and Miguel Urquiola. 2012.
“Anti-lemons: School Reputation, Relative Diversity,
and Educational Quality.” Institute for the Study of
Labor Discussion Paper 6805.
MacLeod, W. Bentley, and Miguel Urquiola. 2015.
“Reputation and School Competition.” American
Economic Review 105 (11): 3471–88.
Manski, Charles F. 1992. “Educational Choice (Vouch-
ers) and Social Mobility.” Economics of Education
Review 11 (4): 351–69.
Mayer, Daniel P., Paul E. Peterson, David E. Myers,
Christina Clark Tuttle, and William G. Howell. 2002.
“School Choice in New York City after Three Years:
An Evaluation of the School Choice Scholarships
Program.” Mathematica Policy Research Reference
No. 8404-045.
McEwan, Patrick J. 2001. “The Effectiveness of Pub-
lic, Catholic, and Non-religious Private Schools in
Chile’s Voucher System.” Education Economics 9 (2):
103–28.
McEwan, Patrick J. 2004. “The Potential Impact of
Vouchers.” Peabody Journal of Education 79 (3):
57–80.
McEwan, Patrick J., and Martin Carnoy. 2000. “The
Effectiveness and Efciency of Private Schools in
Chile’s Voucher System.” Educational Evaluation
and Policy Analysis 22 (3): 213–39.
McEwan, Patrick J., Miguel Urquiola, and Emiliana
Vegas. 2008. “School Choice, Stratication, and
Information on School Performance: Lessons from
Chile.” Economía 8 (2): 1–42.
McMillan, Robert. 2004. “Competition, Incentives, and
Public School Productivity.” Journal of Public Eco-
nomics 88 (9–10): 1871–92.
Mills, Jonathan N., and Patrick J. Wolf. 2016. “The
Effects of the Louisiana Scholarship Program on
Student Achievement After Two Years.” Louisiana
Scholarship Program Evaluation Report Number 1.
Mizala, Alejandra, and Pilar Romaguera. 2000. “School
Performance and Choice: The Chilean Experience.”
Journal of Human Resources 35 (2): 392–417.
Mizala, Alejandra, Pilar Romaguera, and Miguel
Urquiola. 2007. “Socioeconomic Status or Noise?
Tradeoffs in the Generation of School Quality Infor-
mation.” Journal of Development Economics 84 (1):
61–75.
Mizala, Alejandra, and Miguel Urquiola. 2013. “School
Markets: The Impact of Information Approximat-
ing Schools’ Effectiveness.” Journal of Development
Economics 103: 313–35.
Muralidharan, Karthik, and Venkatesh Sundararaman.
2011. “Teacher Performance Pay: Experimental Evi-
dence from India.” Journal of Political Economy 119
(1): 39–77.
Muralidharan, Karthik, and Venkatesh Sundararaman.
2015. “The Aggregate Effect of School Choice:
Evidence from a Two-Stage Experiment in India.”
Quarterly Journal of Economics 130 (3): 1011–66.
Neal, Derek. 2002. “How Vouchers Could Change the
Market for Education.” Journal of Economic Per-
spectives 16 (4): 25–44.
Nechyba, Thomas J. 1999. “School Finance Induced
Migration and Stratication Patterns: The Impact
of Private School Vouchers.” Journal of Public Eco-
nomic Theory 1 (1): 5–50.
Nechyba, Thomas J. 2000. “Mobility, Targeting, and
Private-School Vouchers.” American Economic
Review 90 (1): 130–46.
Nechyba, Thomas J. 2003. “Introducing School Choice
into Multidistrict Public School Systems.” In The
Economics of School Choice, edited by Caroline M.
Hoxby, 145–94. Chicago and London: University of
Chicago Press.
Neilson, Christopher. 2013. “Targeted Vouchers, Com-
petition among Schools, and the Academic Achieve-
ment of Poor Students.” Unpublished.
Patrinos, Harry Anthony. 2002. “Private Education
Provision and Public Finance: The Netherlands as a
Possible Model.” Columbia University National Cen-
ter for the Study of Privatization in Education Occa-
sional Paper 59.
Peterson, Paul E., William G. Howell, Patrick J. Wolf,
and David E. Campbell. 2003. “School Vouchers:
Results from Randomized Experiments.” In The
Economics of School Choice, edited by Caroline M.
Hoxby, 107–44. Chicago and London: University of
Chicago Press.
Peterson, Paul E., and William G. Howell. 2004a. “Ef-
ciency, Bias, and Classication Schemes: A Response
to Alan B. Krueger and Pei Zhu.” American Behav-
ioral Scientist 47 (5): 699–717.
Peterson, Paul E., and William G. Howell. 2004b.
“Voucher Research Controversy: New Looks at the
New York City Evaluation.” Education Next 4 (2):
73–78.
Plott, Charles R. 1967. “A Notion of Equilibrium and
Its Possibility under Majority Rule.” American Eco-
nomic Review 57 (4): 787–806.
Plucker, Jonathan, Patricia Muller, John Hansen, Russ
Ravert, and Matthew Makel. 2006. “Evaluation of
the Cleveland Scholarship and Tutoring Program:
Technical Report 1998–2004.” Bloomington, IN:
Center for Evaluation and Education Policy.
Pop-Eleches, Cristian, and Miguel Urquiola. 2013.
“Going to a Better School: Effects and Behavioral
Responses.” American Economic Review 103 (4):
1289–324.
Rangazas, Peter. 1995. “Vouchers and Voting: An Initial
Estimate Based on the Median Voter Model.” Public
Choice 82 (3–4): 261–79.
Rouse, Cecilia Elena. 1998. “Private School Vouchers
and Student Achievement: An Evaluation of the Mil-
waukee Parental Choice Program.” Quarterly Jour-
nal of Economics 113 (2): 553–602.
Rouse, Cecilia Elena, and Barrow, Lisa. 2009. “School
Vouchers and Student Achievement: Recent Evi-
dence and Remaining Questions.” Annual Review of
Journal of Economic Literature, Vol. LV (June 2017)
492
Economics 1: 17–42.
Rouse, Cecilia Elena, Jane Hannaway, Dan Goldhaber,
and David Figlio. 2013. “Feeling the Florida Heat?
How Low-Performing Schools Respond to Voucher
and Accountability Pressure.” American Economic
Journal: Economic Policy 5 (2): 251–81.
Sacerdote, Bruce. 2011. “Peer Effects in Education:
How Might They Work, How Big Are They and How
Much Do We Know Thus Far?” In Handbook of the
Economics of Education, Volume 3, edited by Eric
Hanushek, Stephen Machin, and Ludger Woess-
mann, 249–77. San Diego and Amsterdam: Elsevier,
North-Holland.
Sandstrom, F. Mikael, and Fredrik Bergstrom. 2005.
“School Vouchers in Practice: Competition Will Not
Hurt You.” Journal of Public Economics 89 (2–3):
351–80.
Tiebout, Charles M. 1956. “A Pure Theory of Local
Expenditures.” Journal of Political Economy 64 (5):
416–24.
Tyrefors Hinnerich, Bjorn, and Jonas Vlachos.
2013. “Systematiska skillnader mellan interna och
externa bedomningar av nationella prov—en upp-
foljningsrapport.” Appendix 5 in Olikheterna ar
for stora. Omrattning av nationella prov i grund
och gymnasieskolan 2013, Skolinspekitionen Dur
U2011/6544/GV, 2013.
Urquiola, Miguel. 2005. “Does School Choice Lead to
Sorting? Evidence from Tiebout Variation.” Ameri-
can Economic Review 95 (4): 1310–26.
Urquiola, Miguel, and Eric Verhoogen. 2009.
“Class-Size Caps, Sorting, and the Regression-
Discontinuity Design.” American Economic Review
99 (1): 179–215.
Valenzuela, Juan Pablo, Cristian Bellei, and Danae de
los Rios. 2010. “Segregacion Escolar en Chile.” In
Fin de Ciclo? Cambios en la Gobernanza del Sistema
Educativo, edited by S. Martinic and G. Elacqua,
209–29. Santiago: Unesco.
West, Edwin G. 1991. “Public Schools and Excess
Burdens.” Economics of Education Review 10 (2):
159–69.
Witte, John F. 1997. “Achievement Effects of the Mil-
waukee Voucher Program.” Unpublished.
Witte, John F., Devon Carlson, Joshua M. Cowen,
David J. Fleming, and Patrick J. Wolf. 2011. “MPCP
Longitudinal Educational Growth Study: Fourth
Year Report.” School Choice Demonstration Project
Milwaukee Evaluation Report 23.
Witte, John F., Devon Carlson, Joshua M. Cowen,
David J. Fleming, and Patrick J. Wolf. 2012. “MPCP
Longitudinal Educational Growth Study: Fifth Year
Report.” School Choice Demonstration Project Mil-
waukee Evaluation Report 29.
Witte, John F., Joshua M. Cowen, David J. Fleming,
Patrick J. Wolf, Meghan R. Condon., and Juanita
Lucas-McLean. 2010. “The MPCP Longitudinal
Educational Growth Study: Third Year Report.”
School Choice Demonstration Project Milwaukee
Evaluation Report 15.
Witte, John F., Troy D. Sterr, and Christopher A.
Thorn. 1995. “Fifth-Year Report: Milwaukee Paren-
tal Choice Program.” Unpublished.
Witte, John F., Patrick J. Wolf, Joshua M. Cowen,
David J. Fleming, and Juanita Lucas-McLean. 2008.
“MPCP Longitudinal Educational Growth Study
Baseline Report.” School Choice Demonstration
Project Milwaukee Evaluation Report 5.
Witte, John F., Patrick J. Wolf, Joshua M. Cowen,
David J. Fleming, and Juanita Lucas-McLean. 2009.
“The MPCP Longitudinal Education Growth Study:
Second Year Report.” School Choice Demonstration
Project Milwaukee Evaluation Report 10.
Wolf, Patrick J., et al. 2010a. “Evaluation of the DC
Opportunity Scholarship Program: Final Report.”
US Department of Education, National Center
for Education Evaluation and Regional Assistance
2010-4018.
Wolf, Patrick J., et al. 2010b. “Evaluation of the DC
Opportunity Scholarship Program: Final Report:
Executive Summary.” US Department of Education,
National Center for Education Evaluation and
Regional Assistance 2010-4019.
Wolf, Patrick J., Babette Gutmann, Michael Puma,
Brian Kisida, Lou Rizzo, and Nada Eissa. 2009.
“Evaluation of the DC Opportunity Scholarship
Program: Impacts after Three Years: Executive
Summary.” US Department of Education, National
Center for Education Evaluation and Regional Assis-
tance 2009-4051.
Wolf, Patrick J., John F. Witte, and David J. Fleming.
2012. “Special Education and the Milwaukee Paren-
tal Choice Program.” School Choice Demonstration
Project Milwaukee Evaluation Report 35.
Wondratschek, Verena, Karin Edmark, and Markus
Frolich. 2013. “The Short- and Long-Term Effects
of School Choice on Student Outcomes—Evidence
from a School Choice Reform in Sweden.” CESifo
Working Paper 4438.
Zimmer, Ron, and Eric P. Bettinger. 2008. “Beyond the
Rhetoric: Surveying the Evidence on Vouchers and
Tax Credits.” In Handbook of Research in Education
Finance and Policy, edited by Helen F. Ladd and
Edward B. Fiske, 447–66. New York: Routledge.