DOES PRIVATIZATION IMPROVE EDUCATION?
THE CASE OF CHILE’S NATIONAL VOUCHER PLAN
Martin Carnoy and Patrick McEwan1
The Framework for Assessing Privatization
The push to expand public funding for private education—in the form of vouchers
or charter schools—is based on several key claims derived from theories of how an
education market should function. Empirical evidence for or against these claims has
been scarce. Yet thanks to a number of recent studies in the United States and in other
countries, particularly Chile, we now know a lot more about educational markets—hence
have much of the information we need to judge whether the claims are valid. The results
of these studies suggest that, despite claims by proponents, markets in education improve
academic performance little if at all. Neither do they provide significantly better
education for the group that needs it most—low-income students. The results also suggest
that successful privately-run schools are better at attracting relatively low cost students
and at raising revenue than increasing educational productivity.
Voucher advocates have consistently claimed that because public education is a
government monopoly it inherently denies consumers free choice in their children’s
education (Friedman, 1955; 1962). This leaves them worse off than they would be under
competitive conditions, for two reasons: many consumers are less than optimally satisfied
since they would choose an educational alternative if they could, and schools are not
compelled to produce as much output as schools would (for the same cost) if they faced
1 The authors are Professor of Education and Economics at Stanford University and Assistant Professor,
School of Education, University of Illinois. This paper is based on research conducted under a grant from
the Ford Foundation in 1998-1999. We would like to thank the Foundation for their support. The views
expressed here are, however, the authors’ and should not be attributed to the Foundation.
competition. Providing vouchers to students would allegedly solve both of these
problems. Vouchers (or their cousins, privately-run charter schools) would induce private
education providers to compete for public school students, giving consumers educational
opportunities not previously available and simultaneously forcing poorly-performing
public schools to improve or lose students.
Friedman’s analysis suggested that voucher schemes would improve the
“welfare” that consumers received from education through greater choice and increased
competition among education providers. Friedman has also consistently argued that
private providers would supply the same quality schooling at a lower price than the
public sector. In addition, other voucher advocates have claimed that private schools are
also more effective, able to deliver higher educational achievement than public schools
without spending more per pupil (Chubb and Moe, 1990).
Some of these claims have been tested in small-scale settings in the United States,
but despite some positive aspects of small-scale, quasi-experimental studies, they also
have serious limitations. They tell us little about the supply response of private education
to vouchers, nothing about the potential impact of competition on the performance of
students in public schools, nothing about whether private schools would produce produce
outcomes at lower cost than public schools.2
Our premise in this paper is that we can gain insights into these issues by
examining school systems where vouchers have been implemented on a large scale, and
where private school supply has increased. Chile is probably the single best example,
2 Derek Neal concludes in his review of Catholic school effectiveness that “…we cannot confidently expect
positive outcomes for [voucher] program participants if the program is large in scale. . . .Large school
voucher programs would likely mean the expansion of many existing private schools and the entry of many
new private schools. How would this expansion and entry affect the quality of private schools or the
worldwide, to study the effects of vouchers. Influenced by Milton Friedman’s proposal,3
Chile’s military government transferred responsibility for public school management
from the national Ministry of Education to local municipalities in 1980 and began
financing public and most private schools with vouchers. 4 Each school’s revenues were
henceforth determined on a month-to-month basis by total enrollments and a
government-determined voucher. Teachers lost their status as civil servants, reverting to
municipal contracts, and school buildings and land were signed over to municipal control.
Initial transfers proceeded rapidly, encouraged by financial incentives. By 1982 most
public schools were operated by municipalities, and the number of private schools was
Education Reform in Chile
The Military Regime, 1973-1980. At the time of the military coup d’etat in 1973,
Chile’s education system was one of the most developed in Latin America. It had
achieved near-universal enrollment in primary education, a feat that still eludes much of
Latin America (Castaneda, 1992; Schiefelbein, 1991). A dirigiste Ministry of Education
assumed exclusive responsibility for administering the public schools. Even so,
important numbers of private schools operated, about half of these under the auspices of
the Catholic Church (Espinola, 1993). Following a long tradition of public support of
private education, many received partial subsidies from the national government that
covered about 30 percent of costs in 1980 (Larranaga, 1995).
quality of remaining public schools? We do not know, and available data shed little light on this question”
(1998, p. 84).
3 Most Chilean economic and social policy during the military government was deeply influenced by the
Chicago School of economics (Valdes, 1995).
Upon assuming power in 1973, the military government disbanded the teachers
union and fired teachers with leftist views (Parry, 1997a; 1997b). It also initiated a
massive administrative reorganization, dividing the country into 13 regions, and the latter
into provinces and over 300 municipalities. At each level the President appointed
governors and mayors, drawn mainly from the ranks of the military (Stewart & Ranis,
1994). During the 1970s the Ministry of Education, in addition to other ministries,
devolved some powers to Regional Ministry Secretariats (SEREMIs) which were charged
with administrative and supervisory duties formerly performed by the central ministry.
Despite the apparent move towards decentralization, the system often functioned as a
military chain of command, organized to implement central government directives (Parry,
1997a; 1997b; Stewart & Ranis, 1994). Mayors of municipalities would not be elected
democratically until 1992.
The Educational Reform, 1980-1990. In 1980 the military government initiated a
sweeping reform. It first transferred responsibility for public school management from
the Ministry of Education to local municipalities.5 Teachers lost their status as civil
servants, reverting to municipal contracts, and schools buildings and land were signed
over to municipal control.6 Initial transfers proceeded rapidly, encouraged by financial
4 The following section draws on the work of other authors who have described the Chilean reforms. See
especially Gauri (1998), Jofre (1988), and Parry (1997a; 1997b).
5 Once transferred to municipalities, schools were placed under the control of one of two kinds of
institutions. Most opted to manage their schools with a Departmento de Administración de la Educación
Municipal (DAEM). DAEMs exist under the larger umbrella of the municipal bureaucracy and, as such,
are governed by municipal rules. Corporations are non-profit organizations that are not subject to direct
mayoral control, though the mayor does preside over a governing board. Their operations are generally
subject to fewer regulations. In contrast to DAEMs, the corporation head is not required to be a teacher and
corporation employees are not subjected to municipal rules regarding the hiring and remuneration of
6 Gauri (1998) and Parry (1997a) describe how property and equipment were leased at no cost to
municipalities for 99 years.
incentives, and by 1982, around 84 percent of schools were operated by municipalities.7
The process was interrupted by economic crisis in 1982 when the central government was
unable to cover the costs of transfers, though all schools were transferred by 1987 (Jofre,
As schools were transferred to municipalities, public teachers were offered severance pay
and became municipal rather than national employees (Castañeda, 1992). Instead of
conforming to the national Escala Única de Remuneraciones, their wages and working
conditions were henceforth governed by the more flexible Código de Trabajo (the latter
was already applicable to most Chilean workers). Teachers lost guarantees of job
security, the right to salary during vacations, standard wage scales, a 30 hour work-week,
and the right to collectively bargain (Rojas, 1998). Teachers in private schools also lost
some legal protections, including minimum wage guarantees and a system of annual
Coupled with decentralization, the government drastically altered how public and most
private schools were financed. Prior to 1980, as in much of Latin America, school
budgets were largely determined by the need to sustain an existing plant of teachers and
facilities. If budgets adjusted in response to the level of student enrollments, they only did
so at a sluggish pace. Many private schools were already being subsidized by the
government before 1980, meeting the rest of their costs with tuition payments and Church
support—about 14 percent of students attended mainly Catholic, subsidized schools and
another 6 percent, high-cost, non-subsidized private schools. Under the reform, the
Ministry of Education began disbursing monthly payments to municipalities based on a
7 Municipalities received an overhead grant of three to five percent on total municipal wages and salaries as
an inducement to begin administering schools (Parry, 1997a; Winkler & Rounds, 1996).
fixed voucher multiplied by the number of students enrolled in their schools; private
schools received equivalent per-student payments if they did not charge tuition. Thus,
payments to public or private schools began fluctuating in direct proportion to student
The law established a base voucher level, which varies according the level of
education and the location of the school.8 Though the real value of the voucher was
originally intended to keep pace with inflation, it was de-indexed following the economic
crisis of the early 1980s. Over the course of the 1980s, as copper prices fell, the real
value of the per-pupil voucher declined precipitously, reaching its lowest point in 1988
(Figure 1). It rebounded thereafter with improved economic growth and has continued to
The voucher plan precipitated a massive redistribution of enrollment across
private and public schools. At the beginning of decade, around 15 percent of students
were enrolled in private voucher schools, and almost 80 percent in public schools. By
1996 around 34 percent of enrollments were in private voucher schools. This growth
occurred mostly at the expense of public schools (see Figure 2). Throughout this period,
between five and nine percent of students private enrolled in elite private schools that
Return to Democracy, 1990. The military ceded power to a democratic government in
1990. The form and function of Chile’s voucher system were largely maintained,
although new policies were grafted onto the existing system. The government focused on
8 Chilean law specifies a factor by which the base voucher is adjusted for students at every grade level.
Furthermore, selected municipalities receive ad-hoc “zone assignments” to compensate for high poverty or
isolation. Since 1987, rural schools within municipalities have received upward adjustments. For details,
see Parry (1997a).
improving the quality of poor primary schools through direct resource investments. The
“900 Schools Program,” referred to as P-900, was targeted at high-poverty and low-
achieving schools (Garcia-Huidobro, 1994). Classrooms received a package of basic
teaching materials and infrastructure improvements, while teachers received additional
in-service training. Funds were also provided train and employ local secondary graduates
as tutors for the lowest achieving students. Eventually, P-900 expanded to include about
2,300 schools. In 1992, the Program to Improve the Quality and Equity of Pre-primary
and Primary Education (MECE) was initiated with World Bank financing. More
ambitious in scope than P-900, it sought to endow all publicly-funded schools with
textbooks, libraries, and some infrastructure improvements (Cox, 1997).
About 43 percent of Chilean children now attend private primary (K-8) schools,
all but 8 percent of those, private voucher schools. Private voucher schools are both
religious (almost all Catholic) and, approximately two thirds, for profit. For profit schools
account for much of the 23 percentage point growth of private enrollment after 1980.
Fifty-seven percent of primary pupils still attend municipally run public schools also
largely financed by vouchers. This figure is considerably down from the 80 percent who
attended public primary schools in 1980. However, the proportion of pupils attending the
different types of schools have remained relatively stable since 1990, when Chile
replaced the military dictatorship that implemented the voucher plan with a
democratically elected government. It was in 1990 that the Chilean government began a
concerted effort to improve all low performing schools, both public and private voucher
The return to democracy in 1990 brought renewed political pressures from
teachers seeking improved wages and working conditions. Negotiation between the
government and teachers resulted in the passage of the 1991 Estatuto Docente, a national
law that subjected the teacher labor market—particularly for public school teachers—to
additional regulation (Rojas, 1998). Wage floors were set for teachers with various levels
of experience and training; these minimum wages were legislated to vary in lockstep with
the voucher’s value. Limits on hiring and firing of public teachers were also introduced.
Public school teachers could either be hired as tenured or contracted teachers.9 Tenured
teachers were to be hired through public contests in each municipality and severe
restrictions were placed on their firing or reassignment. Contracted teachers had fewer
restrictions placed on their hiring and firing, but could account for no more than 20
percent of a municipality’s teacher workforce. The contracts of private teachers were still
governed by the Código de Trabajo, which permitted significantly more flexibility in
hiring and firing.
Assessing Voucher Plans
The Chilean voucher plan was supposed to improve Chilean education and make
Chilean parents feel better off because they had a wide choice of where to send their
children to school, including the option of private education. Before assessing Chile’s
voucher plan, we need to be clear on what vouchers are supposed to have accomplished.
Choice. Voucher proponents offer the empirical evidence that parents who send
their children to voucher or charter schools are “more satisfied” than when their children
were in their local public school. But who takes advantage of choice and what the factors
9 Chilean Spanish makes the distinction between titulares and contratados.
influencing choice (revealed preference) are also important to understanding consumer
“satisfaction.” Voucher and charter plans do not only provide increased choice for
consumers. Many privately run schools also have a choice of students. The most
desirable schools from a consumers’ point of view may be those that have the longest
waiting lists, hence the greatest opportunity to exercise choice. These are the schools that
are able to deliver the greatest peer effect, one of the few elements of educational quality
that parents can recognize.10 If the level of student performance is crucial to schools’
attractiveness to parents, vouchers and charter schools will have enormous incentive to be
selective, no matter what constraints, such as lotteries, are placed on them. Does this
affect overall consumer satisfaction?
Competition. A fundamental benefit of privatization is the alleged gain in
effectiveness and efficiency in public education resulting from competition with
alternative providers of educational services. Testing for the effect of competition has
been elusive. Private schools are more likely to locate in areas with more educated
parents, who have more income and whose children are, on average, easier to bring to
high levels of achievement. Private schools may also be more likely to locate in areas
where public schools have lower than “expected” student performance. The potential
positive effect of competition (a higher density of private schools) on academic
10 The current array of public school choice arrangements in the United States and other countries provides
ample evidence for the fact that parents who place high value on education generally seek more over less
selective situations, giving schools (even some public schools) considerable options to be selective. The
most general form of selectivity is real estate prices and real estate taxes, determining who lives in one
school district or another. Magnet (selective) schools within a district have served to provide a different
kind of public choice, again based on choice by parents and schools. New York City has long had
specialized, selective public high schools (entry by test or audition only), such as Bronx Science,
Stuyvesant, and Music and Art (now called Laguardia). In almost all other countries, parents often seek to
gain entry into “better” public schools outside their neighborhoods. These generally have waiting lists, and
exert choice over incoming students through examinations, interviews, and willingness by the parents to
donate financially to the school.
achievement in public schools can be confounded the simultaneity of private schools
responses to lower public school scores or higher parent income. A third problem is that
it is difficult to separate positive competition effects from negative effects on public
school student performance due to “cream-skimming.” education. mayof a higher
concentration of private schools for these reasons. In the United States, the issue is
further complicated by the dominance of religious education in private alternatives. Even
theoretically, competition from religious schools should not necessarily induce better
academic performance from public schools if they perceived such private competitors as
catering to a different market niche, namely parents seeking a religious environment.
Academic Efficiency. Empirical evidence of improved student outcomes under
voucher and charter plans—specifically student performance in privately run schools—is
important to the argument that they make consumers better off. The gain for low-income
students is especially crucial given current privatization politics. Do students in private
schools do better academically than in public? The main obstacle to showing whether this
is true or not is selection bias. Ideally, we would like to know whether a student placed
randomly in a privately or publicly managed school spending the same amount of
resources per student would do better academically. Experimental data are best suited for
such estimates, and some are available, but even they have selection bias problems.
Further, we need to understand whether there exists a logic of greater private school
efficiency, as implicitly claimed by some analysts (for example, Chubb and Moe, 1990;
Bryk, Lee, and Holland, 1993), and, if so, the source of such efficiency.
Equity Effects. To the extent that a society values educational equity, we want to
know whether vouchers and charters would alter the present distribution of access to
educational quality. The theoretical stream of the voucher movement puts its stock in “the
rising tide lifts all boats” argument, claiming that educational efficiency gains would be
large enough to raise consumer welfare among all groups. For political reasons, most
voucher/charter proponents now argue that the greatest need for vouchers and charter
schools is among low-income groups, for whom choice is most limited. Even with
limited vouchers and charters, they could produce greater inequality of access to
educational quality. Yet society may accept greater inequality as the price of improving
academic performance for some, if that gain is large enough. The question, then, is the
size of the inequality effect and the size of the potential gain for the winners.
Lessons from Chile
The Chilean data provide important insights into a privatized choice reality.
Chile’s reforms encouraged a rapid growth in private school enrollment in the 1980s that
was driven by an expansion of non-religious and profit maximizing voucher schools. Our
estimates using extensive data on test scores in the 1980s and 1990s show that, on
average, this type of privately-run school is marginally less effective than municipal
schools in producing Spanish and mathematics achievement in the fourth grade (or, at
best, similarly effective). Further results suggest that non-religious private voucher
schools are even less effective than municipal schools when they enroll lower-SES pupils
or are located outside of the capital. Some evidence suggested that the gap is explained
by different resources in private schools, such as a greater percentage of teachers with
short-term contracts (McEwan and Carnoy, 1999a)..
Although they produce somewhat lower test scores, non-religious private schools also
cost about 13-17 percent less than public schools once outputs and student background
are held constant (the gap may be slightly larger in Santiago). Direct evidence suggests
that a rather small portion of the difference is due to some exogenous constraints on
resource allocation such as the number of teacher contract hours and class size. Indirect
evidence suggests that the difference is probably attributable to other
constraintsexternally imposed on municipal schools by the regulatory
environmentsuch as lower private sector wages and increased private sector flexibility
in managing infrastructure investments. Our evidence cannot rule out the possibility that
different regulations on public schools (such as less restrictive teacher legislation) could
alter relative efficiency, quite independently of a voucher plan.
In contrast to non-religious voucher schools, Catholic schools are more effective than
public schools at producing achievement for similar students. Catholic schools spend
more per pupil than public schools (they charge higher tuition, on average, and have more
expensive, largely donated facilities), but because they apparently “add” more academic
value, are about as cost-effective in the production of achievement.
These results are inconsistent with advocates’ claims that, on the whole, privately
managed voucher schools produce significantly higher achievement than public schools
for pupils with similar socioeconomic backgrounds. Nevertheless, a large category of
private schools is more cost-efficient than publicly run schools. Another category,
Catholic voucher schools, is able to achieve higher test scores for similar students but
only by spending more. Such results deliver a mixed message, suggesting that more
money may be needed to produce higher student achievement even in private schools, but
that private schooling (or deregulation) may save money. Although it would be difficult
to argue for a strategy that reduces costs per student at the expense of student
achievement, poor countries with limited resources may find vouchers to be attractive.
At the very least, cost savings from voucher programs could be re-directed to more
traditional efforts at compensatory education for low-achieving students.11 However, it
would be naïve to assume that a struggling democracy couldor shouldimplement a
voucher plan as swiftly and decisively as Chile. Even if most parents suppport choice
and privatization, as they probably did in Chile, reality in most democratic countries
would make Chilean-style implementation difficult. After all, the Chilean plan was
implemented by an authoritarian regime that systematically and violently squelched
opposition, including forcibly dismantling teachers’ organizations.
We were also able to estimate choice functions for families in Santiago, Chile’s
capital. Results using a simple model with no interaction variables suggest that families
of higher socioeconomic status are more likely to choose private schools, but that all
families derive similar utility from school attributes such as higher test scores and the
average schooling of other parents. Including interaction effects, however, we found that
parents with more schooling derive the greater utility from increasing amounts of these
attributes. Our results therefore reject the hypothesis that less-educated parents respond to
the offer of higher performing, higher social class schools to the same degree as do more-
educated parents, even when these schools are available in equal numbers and even when
their cost is approximately the same.
This should not be interpreted as meaning that less-educated parents are
“irrational.” Wells and Crain (1992) argue that school choice is governed not only by
11 Chile’s P-900 program, implemented after the return to democracy, was just such as attempt.
resource availability, but also by access to information and internalized viewpoints
associated with social status. Lower income parents may not have full information
concerning school quality because such information may be costly to obtain or interpret,
a sentiment echoed by Levin (1991). Even with relevant information, members of lower
social class groups could be “either intimidated by, distrustful of, or resistant to members
of the dominant group and therefore [would] remove themselves from competition for
seats in the ‘best’ schools” (Wells and Crain, 1991, pp. 77-78). Lower income parents
might not be as likely to choose higher-performing schools with higher social status
student bodies even when their children might qualify for these schools or they could
afford to pay the somewhat higher costs associated with them. Their self-perception as
not “belonging” in these better public or private schools could explain why the utility
functions apparently vary by parents’ education. Less-educated parents’ perceptions of
their position in the social structure may prevent them from making choices to send their
children to higher status schools.
Schools with educated parents may reinforce these perceptions by dissuading less-
educated parents from placing their children there. This would be rational behavior on the
part of schools if they believed that bringing in lower-status children could affect the
school’s desirability to other parents, especially those with more education. Indeed, our
results tend to bear out the rationality of such selectivity by profit-maximizing schools.
We also find that lower class size is not an important factor in influencing the
attractiveness of a school to more educated parents. If this is the case, as our results
appear to indicate, school administrators can also lower costs by raising class size as long
as they simultaneously maintain a high average parent-education clientele. The reverse
side of this coin is that although urban municipal schools tend to have about the same
class size as private voucher schools and do better on the national tests once we adjust
their score for the socio-economic background of the pupils, this is not enough for them
to be attractive to parents with higher education. Such parents pay more attention to
school status (as measured by average parents’ education and average test score
unadjusted for socio-economic background) than “value-added.”
Choice studies using U.S. data find similar tendencies. What distinguishes ours
from those is the availability of detailed data in Chile on families and schools linked by a
household survey instrument that actually identifies the school attended by each child in
the family. Our results are based on actual choices in an education market that gives
families a variety of choices. If less educated parents in Chile have “revealed” utility
functions regarding school choice that differ from those of higher educated parents, it is
likely that the same would be true in the United States were school choice widened
beyond residential movement by voucher plans. The main counter-argument to this
assertion is that at higher average national income per capita, even less-educated parents
have such high education that the differences we observe in Chile would disappear. But
many sociologists would contend that behavior based on social position is independent of
national income and education level.
There exists another argument that might lessen the significance of our results:
different utility functions are immaterial if the resulting choice patterns do not widen the
earnings or other welfare differences between the children of more and less-educated
parents. The fact that less-educated parents choose lower status schools and higher-
educated parents choose higher status schools may not mean much if these same schools
raise their children’s academic achievement equally or more than higher status schools.
But lower-income children may be penalized in terms of both peer effect and attainment
by attending schools of lower socioeconomic status, hence significant potential losses in
welfare. .Higher status schools have, on average, higher scoring students and build
“cultural capital” that raises students’ academic expectations for a given level of
academic achievement (Cookson and Persell, 1985). This also suggests that higher
educated parents are not irrational either. They make not get higher value added from
sending their children to private schools (and paying tuition besides), but they probably
do get a significant peer effect on their child’s achievement and a potentially large
educational attainment effect.
Another interesting aspect of these results is what they tell us about how private
schools operate. Catholic schools in Chile—most pre-dating the 1981 voucher reform—
are able to achieve higher test scores than public schools, even when we account for pupil
social class differences, and do so by spending more per pupil. For profit schools, which
sprang up like mushrooms in Chilean cities when vouchers were made available, are no
better at producing high achievement scores for a child of a given socioeconomic
background than public schools. They are better at attracting higher scoring students.
They do that by locating, on average, in higher income neighborhoods and tending to be
selective (Parry, 1996). They are also better at reducing costs per pupil than public
schools (public schools in Chile are more constrained in their hiring practices), and do
that by paying their personnel less. 12 This implies that the line of least resistance for
12 To some extent, private schools in urban areas, particularly in Santiago, are able to “free ride” on the
public education system, which provides the main employment draw for individuals into the teaching
profession and an employment and pay “base.”
private for profit voucher schools in competition with public schools is to attract higher
scoring students by establishing themselves as somewhat higher socially than the public
schools in their area, then holding down costs. This can be interpreted as a positive effect
of markets, but it hardly fulfills the promise of raising a community’s pupil performance.
We also used panel data of test results and the density of private schooling at the
municipal level from the early 1980s to the mid-1990s to estimate the possible effect that
increased competition from private education might have had on public school
performance and overall educational performance (McEwan and Carnoy, 1999b).
We found that the effect of competition is positive in the Metropolitan Region,
though modestly so, accounting for roughly 0.2 of a standard deviation increase in test
scores over 15 years. These overall gains appear to be driven by public schools in which
parents have lower-middle or middle levels of educational attainment. Outside the
Metropolitan Region, where three-quarters of Chile’s primary students live, competition
has slightly negative effects. Evidence suggests that the P-900 program had positive
effects of 0.1-0.2 standard deviations on Spanish and mathematics achievement. These
findings are consistent with an extensive literature in developing countries that shows
positive effects of basic resource investments in primary schools (Fuller & Clarke, 1994).
Our results neither refute nor provide strong support for the notion that
competition will lead to improvements in the quality of public schools. Instead, they
suggest that effects of competition may exist in some contexts, but not in others (in this
respect, they resemble the rather mixed findings of U.S. research, which has relied on
data from a variety of states, time periods, and institutional contexts. Existing theory and
evidence are unable to provide much guidance. If economists are to provide credible
advice to those who design and implement voucher plans, we require a fuller
understanding of the conditions under which competition may or may not produce
We often presume that vouchers can be implemented just as described in planning
memoranda. However, the Chilean experience demonstrates that many stakeholders will
seek to alter, often with great success, the form and function of voucher policies. This
modifies the incentives and constraints faced by public school managers and, ultimately,
the effects that vouchers will have on student outcomes. A cogent lesson from Chile is
that an economic understanding of vouchers and competition cannot divorce itself from
the larger political economy of school choice.
It is worth noting that between 1980 and 1990, the Chilean voucher plan was
“unusual” in that it was implemented in a dictatorship. In many respects, its provisions
during that first decade met many of the conditions favored by voucher advocates (e.g.,
the abolition of teachers’ unions and unregulated expansion of for-profit private
schooling). If the effect of competition on public education was attenuated in Chile’s
political context, it seems likely that it would also be attenuated in politically democratic
societies, where political opposition to vouchers would flourish.
We observed in Chile that competition is only one of several potential effects that
vouchers may have on schools and students. Vouchers encourage a large-scale sorting of
students across public and private schools. This certainly occurred in Chile, and some
empirical evidence suggests that it assumed the form of “cream-skimming”, in which
able or privileged students were the first to exit public schools (Gauri, 1998; Parry,
1996). If peer effects are important, then it is probable that the exit of these students
negatively affected the outcomes of students remaining in public schools. We don’t have
empirical evidence on the existence or magnitude of such an effect. However, it is clear
that a full evaluation of the Chilean reform—or any voucher plan, for that matter—should
weigh the beneficial effects of competition against the potentially harmful effects of
A curious feature of the voucher literature is that advocates and critics have a
tendency to emphasize one of these effects, and ignore or downplay the others. For
example, advocates have emphasized the positive effects of competition (e.g., Hoxby,
1998), while skeptics have focused on the negative effects of sorting and cream-
skimming (e.g., Fuller, Elmore, & Orfield, 1996). To adequately evaluate the impact of
vouchers, we need to consider both.
The Chilean experience with a nationwide voucher plan suggests that “marketizing”
education will increase school choice for a certain fraction of parents, but is unlikely to
improve educational delivery for more than a small fraction of the school population.
Chile’s experience also suggests that vouchers increase inequality in the school system,
mainly through peer effects. The Chilean results are generally consistent with much
smaller voucher experiments and other choice plans in the U.S. (Levin, 1998).
Such results should not be surprising. For those who promulgated the Chilean
reform, and for most of the architects of voucher plans in the U.S. and elsewhere, the
main motivation for privatizing education is a profound belief that a public education
monopoly restricts individual choice. For them, expanding choice, in and of itself,
improves public welfare—even if it also produces greater inequality.
The main reasons that increased choice seems to lead to greater inequity are that
“better” privately-run alternatives to public schools are more likely to locate in areas
where they can attract “lower cost” students, and that many parents do not realize their
first choice of schools. If schools operate for profit, or even as non-profit private
organizations, it is almost impossible to prevent them from selecting where to locate and
from selecting students. Even in the allegedly strictly controlled Milwaukee experiment,
the private schools involved managed to turn away students with “special problems.”
One way to reduce the disequalizing side effects of vouchers is to target them. By
limiting vouchers to low-income families, as they are now in Milwaukee, Cleveland, and
other cities, it is more likely that some high quality private for –profit or non-profit
providers would enter the inner city market. But to achieve that goal, voucher values will
have to be as high or perhaps higher than current public school costs per student,
especially if the number of vouchers is larger than existing Catholic school excess
capacity.13 The original Friedman idea that private schools would deliver education equal
in quality to current public education at one-half the price has never been realized in
practice. Private providers will demand at least as much as per student costs in local
public schools and will avoid taking special education students. Witness the rise in
voucher values in Milwaukee, and Cleveland’s Hope Schools (the only for-profit
provider in the Cleveland voucher plan) shifting to much higher “vouchers” associated
with charter status. The political question then becomes whether legislators will pass
voucher plans that provide large vouchers to low-income families but not to middle class
This still does not solve the problem of improving the quality of education even
for those low-income students able to get into private schools. Our data from Chile
suggest that even in the best of cases, fifteen years of intense competition improved
achievement in public schools by only a small amount. “Value added” in U.S. private
schools is by the most optimistic accounts only slightly higher than in public schools.
Studies comparing voucher or private schools with public schools find no differences or
only small differences in achievement. So privatization solves neither the gap in
achievement between low-income children and higher-income children nor the gap in
access to high quality schools. With vouchers, the vast majority of low-income children
still get less than adequate education, even though some will switch schools. In Chile, the
measure that most effectively addressed the quality of education problem in low
performing public schools was not increased competition from privately-run schools, but
effective Ministry of Education intervention in building capacity—new curriculum
materials and training teachers to use them.
13 Catholic schools will accept smaller vouchers because they are currently operating at less than capacity.
Vouchers can also be used to subsidize current Catholic school students. Once capacity is filled, however, it
is unlikely that the parish or archdiocese will open more schools at the same low price.
Table 1. Characteristics of primary students, teachers, and schools
Female (%) 48.7 48.2 57.2 49.4 46.2 50.5
Mother’s schoolinga 7.49 8.97 10.70 9.95 9.28 14.20
(4.80) (2.77) (3.45) (3.17) (2.98) (2.26)
Father’s schoolinga 7.68 9.44 11.18 10.53 9.64 15.35
(5.01) (2.92) (3.56) (2.93) (3.06) (2.36)
Monthly household incomeb 1.65 2.29 3.02 2.61 2.88 11.17
(2.34) (2.23) (3.19) (1.96) (3.11) (31.01)
N 16,707 2,740 2,622 227 3,125 1,159
Female (%) 71.9 77.9 75.6 69.8 71.8 73.5
College graduate (%) 97.8 97.0 96.5 94.7 91.9 96.9
Age (mean) 46.1 46.1 40.7 38.5 39.7 39.1
(8.4) (8.6) (10.0) (9.1) (9.8) (9.8)
Contractors (%) 9.3 12.3 10.4 17.9 19.2 14.6
Moonlighting (%) 10.6 20.1 21.8 21.5 30.7 15.9
N 35,683 14,804 7,495 1,132 15,511 10,377
Class sizec 22.8 28.6 38.7 32.3 28.0 21.8
(8.9) (10.3) (5.9) (8.7) (9.6) (8.8)
Teacher contract hours per class 38.1 37.9 40.7 38.3 34.0 45.1
(8.8) (7.9) (8.5) (8.3) (8.4) (17.4)
Enrollment in grades 1-8 215 414 576 334 286 292
(293) (390) (282) (245) (368) (258)
N 3,823 972 337 87 1,455 499
aMeans of these variables exclude observations for children whose mothers or fathers are absent from the household.
bVariable divided by 100,000.
cClass size is calculated as total primary enrollments in each school divided by the number of primary classes.
Source: Student data are from the 1994 CASEN household survey, Ministry of Planning. Teacher data are from the
1996 teacher census, Ministry of Education. School data are from Ministry of Education enrollment files, 1996.
Student data are from a sample, while other data have census coverage.
Note: Standard deviations for continuous variables are in parentheses. Student observations are weighted in order to
account for unequal probabilities of selection into the CASEN sample. Thus, the distribution of student observations
across school types does not reflect the population distribution.
Table 2. Fourth-grade achievement differences between Public DAEM and other school types
Dependent variable: SPANISH Dependent variable: MATH
1990 1992 1994 1996 Mean
1990 1992 1994 1996 Mean
Public corporation 0.27 0.22 0.11 0.05 0.16 0.22 0.21 0.10 n.s. 0.13
Catholic voucher 1.11 1.09 1.07 0.99 1.07 1.02 0.98 0.91 0.87 0.94
Protestant voucher 0.40 0.31 0.47 0.39 0.39 0.37 0.22 0.36 0.33 0.32
Non-religious voucher 0.48 0.40 0.40 0.34 0.40 0.44 0.33 0.33 0.27 0.35
Private non-voucher 1.93 1.89 1.90 1.61 1.83 1.94 1.75 1.80 1.56 1.76
Adjusted for SES, location:
Public corporation -0.04 -0.06 -0.08 -0.08 -0.06 -0.04 -0.03 -0.06 -0.09 -0.06
Catholic voucher 0.31 0.23 0.25 0.27 0.27 0.28 0.19 0.17 0.24 0.22
Protestant voucher -0.17 -0.21 -0.01 -0.16 -0.14 -0.18 -0.27 -0.09 -0.15 -0.17
Non-religious voucher -0.05 -0.10 -0.07 -0.07 -0.07 -0.04 -0.10 -0.08 -0.08 -0.07
Private non-voucher 0.63 0.61 0.66 0.38 0.57 0.67 0.58 0.65 0.40 0.57
Adjusted for SES, location,
Public corporation --- --- --- -0.08 --- --- --- --- -0.08 ---
Catholic voucher --- --- --- 0.26 --- --- --- --- 0.22 ---
Protestant voucher --- --- --- -0.09 --- --- --- --- -0.08 ---
Non-religious voucher --- --- --- 0.06 --- --- --- --- 0.06 ---
Private non-voucher --- --- --- 0.42 --- --- --- --- 0.44 ---
n.s. indicates not statistically significant at 5 percent.
Note: Unadjusted differences are from achievement regressions in which school type dummies are the
only independent variables (omitted category: Public DAEM). Adjusted differences are from regressions
that control for additional independent variables
Table 3. Mean annual per-student costs, divided by category (1996 pesos)
National voucher payments 185,882 163,084 165,499 182,083 158,848 ---
Municipal contributions 23,834 36,258 --- --- --- ---
Parent contributions 164,224 193,794 252,312 211,965 232,363 731,125
Imputed rent on land and buildings 52,244 44,257 74,425 59,911 --- ---
P-900 program 4,133 2,605 265 1,579 1,904 ---
Total 430,316 439,998 492,501 455,538 393,115 731,125
(120,913) (99,291) (97,937) (115,373) (96,596) (120,318)
N 1,278 386 125 21 766 369
Notes: Standard deviations are in parentheses.
Table 4. Cost differences between Public DAEM and other school types, 1996
Difference adjusted for:
SES, location SES, location,
Public corporation n.s. n.s. n.s.
Catholic voucher n.s. n.s. 5.0%
Protestant voucher -9.3% n.s. n.s.
Non-religious voucher -14.9% -13.2% -11.2%
Private non-voucher 13.8% 14.0% 12.2%
n.s. indicates not statistically significant at 5 percent.
Table 5. The effects of private enrollment share on fourth-grade mathematics achievement, by Region and
Dependent variable: MATH
Metropolitan Region Other Regions
Percent of parents in school with at least some
Percent of parents in school with at least
sample ≥0 and
sample ≥0 and
PERPRIV 0.222 -0.459 0.469 0.369 -0.554 -0.219 -0.171 -0.112 -0.507
(0.133) (0.380) (0.191) (0.327) (0.504) (0.091) (0.207) (0.130) (0.172)
PERPRIV2 -0.171 0.678 -0.496 -0.203 0.297 0.229 -0.223 0.300 0.711
(0.125) (0.462) (0.213) (0.283) (0.381) (0.169) (0.384) (0.249) (0.293)
P900 1.086 1.675 0.698 1.508 8.220 2.239 2.632 1.956 1.959
(0.672) (1.379) (0.869) (1.866) (7.879) (0.322) (0.509) (0.496) (0.708)
NATIONAL 2.219 7.391 -5.389 5.715 8.738 0.621 -8.034 4.532 3.595
(2.203) (11.506) (4.478) (3.131) (6.136) (2.170) (6.290) (1.116) (4.009)
SES1 0.829 1.546 -0.035 27.477 --- 0.502 0.180 1.245 -8.242
(1.949) (2.559) (2.956) (13.650) (0.588) (0.738) (1.048) (3.695)
SES3 -0.684 -7.622 -1.176 1.037 -2.073 -0.302 1.028 -0.729 -0.169
(0.581) (2.122) (1.191) (0.691) (1.326) (0.354) (0.963) (0.682) (0.439)
SES4 -0.143 --- --- --- -2.014 -0.350 -18.654 --- -1.484
(1.351) (1.805) (1.515) (1.440) (2.381)
N 2299 406 1054 618 221 9404 3443 3550 1831
R2 0.077 0.094 0.090 0.101 0.160 0.052 0.035 0.068 0.103
Note: All regressions use the difference-in-difference specification and include year dummies. Huber-White standard
errors are in parentheses. Coefficients and standard errors for PERPRIV2 are multiplied by 100.
PERPRIV = percent private enrollment in municipality.
Figure 2: Evolution of private and public enrollment shares, 1981-1996
Source: Vargas (1997)
81 82 83 84 8 5 86 87 88 89 9 0 91 92 93 94 95 96
Priv ate vouch er
Bryk, A. S., Lee, V. E., & Holland, P. B. (1993). Catholic schools and the common good.
Cambridge, MA: Harvard University Press.
Castaneda, T. (1992). Combating poverty: Innovative social reforms in Chile during the 1980s. San
Francisco: ICS Press.
Chubb, J. E., & Moe, T. M. (1990). Politics, markets, and America’s schools. Washington, DC:
The Brookings Institution.
Cookson, P. & Persell, C. H. (1985). Preparing for power: America's elite boarding schools. New
York: Basic Books.
Cox, C. (1997). La reforma de la educacion chilena: Contexto, contenidos, implementacion, Programa de
Promocion de la Reforma Educativa en America Latina (PREAL).
Espinola, V. (1993). The educational reform of the military regime in Chile: The system’s
response to competition, choice, and market relations. Unpublished Ph.D. dissertation,
University of Wales, United Kingdom.
Friedman, M. (1955). The role of government in education. In R. A. Solo (Ed.), Economics and
the public interest (pp. 123-144). New Brunswick, NJ: Rutgers University Press.
Friedman, M. (1962). Capitalism and freedom. Chicago: University of Chicago Press.
Fuller, B., Elmore, R. F., & Orfield, G. (Eds.). (1996). Who chooses, who loses? Culture, institutions, and
the unequal effects of school choice. New York: Teachers College Press.
Garcia-Huidobro, J. E. (1994). Positive discrimination in education: Its justification and a Chilean example.
International Review of Education, 40(3-5), 209-221.
Gauri, V. (1998). School choice in Chile: Two decades of educational reform. Pittsburgh:
University of Pittsburgh Press.
Jofre, G. (1988). El sistema de subvenciones en educacion: La experiencia chilena. Estudios
Larranaga, O. (1995). Descentralizacion de la educacion en Chile: Una evaluacion economica.
Estudios Publicos(60), 243-286.
Levin, H. M. (1991). The economics of educational choice. Economics of Education Review, 10(2), 137-
Levin, H. M. (1998). Educational vouchers: Effectiveness, choice, and costs. Journal of Policy
Analysis and Management, 17(3), 373-391.
McEwan, P. J., & Carnoy, M. (1999). Competition, decentralization, and public school quality:
Longitudinal evidence from Chile's voucher system, unpublished manuscript, Stanford
Neal, D. (1998). What have we learned about the benefits of private schooling? Federal Reserve
Bank of New York Policy Review, 4(1), 79-86.
Parry, T. R. (1996). Will pursuit of higher quality sacrifice equal opportunity in education? An
analysis of the education voucher system in Santiago. Social Science Quarterly, 77(4),
Parry, T. R. (1997a). Achieving balance in decentralization: A case study of education
decentralization in Chile. World Development, 25(2), 211-225.
Parry, T. R. (1997b). Decentralization and privatization: Education policy in Chile. Journal of
Public Policy, 17(1), 107-133.
Parry, T. R. (1997c). Theory meets reality in the education voucher debate: Some evidence from
Chile. Education Economics, 5(3), 307-331.
Rojas, P. (1998). Remuneraciones de los profesores en Chile. Estudios Publicos,(71), 121-175.
Schiefelbein, E. (1991). Restructuring education through economic competition: The case of Chile. Journal
of Educational Administration, 29(4), 17-29.
Stewart, F., & Ranis, G. (1994). Decentralization in Chile, Human Development Report
Occasional Papers No. 14,
Valdes, J. G. (1995). Pinochet's economists: The Chicago School in Chile. Cambridge:
Cambridge University Press.
Wells, A. S., & Crain, R. L. (1992). Do parents choose school quality or school status? A sociological
theory of free market education. In P. W. Cookson (Ed.), The Choice Controversy (pp. 65-81).
Newbury Park, Calif.: Corwin Press.
Winkler, D. R., & Rounds, T. (1996). Municipal and private sector response to decentralization
and school choice. Economics of Education Review, 15(4), 365-376.