Bringing the schools back in: the stratification of educational achievement in the chilean voucher system

International Journal of Educational Development j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / i j e d u d e v Bringing the schools back in: the stratification of educational achievementin the Chilean voucher system§ a Center for Applied Economics, Department of Industrial Engineering, Universidad de Chile, Santiago, Chileb Department of Sociology, New York University, United States This paper analyzes the socioeconomic stratification of achievement in the Chilean voucher system using a census of 4th and 8th graders, a multilevel methodology, and accounting for unobserved selectivity into school sector. Findings indicate that the association between the school’s aggregate family Socioeconomic stratificationMultilevel methodology socioeconomic status (SES) and test scores is much greater in the private-voucher sector than in thepublic one, resulting in marked socioeconomic stratification of test scores in the Chilean voucher system.
We also find that the amount of tuition fees paid by parents in private-voucher schools has no bearing ontest scores, after controlling for the socioeconomic makeup of the school. Implications of these findingsfor educational inequality in the context of a universal voucher system are discussed.
ß 2010 Elsevier Ltd. All rights reserved.
Evaluation of these small-scale, short-term voucher experiments leaves unanswered, however, the important question about the Among the diverse policies to improve the quality of schooling, general equilibrium outcomes of a universal voucher system educational vouchers are one of the most debated. Proponents highlight that, by expanding educational choice and stimulating Chile provides a unique case to explore this question. In 1981, in the competition among schools, vouchers will provide alternatives to context of a market transformation of the Chilean economy, a low-resource families trapped in underperforming public schools universal voucher mechanism was implemented. In the new system, the government grants a per-student subsidy to all public and private schools provided that they do not charge tuition; and all that voucher schools will skim off students with higher perform- families are allowed to use their voucher in the school of their choice ance and more socioeconomic resources, furthering segregation without improving overall educational outcomes ( led to a massive reallocation of students from the public to the newly established private-voucher sector. Public sector enrollment evaluation of experimental voucher programs in the US is not dropped from 78% in 1981 to 53% of the total enrollment in 2002 conclusive. Evaluation of programs such as the Milwaukee Parental Several studies have examined the educational outcomes of the Chilean voucher system. Lacking experimental evidence, report a range of estimated effects from no researchers have used observational data, concentrating on two improvement to small gains, with effects sensitive to sample questions: (1) Do voucher schools yield higher educational decisions, and varying across students’ gender and race.
achievement than public ones, net of the characteristics of theirstudent bodies? and (2) has the competition in local educationalmarkets promoted by voucher schools improved educational outcomes? Virtually all studies of the Chilean voucher system We are grateful to the SIMCE office at Chile’s Ministry of Education for providing focus on the differences across school sector – public vis-a-vis us with the data. Carolina Ostoic and Marcelo Henrı´quez provided excellentresearch assistance. Mizala acknowledges funding from FONDECYT project N8 private-voucher – implicitly assuming that the variance in 1061224 and PIA-CONICYT project CIE-05.
achievement between sectors is more relevant than the variance * Corresponding author at: Center for Advanced Research in Education, across schools within sector, thereby inadvertently neglecting the Universidad de Chile, Chile. Fax: +562 9784011.
school as a unit of analysis. The reason is understandable as, as we will document, there are significant differences in educational achievement across sectors. However, if substantial variation in 0738-0593/$ – see front matter ß 2010 Elsevier Ltd. All rights reserved.
doi: A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 achievement between schools exists, students’ outcomes may be 2. Stratification and achievement in the Chilean voucher more closely related to the characteristics of the school they attend than to whether the school is private-voucher or public. No studyto date examines differences in attainment between and within Beginning in the early 1980s, far-reaching reforms were schools across educational sector in the Chilean voucher system, implemented in the Chilean educational system by an authoritari- an regime that came to power in 1973. The reforms involved the This paper attempts to fill this gap. Using a multilevel decentralization of the public school system and the handing over formulation, and controlling for unobserved selectivity in the of school administration to local governments (municipalities).
allocation of students into school sectors, we examine the The most important component of the reform was a new financing socioeconomic distribution of achievement within and between mechanism for public and private schools through a nationwide schools in the public and private-voucher schools. We find that per-student subsidy, which allowed families to select the school, the Chilean voucher system has given rise to a particular form of stratification. Contrary to a simplified vision of sorting, in which Before the reform, three types of schools existed in Chile: public voucher schools homogeneously skim-off the ‘‘best’’ public schools (accounting for 80% of the enrollment), private subsidized school students, we find that, as a sector, voucher schools serve (14%) and private fee-paying schools (6%). Both public and private- a broad cross-section of the population, but each individual subsidized schools were free and funded by the government. The voucher school is characterized by high homogeneity in the latter type of school received a lump-sum subsidy, substantially socioeconomic status (SES) of its student body. This configura- smaller than the per-student spending in the public sector. Most of tion, we suggest, is contingent on the institutional design of the them were Catholic and operated as a form of charity ).
Chilean voucher system. Until recently, the Chilean voucher was Fee-paying private schools charged high tuition fees and served the flat, i.e. it did not vary with family socioeconomic resources; and Chilean elite. The 1981 reform sparked the emergence of a new voucher schools were allowed to select students at will. This sector, which we will call ‘‘private-voucher’’ to distinguish it from configuration, we argue, provides the incentives and the means the private-subsidized institutions that existed before. In the new for private-voucher schools to specialize in different market system, a per-pupil subsidy is paid by the government to all niches. We then address the question about socioeconomic schools – public or private – participating in the voucher system. In distribution of achievement between and within schools in the contrast to US experiences, in which the subsidy is given directly to private-voucher and public sectors. We find that the association the family, in the Chilean design funds are allocated directly to the between individual SES and test scores is slightly stronger in the school selected by the family based on the number of students private-voucher than in the public sector—signaling a slightly enrolled, a system known as ‘‘funds follow the student’’ ( more unequal distribution of achievement in the former sector. In It is important to note that a given private- contrast, the association between the school’s aggregate family SES voucher school receives the same per-pupil voucher payment as a of the student body and achievement is more than twice as strong in municipal school of similar characteristics. Public schools can the private-voucher sector, resulting in pronounced socioeco- receive subsidies from municipalities, with the amount transferred nomic stratification. In other words, the educational achievement varying according to the financial capacity of the municipality.
of a child attending the private-voucher sector depends consid- As a result of the voucher reform, a substantial migration from erably more on the aggregate SES of her school than on her own the public sector to this new type of school ensued. By 2002 private-voucher schools reached 38% of the enrollment, at the A final piece of this analysis examines the influence of a expense of the public sector, which dropped to 53%, by 2004 financing reform introduced in the Chilean voucher system in private-voucher enrollment had reached 41%. Students who 1993. This reform allowed private-voucher primary and high migrated to the private-voucher sector were, on average, of higher schools (and public high schools) to charge add-on fees to parents socioeconomic status than those who remained in the public to complement the government voucher. While supporters of the sector, suggesting that sorting followed the voucher reform parental tuition fees argue that they contribute needed funds to education, critics warn that they may exacerbate educational Private-paid schools were conspicuously unaffected by this stratification. If private-voucher schools use add-on fees to select transformation. Their fees were, on average, five times the per- economically advantaged families, and use tuition resources to student voucher. As a result these schools did not enter the offer education of better quality, add-on fees may explain the competitive educational market created by the reform. They strong association between aggregate family SES at the school level remained serving a small group of high-income families and do not and achievement in the private-voucher sector. We analyze the constitute a reachable alternative for the large majority of influence of tuition funds levied on parents on educational Chileans. For this reason, we do not include this type of school outcomes in private-voucher schools. We find virtually no in our analysis, concentrating instead on the public and private- association between parental add-on fees and test scores after voucher sectors that serve more than 90% of the Chilean the school-level SES is accounted for. In other words, the economic resources contributed by families to voucher schools reflect the Public schools are everywhere in the country; however, the ability to pay of the student body, but they do not appear to distribution of private-voucher schools is uneven throughout the country. In 10% of municipalities, private-voucher enrollment Our paper proceeds as follows. The next section describes the stands at more than 50% but nearly 63 out of 345 municipalities, Chilean voucher system, evaluates extant research on the Chilean mostly rural and poor, have no private-voucher schools at all.
case, and introduces the question about the socioeconomic Voucher schools are allowed to operate as for-profit institutions, stratification of achievement within and between schools. Section and about 70% of them do so. In terms of religious differentiation, three introduces the data, variables, and methods. Section four 35% of them are religious, mostly Catholic, institutions ( presents the multilevel analysis of the social distribution of achievement between and within schools in the public andprivate-voucher sectors and of the role of add-on tuition in the 1 As a robustness check we reproduce all analyses including students in private private-voucher system. The final section concludes and discusses non-voucher schools. The results for private-voucher and public schools are nearly A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 Since 1990, after the reestablishment of democracy, the Chilean These institutional features of the Chilean voucher system government has devoted substantial resources to improving the likely promote socioeconomic stratification. The literature indi- quality and equity of educational outcomes, and has implemented cates that low-SES students have, on average, lower educational targeted programs focused on the poorest, lowest-performing performance and are usually more demanding in terms of voucher system – ability to choose and competition between As a result, a flat voucher provides strong incentives for schools – have, however, remained intact for the last quarter- private-voucher schools to select socioeconomically advantaged students to lower their costs, while regulations about studentselection allow them to do so. Furthermore, rigid regulations in the 2.1. Institutional arrangements and educational stratification in the public sector prevent dismissing low-performing educators and school principals, providing incentives for high-resource, motivat-ed families to search for alternatives in the voucher sector. The The notion of ‘‘a’’ voucher system is misguided insofar as ‘‘shared financing’’ system may contribute to stratification by institutional arrangements shape the outcomes of the specific providing an additional avenue for voucher schools to select students based on their socioeconomic resources and preventing ). Four institutional features are relevant in the Chilean case: access of low-income students to private-voucher schools that the amount of the per-student voucher, rules about admission and charge fees. Even though by law schools that charge fees must expulsion of students, teachers’ regulations, and alternative provide scholarships for low-income students, the law requires only between 5% and 10% of the amount of the fees charged to be Since its inception, the Chilean voucher has provided a flat per- student subsidy without adjustments for students’ socioeconomic In 2008, a law was passed that implemented two important changes to the Chilean voucher system. The law established an schools can establish their own admission and expulsion policies, extra per-student subsidy for economically disadvantaged stu- whereas public schools have to accept all applicants unless they dents (as determined by the Ministry of Education), and for schools are oversubscribed and constitute, effectively, suppliers of last with a high concentration of disadvantaged students. This change resort. Evidence shows that private-voucher schools intensively emerges from the recognition that it is more expensive to educate use selection mechanisms such as entry exams and parental low-resource students and it effectively implies transforming the interviews to shape their student bodies (). A survey of flat voucher system into a means-tested one. In addition, the law 4th grade parents found that 44% of voucher schools give prohibited the use of parental interviews and admission tests to admission exams, and 36% request parental interviews, indicating select students among participating schools. Although it is too that many, but not all voucher schools select their students ( early to examine the consequences of these recently implemented changes, we discuss their potential effect for the stratification of Thirdly, there are differences across school sector in terms of educational achievement in light of our findings.
teachers’ contracts and regulations. Public school teachers aregoverned by special legislation (the Teacher Statute), involving 2.2. Extant research on the Chilean voucher system centralized collective-bargaining, with wages based on uniformpay-scales independent of merit, making it nearly impossible to Evaluations of the Chilean voucher system have focused on two dismiss under-performing educators. Private-voucher schools, in issues: the relative effectiveness of private-voucher vis-a`-vis contrast, operate as private firms with flexible criteria for public schools, and the effect of school competition on student personnel recruitment, dismissal and promotion.
academic outcomes. Lacking randomized designs, researchers Finally, public and private-voucher schools differ in the ability have addressed the first question by comparing the achievement of to raise additional funds. A 1993 reform allowed primary and students who attend public and private schools with controls for secondary private-voucher schools (but only secondary public their observed and (more tentatively) unobserved characteristics.
schools) to charge ‘‘add-on’’ fees to parents to supplement the Given that achievement data was available at the school but not government voucher, under a withdrawal schedule that reduces the individual level until 1997, early studies of relative effective- the subsidy as parental fees increase. This system – known as ness across school sector used aggregate school averages. ‘‘shared financing’’ (financiamiento compartido in Spanish) – concluded that voucher schools did not perform expanded rapidly from 16% of the voucher sector enrollment in better than public schools given similar resources. 1993 to about 80% in 1998, stabilizing thereafter.Private-voucher found that when sufficient control variables are schools differ in the amount of fees they charge. In 2002, 20% of added, there are no consistent differences in achievement between them were free, 44% charged less than nine dollars per month, 29% the public and private-voucher sectors. Moreover, charged between 9 and 17 dollars, and the remaining 27% charged found that public schools have advantages in educating students between 17 and 68 dollars (the government subsidy is fully from disadvantaged family backgrounds.
withdrawn at 68 dollars). Furthermore, the supply of fee-charging Availability of individual-level data since 1997 induced a new voucher schools varies across the country. Based on the Ministry of generation of studies which include controls for students’ Education’s school directory, 2% of municipalities have only fee- resources and attempt to account for selection into different paying voucher schools, 57% have both free and fee-paying, and school sectors. Most studies using individual-level data found that 41% have only free voucher schools. Supporters of the ‘‘shared students attending voucher schools have slightly higher educa- financing’’ system claim that it brings badly needed resources to tional outcomes (about 0.15–0.2 standard deviations in test scores) education, allows targeting public resources to the poorest schools, than those from public schools, net of individual attributes and promotes parental involvement critics worry that it furthers socioeconomic sorting in an already unequal system students and a novel identification strategy, found that private-voucher education leads to small (4–6% of one standard deviation in test scores), sometimes not statistically significant differences in Even though public high schools are allowed to charge parental fees, very few A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 The second line of research has attempted to identify the effect achievement substantially varies across school sector. Scores are of competition between schools on students’ achievement.
lowest in the public sector – the average of 235 is almost a half standard deviation lower than in the private-voucher sector (254) that private-voucher schools skim-off more advantaged families – while private fee-paying schools serving a small number of elite while relegating disadvantaged ones to the public sector, and that families average scores of 298 places them at a far distance from the net aggregate effect of competition on student performance is both public and private-voucher institutions.
A less-explored dimension of socioeconomic stratification is found that greater competition significantly that which occurs across schools within each sector. As a preliminary raises test scores, although the endogenous entry of voucher examination of the role of schools as units of stratification in the schools into local markets is a lingering concern.
Chilean voucher system, we partition the total variance in family In sum, the most recent estimated effects of private-voucher SES into its between-school and within-school components. When education on academic achievement are much lower than those we consider the total population of 4th graders, the proportion of obtained by the previous literature on Chile. The influence of SES variation that occurs between schools is extremely large in competition on students’ achievement remains very much an open Chile, reaching 62%. This indicates that the school is a pivotal unit of stratification. However, when we examine the variance withinand between schools across school sectors, substantial differences 2.3. Socioeconomic stratification across school sector in Chile emerge. First, the amount of variance between schools substan-tially drops—an expected finding insofar as sector organizes With this background information, we now provide introduc- socioeconomic inequality in the Chilean educational system. In tory information about differences in economic status and addition, substantial differences across sectors emerge. The SES educational achievement across school sector. presents variance that is between-schools is only 24% in the public sector the distribution of school sector by household SES decile for but it reaches 47% in the private-voucher one. In other words, Chilean 4th graders. SES combines standardized measures of while the voucher sector serves a diverse population, voucher mother’s years of schooling, father’s years of schooling and total schools are socioeconomically homogeneous—some of them family income to provide a comprehensive description of family appear to concentrate better-off families, while others focus on resources. shows the profound socioeconomic stratifica- poor communities. This descriptive evidence qualifies the claim tion in the Chilean educational system. Private fee-paying schools that private-voucher schools uniformly skim off more advantaged serve the upper class, with 94% of enrollment coming from the two students, and suggests a more complex configuration in which wealthiest deciles. Public schools mostly serve the lower and the private-voucher schools specialize in distinct niches of the market lower-middle class, with two-thirds of their students coming from in order to accomplish their diverse economic and educational the bottom half of the SES distribution. Private-voucher schools recruit broadly from the middle and upper-middle strata. There is, This evidence introduces a central question of our study: what however, substantial socioeconomic overlap between the public is the association between individual-level and school-level and private-voucher sectors—both sectors recruit about two-thirds socioeconomic resources and students’ achievement across school of their students from the middle six SES deciles.
sector? While much research explores the association between The second panel in compares educational achievement individual-level SES and achievement, the aggregate level of SES across sector. The metric is math and language test scores in a resources in the school may strongly shape test scores, contribut- national standardized test (Sistema de Medicion de la Calidad de la ing to the socioeconomic stratification of achievement. The Educacion SIMCE, in Spanish) administered by the Ministry of association between school-level SES and achievement is de- Education to 4th graders in 2002. The test scores are standardized scribed as a contextual or compositional effect, to highlight the fact to have a mean of 250 and a standard deviation of 50. As expected that it emerges from the socioeconomic makeup of the school given the different socioeconomic makeup of their student bodies, body, net of the influence of individual socioeconomic resources.
An important US-based literature has explored contextual effects of SES on educational achievement and its variation across Table 1Enrollment in school sector by family SES decile (percent distribution) and test school sectors. This literature is mostly concerned with the scores across school sector. 4th graders, Chile difference between Catholic and secular public schools. Early workby Coleman found that Catholic schools have a higher mean and a more equitable distribution of achievement within schools ). Subsequent analyzes support this result, which suggests that the ‘‘Catholic advantage’’ is accounted for by aspects of the normative environment and academic organization such as a better disciplinary climate and ). Other studies qualify this finding, indicating that differences The literature comparing Catholic and public schools in the US has also found that the association between aggregate school-levelSES and achievement is relatively similar in Catholic and public We hypothesize that, in contrast to the US case, the association between school-level SES and test scores may be stronger in the private-voucher than in the public sector in Chile, resulting in an Source: Authors’ calculations, based on the SIMCE standardized test and SIMCE overall stronger stratification of achievement in the former sector. We parental questionnaire, 4th grade students, 2002.
base this hypothesis on the institutional characteristics of the Family SES obtained from a factor analysis of mother’s years of schooling, father’s years of schooling and total family income.
Chilean voucher system. The ‘‘shared financing’’ system allows A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 private-voucher schools to extract resources from families, nomic resources. School-level SES is obtained by averaging potentially inducing a strong association between mean family individual-level SES within school. Given that the SIMCE is a SES and students’ outcomes. The more flexible regulations in the census rather than a sample of schools, this variable provides a private-voucher sector may enhance these schools’ capacity to very precise measure.We add school-level controls based on the translate the economic advantage of the families they serve into educational production function literature achievement. For example, private-voucher schools serving ). They include urban/rural location of the school, teachers’ wealthier families may be able to attract better teachers than years of experience, student-teacher ratio, school size (natural log schools serving more deprived populations, successfully capital- of the number of students enrolled in the school), the standard izing on the resources of their student body. Importantly, the deviation of family SES within school as a measure of diversity in ability and incentives of private-voucher schools to select their the socioeconomic resources of the student body, and a dummy for student body may also result in a strong contextual effect of SES. If religious private-voucher schools. Unfortunately, no variables private-voucher schools recruit students based on attributes capturing schools’ normative environment or organizational correlated with SES such as ability, cultural capital, or motivation, practices exist to date in the data. present their selection of students may result in a stronger contextual descriptive statistics at the individual and the school levels across effect of SES, driven by these attributes.
In general, we expect a closer association between the socioeconomic composition of the students’ body and achievement in the private-voucher sector than in the public one insofar asinstitutional regulations leave ample room for sorting and impose The analysis is based on a two-level hierarchical linear model less redistributive constraints on private-voucher schools. In what (HLM). The first-level units are students (within-school model), follows we examine whether schools are important units of and each student’s outcome is represented as a function of a set of stratification in the Chilean voucher system. We test the individual characteristics. In the second level (school-level model) hypothesis that the contextual effect of SES is more pronounced the regression coefficients in the level-1 model are treated as in the private-voucher sector, and examine whether this is outcome variables hypothesized to depend on specific school accounted for by the amount of parental add-on funds charged characteristics. The HLM methodology explicitly recognizes the and other school-level characteristics.
clustering of students within schools and allows simultaneousconsideration of the association between school factors and average school achievement; the relationships between individualcharacteristics and outcomes, and the variation across schools in The analysis is based on merged data from three sources. The the relationships between individual characteristics and outcomes first one is the SIMCE (Sistema Nacional de Medicio´n de la Calidad de la Educacio´n—Educational Quality Measurement System), stan- dardized tests in math and language. We utilize the 4th grade But the allocation of students to school sector is not random and (2002) and 8th grade (2004) SIMCE dataset to evaluate our depends on unobserved attributes, such as motivation, ability, and hypotheses in different grades, years and subject matters. The ambition, which are correlated with educational outcomes. We dataset is compiled by the Chilean Ministry of Education and it control for unobserved selectivity into school sector by estimating includes the entire population of public and private-voucher a two-step model The first step is a choice model schools and their students (5204 schools and 196,212 students in in which the dependent variable is the type of school attended by 2002; 4888 schools and 173,907 students in 2004). The second the student. The model considers that each student has two data source is a survey of parents of the students who took the choices—to attend a private-voucher school or a public school. In SIMCE tests. This questionnaire provides information about the order to be correctly identified, the choice model must contain at socioeconomic characteristics of students, including family income least one variable that is uncorrelated with the error term of the and parents’ education. The third source of data is administrative records from the Ministry of Education, which we used to produce satisfy the exclusion restriction, we use the supply of schools of several school-level characteristics, including school sector, school different sectors in the municipality where the family lives, i.e. the enrollment, teachers’ years of experience, the religious affiliation number of public and private-voucher schools per squared- of schools, and the amount of add-on tuition charged by private- kilometer in the students’ municipality (for a similar strategy voucher schools, which were merged to the SIMCE datasets.
see ). As it is conventional, the inverse mills-ratio The dependent variables are the math and language SIMCE obtained from the choice model is added to the achievement standardized test scores. The independent variables include equations to correct for potential selectivity.
characteristics of students and schools. The central predictor atthe student level is family socioeconomic status (SES), obtained from a factor analysis of mother’s education, father’s education andfamily income, and standardized to have a mean of zero and a The analysis is organized in four steps. The first step in any HLM standard deviation of unity. In addition, we control for students’ is the decomposition of the variance in the outcome of interest into gender (female = 1), number of books at home – a proxy for its between- and within-group parts. This step estimates a fully cultural capital and the value of scholarly culture – parental unconditional ANOVA model, and allows us to compute the expectations (a dummy coded 1 if parents expect that the child will proportion of the total variance in math and language test scores attain post-secondary education). The SIMCE tests do not track that is between schools across school sector. The second step is the students over time, so it is not possible to assess school effects on within-school model. It estimates how student characteristics achievement gains. In order to control for children’s prior affect test scores within schools. We evaluate the influence of achievement, indicator variables for whether the students family SES, gender, books at home, parental expectations, attended preschool (preschool = 1), and whether they haverepeated a grade (repeat = 1) are also included.
3 This is a major advantage over alternative databases such as TIMSS and PISA, The central predictors at the school level are school sector – whose smaller sample sizes of students and, particularly, schools result in limited public and private-voucher schools – and school-level socioeco- A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 Table 2ANOVA model. Percent of total variance in SIMCE test scores between schools. 4th grade 2002 and 8th grade in 2004, Chile.
preschool attendance, repetition history, and the selectivity terms questions focus on comparing schools across sector, we need to on test scores. The third step presents the between-school model, control for all independent variables across the entire sample.
which adds school-level characteristics to the previous specifica- Grand-mean centering accomplishes this objective (e.g. tion. This step allows us to address three questions: what is the relationship between individual-level and school-level SES and variation in individual-level coefficients across school sector. We educational achievement? Do the individual and the contextual find significant variation across sector for family SES and parental effect of SES vary across school sector? And, is the contextual effect expectations (p < .001) only, and allow for such variability by of SES accounted for by school-level characteristics and resources? adding cross-level interaction terms.
The final step of our analysis evaluates the association between Each student-level characteristic is significantly related to the parental add-on tuition fees and students’ test scores in the outcome in the expected direction and results are strikingly similar private-voucher sector. It examines whether, net of individual and across grades. Boys perform better in math and worse in language, contextual effect of SES on achievement, parental fees contribute signaling a ‘‘gender division of learning’’ similar to most countries in the world (e.g. ), which increases from 4th to 8th grade.
Books at home and parental expectations display a positive correlation with achievement, with a larger influence of expecta-tions found at public than at private-voucher schools in 4th grade, 4.1. Partitioning the variance in math and language test scores and no differences across school sector in 8th grade. Havingrepeated a grade has an expected substantial negative association Calculations indicate that about 20% of the variance in students’ with test scores. Note that with the exception of gender, the test scores is between-schools, a proportion virtually identical patterns of effects are nearly identical for math and language, across test subjects and grades. Central to our question, the indicating that the results are not an artifact of a particular subject proportion of test score variance that is between-schools differs substantially across school sector. It reaches approximately 27% inthe private-voucher sector, but only 14% in the public sector, consistently across subject matter and grade. Put another way, it ismuch more common that the worst student at a ‘‘good’’ school will We now consider the association between school-level and score lower than the best student at a ‘‘bad’’ school in the public individual-level SES and test scores across sectors. The first school- sector than in the private-voucher one. This renders the school a level model (Model 2 in evaluates the gross crucial unit of stratification of achievement in the private-voucher association between the mean school SES and achievement across sector, without controls for school-level characteristics. Thesecond one (Model 3) adds school-level characteristics. We first note that adding indicators for school sector and mean SES at theschool level in Model 2 results in a large decrease in the between- display models predicting language and math school variance in the test scores. As reported by test scores for 4th graders and 8th graders, respectively. Model 1 the decline in unexplained variance in test scores is about 20% (as presents the coefficients for the individual-level model capturing measured by the decline in b0 between Models 1 and 2). When the association between students’ SES and test scores. This and the additional school characteristics are included in Model 3, only a following models account for non-random selection of students slight additional reduction is obtained (of about 5%). In brief, school into school sector by adding the inverse-mills ratio (IMR) terms sector and socioeconomic composition of the student body obtained from the choice equation, (reported in accounts for a substantial portion of the test scores variance.
for 4th and 8th grade, respectively). All independent variables are Net of sector and socioeconomic status, school resources and centered at their grand means, except for indicator variables that demographic characteristics account for little additional variation use the natural metric. As suggests, students who attend private-voucher schools differ from those who attend the public Moving to the central question of our study, we examine the sector in terms of socioeconomic status. Given that our research relationship between individual-level and school-level SES and A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 Table 3AHLM model of achievement, language and math 4th grade, 2002. Fixed effects.
No fees (omitted category)Parental fees LT $9 Notes: Standard errors in parenthesis. References dummy: urban schools, non-religious schools, male students, parents expect the student will only finish high school, no add-on fees. PV: private vouchers schools. IRM is the inverse-mills ratio obtained from the choice equation ().
test scores. We find that the positive association between student associated with a 1-unit increase in school-level SES are less than SES and test scores is slightly stronger in the private-voucher 10% in the public sector and approximately 35% in the voucher sector than the public one. In other words, a student’s achievement sector. These differences are substantial, and they signal a is less determined by his/her socioeconomic status in the public pronounced stratification of achievement in the private-voucher than in the private-voucher sector. The difference is statistically schools: ceteris paribus, students who attend a high-SES voucher significant but small in 4th grade and statistically insignificant in school will perform much better than those in low-SES voucher 8th grade. Net of students’ SES, the association between school SES and test scores is positive in both sectors but it is much stronger in The sizable socioeconomic stratification of achievement in the the private-voucher sector. Among 4th graders, a 1-unit increase in private-voucher sector may emerge from school-level resources average school SES results in an increase in test scores of five points and characteristics. To address this possibility, we include school- in the public sector, and about 20 points in the private-voucher level attributes in Model 3 of , including rurality, sector. Given that the standard deviation of SIMCE test scores is student-teacher ratio, teachers’ experience, SES standard devia- around 50; this implies an improvement of 10% of a standard tion, school size, and religion. The answer is clear: these factors do deviation in the public sector, but a high 40% in the private- not account for the large contextual effects of SES in voucher voucher sector. Among 8th graders, the comparable increases schools. The contextual effect remains almost twice as large in the A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 (although teacher–student ratio is not significant among 8th HLM model of achievement, 4th grade, 2002, random effects.
graders). Interestingly, the standard deviation of family SES within the school has a positive influence on achievement in thepublic sector—this suggests that socioeconomic diversity within the school is not detrimental, and it may even be beneficial for learning. We speculate that this may result from the advantageous effects of having a group of high-resource students in disadvan- taged schools, probably driven by the influence of these students and their families on teachers’ expectations and peer interactions.
The main finding of this analysis indicates that contextual effect of SES is much stronger in the private-voucher as in the public sector. Based on Model 3 in the contextual effect of SES in the public sector is about 20% of a standard deviation of test scores—slightly stronger that the influence of individual-level SES.
In contrast, in the private-voucher sector the contextual effect of SES reaches 40% of a standard deviation, almost twice as much as the individual-level SES. This benefit is substantial—it compares with a four-decile increase in family-level SES, or to almost 300 additional books at home. Among students attending the private- voucher sector, the aggregate socioeconomic resources of their school are much more consequential for achievement than their own family SES. This finding supports the hypothesis that voucher schools are more able than their public counterparts to ‘‘convert’’, unmodified, the socioeconomic advantages of their student bodies into achievement gains. Rather than leveling the playing field, private-voucher schools produce a distribution of educational achievement that closely mirrors the socioeconomic resources of The substantial association between school-level SES and test scores in the private-voucher sector may be accounted for by add- voucher sector as in the public one, net of school characteristics.
on tuition fees paid by parents. The strong contextual effect of SES This finding holds for both grades and both test subjects. The in the private-voucher sector may reflect the ability of schools to comparison between Model 2 (without school-level controls) and extract additional resources from better-off parents via add-on Model 3 (adding school-level controls) gauges the extent to which tuition and translate these resources into higher educational the contextual effect of SES is accounted for by school-level achievement. If this hypothesis is true we should observe that the attributes. The answer varies across sector. In the private-voucher contextual effect of SES diminishes or disappears altogether after sector, the contextual effect of SES declines only by about 20% after controlling for the amount of add-on tuition charged by the controlling for the influence of school size, rurality, student– school. Alternatively, tuition fees and school-level SES may have teacher ratio, teachers’ years of experience and religious school. In independent beneficial effects on achievement, indicating that the public sector, in contrast, the contextual effect of SES increases the resources provided by tuition add to the benefits associated by about 50%. This increase suggests that one or more school-level with the socioeconomic makeup of each school. This hypothesis variables work as suppressors of the socioeconomic stratification will result in significant achievement gains associated with higher across public schools. Step-wise regression models (not shown, tuition, without decline in the influence of school-level SES. A available from the authors upon request) indicate that the variable third alternative suggests that private-voucher schools that operating as a suppressor is the indicator for rural school. Rural charge higher tuition may be able to select higher-SES families, schools display much higher achievement than expected given but tuition fees may not have a positive influence on achievement their SES levels, so controlling for rural residency results in a net of the average socioeconomic resources of the families stronger influence of aggregate SES resources at the school level on selected by the school. If this third hypothesis is true, we should achievement in the public sector in the Chilean educational observe that the positive association between tuition fees and system. Consistently, holding socioeconomic and other character- achievement declines or disappears after controlling for school- istics of the students constant, rural schools perform better than their urban counterparts by 20% of a standard deviation, a finding To examine these alternative hypotheses, the last columns in that is consistent across the public and private-voucher sector. This (Model 4) add tuition fees to the model. We is likely accounted for by the government programs in place since measure parental add-on tuition as a set of dummies, distinguish- the mid-1990s, which provide substantial additional financial and ing four ordered categories: no tuition fees, monthly fee of less pedagogical assistance to public rural schools ( than nine dollars, 9–17 dollars, and 17–68 dollars, with ‘‘no tuition’’ as the reference category. This formulation allows us to In line with previous research on the Chilean voucher system, capture potential non-linearities in the association with test we find that religious private-voucher schools feature higher achievement than secular ones with an average advantage of 10%of a standard deviation in test scores (Schoolswith lower teacher–student ratios and more experienced teachers 4 Alternative models were estimated with a linear formulation of add-on tuition perform better, a pattern that is uniform across school sector fees. Results are substantively identical to those presented here.
A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 Table 4AHLM model of achievement, language and math 8th grade, 2004. Fixed effects.
No fees (omitted categories)Parental fees LT $9 Notes: Standard errors in parenthesis. References dummy: urban schools, non-religious schools, male students, parents expect the student will only finish high school, no add-on fees. PV: private vouchers schools. IRM is the inverse-mills ratio obtained from the choice equation ().
After controlling for tuition fees, the contextual effect of SES nomic makeup of the student body selected by each school has declines marginally for 4th graders and remains unmodified for been accounted Our finding is all the more striking if we 8th graders, indicating that parental fees do not account for the consider that, net of their socioeconomic resources, families who beneficial effect of school-level SES on test scores. This result is are willing to pay fees may be positively selected on unobservables consistent with previous research, which has reported no (if they hold education in higher value or are more motivated), differences in performance between students in private-voucher which will result in our overestimating the association between schools that charge add-on fees and those that are free parental tuition fees and achievement.
Furthermore, the association between tuition fees andtest scores, net of school-level SES is very small and statisticallyinsignificant except for one category ($9–17 monthly fees) in 4th Note that parental fees charged by private-voucher schools do not fully translate into school revenue because the amount of government is reduced as This evidence is consistent with the third hypothesis. It tuition add-on tuition fees increase. This reduction is, however, very small—it is 0% suggests that financial contributions by parents are not associated of the subsidy up to U$9 of add-on fees, 10% between U$ 9 and 17 and 25% between with gains in students’ achievement after the aggregate socioeco- A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 stronger – twice as much – in the private-voucher than in the HLM model of achievement, 8th grade, 2004, random effects.
public sector, leading to a pronounced socioeconomic stratification of achievement. In other words, for students attending private- voucher schools, their educational achievement is more closely related to the aggregate SES of their school than to their own family’s socioeconomic resources. These findings are strikingly consistent across grades (4th and 8th) and test score subject (math and language) suggesting that they identify a general attribute of voucher schools rather than idiosyncratic patterns.
One likely mechanism for the strong contextual effect of SES on test scores in the private-voucher sector is the add-on fees that these schools have been allowed to charge since the 1990s.
By imposing fees, schools can select better-off families and translate the additional tuition funds into higher educational achievement. Our findings are not consistent with this hypothe- sis. We find that the contextual influence of SES on test scores does not decline after accounting for add-on fees, and that net of school-level SES, the amount of tuition fees levied on parents is not associated with higher achievement. In sum, the financial resources contributed by parents do not appear to translate into higher test scores once socioeconomic resources at the school Why is that private-voucher schools that charge add-on fees are able to extract resources from parents if their students do not outperform free private-voucher schools, net of individual resources? One possible answer is that parents care about peer socioeconomic makeup in itself, regardless of achievement (see parents may be able to assess average school performance, but not the value added by the school. Given the strong correlation between socioeconomic status and students’ performance (e.g.
choosing a high-SES schools is a rationalstrategy to maximize their children’s achievement. Even thoughour research design does not allow us to formally test whether add-on tuition fees induce sorting across schools, our findingssuggest that the ‘‘shared financing’’ system may provide a vehicle Virtually all research on the Chilean voucher system focuses on for socioeconomic stratification across schools, which contributes differences between school sectors—in particular, the relative to the inequality in test scores without improving the overall level effectiveness of private-voucher versus public schools. In contrast, this paper addresses the socioeconomic distribution of achieve- Further exploring why the contextual effects of SES matter so ment within and between schools across school sectors. We examine much in the private-voucher sector – substantially more than in educational achievement measured by standardized math and the public sector – is also an important task for future research. A language test scores among 4th and 8th graders using a rich literature on school effects suggests diverse pathways of hierarchical linear methodology, and accounting for non-random influence: aggregate family SES at the school level may be a proxy selectivity of students into school sector.
for beneficial peer interactions, teachers’ expectations, school The basic premise of this study is that schools are important normative climates, curriculum, basic infrastructure resources, or units of educational stratification among voucher schools. Three findings emerge from the analysis in support of this premise. First, a much larger proportion of the variance in socioeconomic status is between schools in the private-voucher sector than in the public one. This pattern suggests that while the private-voucher sector Most likely, several of these dimensions are at play serves an economically diverse population, each voucher school and feedback dynamics among them exist. For example, higher- focuses on a socioeconomically homogeneous community. Given SES student bodies likely attract more motivated families and the institutional design of the Chilean voucher system – in provide an incentive for schools to select them. High concentration particular, a flat voucher, independent of students’ need, and the of more advantaged families may induce a normative environment ability of private-voucher schools to select students according to more conducive to learning. This, in turn, may increase the ability the criteria of their choice – we interpret this finding as suggesting of better-off schools to attract more capable and motivated that voucher schools use of the flexibility provided by the students and teachers, in a dynamic that widens the socioeco- educational regulations to shape their student body and manage nomic gap in achievement across schools, creating unequal their teaching staff, thereby specializing in distinct market niches to accomplish their diverse financial and educational objectives.
Disentangling the mechanisms driving the strong contextual The substantial variation in socioeconomic makeup of the effect of SES in private-voucher schools has important policy student body across private-voucher schools raises the next implications. If school-level SES affects voucher school students’ question: does school-level SES matter for test scores, net of achievement largely because of its relationship to potentially student-level resources? The answer is a clear yes. The association alterable school organizational features, resources, or practices between aggregate school-level SES and test scores is much such as curriculum, teachers’ expectations or infrastructure, then A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 socioeconomic stratification itself may not be an important issue.
none of the school-level variables currently available in the data Policies targeted to increase school resources and to reform school adequately captures organizational and normative features at structures may go a long way towards addressing the substantial the school level. Obtaining such data is, however, possible in socioeconomic achievement gap in the private-voucher sector. If, Chile given the good educational data collection infrastructure in contrast, the contextual effect of SES cannot be traced to school that exists in the country. The SIMCE test is a census of students characteristics potentially modifiable by policy, then peer effects and schools administered annually to pupils of a specified grade emerging from socioeconomic segregation itself may be a concern level with a schedule that, since 2005, gives the SIMCE test every year to 4th graders and rotates between 8th, and 10th grades. It mate factors explaining contextual effects of SES are school already includes parental, teacher, and principal questionnaires, organizational features, these factors may depend on the social to which inquiries about normative and organizational char- makeup of the students attending each school. For example, acteristics of schools can be added at minimal cost. Furthermore, educators and school officials may respond to poor students by the grade schedule of the SIMCE test can be arranged so that lowering expectations and offering less demanding curricula. In individual students can be followed over time providing this case, socioeconomic stratification may be the fundamental longitudinal information on students’ test score gains, allowing cause of the observed socioeconomic achievement gap. As researchers to capture the value added by the school. These feasible changes would go a long way to help decipher the schools respond to the demands and political influence of their different paths for the strong influence of the socioeconomic constituents, higher-resource communities may be able to composition of schools on educational achievement in Chilean successfully lobby for more resources and reform in their schools.
In such circumstance, reducing the socioeconomic stratification Finally, the institutional design characteristics of the Chilean across schools may be necessary for equalization of educational voucher system are undergoing a substantial transformation. As mentioned, a recent 2008 law establishes a means-tested voucher This task transcends the educational system and involves and forbids private-voucher schools from selecting elementary addressing residential segregation, which is pronounced in the school’s students based on entry exams and parental interviews.
Chilean context (). In the U.S., children usually These measures should alter the incentive structure facing voucher have to attend schools in the educational system where they live, schools, reducing the incentives and ability to recruit socioeco- so that ‘‘school segregation and residential segregation are nomically advantaged students. These changes could go a long way inextricably entwined’’ : 795). In the Chilean choice in reducing the socioeconomic segregation across private-voucher system, families are formally allowed to enroll their children in any schools and could weaken the influence of school-level SES on public or private-voucher school they choose, and the influence of students’ test scores. Although it is still too early to evaluate this socioeconomic segregation is less explicit but likely as powerful to hypothesis, we hope to have provided a needed missing piece for the extent that no compensation for transportation costs – understanding of socioeconomic stratification in the Chilean substantial for poor families – is provided.
universal voucher system, and a baseline to evaluate the Our analysis shows that school-level characteristics such as school size, teachers’ experience, rurality, religious schools, orparental add-on fees have a small influence on achievement afteraccounting for the socioeconomic composition of the student body, and they play almost no role in accounting for theinfluence of aggregate school-level SES on test scores. However, Table A1Summary statistics all schools and by sector. Chilean 4th graders, 2002.
A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 Table A2Summary statistics all schools and by sector. Chilean 8th graders 2004.
Table A3Logit choice model: determinants of attending private-voucher school versus public Aedo, R., 2000. La Educacio´n privada en Chile: un estudio histo´rico-analı´tico desde el periodo colonial hasta 1990. (Private Education in Chile: a Historical-analyt- ical Study from Colonial Times to 1990).RIL, Santiago.
Alexander, K., Pallas, A., 1985. School sector and cognitive performance: when is a little a little? Sociology of Education 58 (2), 115–128.
Anand, P., Mizala, A., Repetto, A., 2009. Using school scholarships to estimate the effect of private education on the academic achievement of low income students in Chile. Economics of Education Review 28, 370–381.
Auguste, S., Valenzuela, J.P., 2003. Do Students Benefit from School Competition? Evidence from Chile. University of Michigan, Mimeo.
Baker, D., Goesling, B., LeTendre, G., 2002. Socioeconomic status school quality and national economic development: a cross-national analysis of the Heyneman– N of public schools per km2 in municipality Loxley effect on mathematics and science achievement. Comparative Education Chubb, J.E., Moe, T., 1990. Politics Markets and America’s Schools. Brookings, Coleman, J.S., Hoffer, T., Kilgore, S.B., 1982. High School Achievement: public Catholic and Other Schools Compared. Basic Books, New York.
Coleman, J.S., Hoffer, T., 1987. Public and Private High Schools the Impact of Coons, J., Sugarman, S., 1978. Education by Choice: the Case for Family Control.
University of California Press, Berkeley.
Cox, C., Lemaitre, M., 1999. Market and state principles of reform in Chilean educa- tion: policies and results. In: Perry, G., Leipzinger, D. (Eds.), Chile: recent PolicyLessons and Emerging Challenges. World Bank, Washington, DC (Chapter 4).
Denton, N., 1996. The persistence of segregation: links between residential segre- gation and school segregation. Minnesota Law Review 80, 795–824.
Ducombe, W., Yinger, J., 2000. Financing higher student performance standards: the case of New York State. Economics of Education Review 19, 363–386.
Elacqua, G., Schneider, M., Buckley, J., 2006. School choice in Chile: is it class or the Logit choice model: determinants of attending private-voucher school versus public classroom. Journal of Policy Analysis and Management 25 (3).
Elacqua, G., 2006. Public, Catholic, and For-profit School Enrollment Practices in Response to Vouchers: evidence from Chile. National Center for the Study of Privatization in Education, Teachers College, Columbia University (Paper 125).
Fiske, E.B., Ladd, H.F., 2000. When School Compete: a Cautionary Tale. Brookings, Fuller, B., Clarke, P., 1994. Raising schools effects while ignoring culture, local conditions and the influence of classroom tools rules and pedagogy. Review of Educational Research 64 (1), 119–157.
Gallego, F., 2002. Competencia y resultados educativos: teorı´a y evidencia para Chile.(Competition and educational outcomes: theory and evidence for Chile).
N of public schools per km2 in municipality Cuadernos de Economı´a 39 (118), 309–352.
Gallego, F., 2006. Inter-school Competition Incentives and Outcomes: evidence from Chile Mimeo Department of Economics. MIT, USA.
Gamoran, A., Long, D., 2006. Equality of Educational Opportunity: a 40-Year Retrospective. Wisconsin Center for Education Research Working Paper Garcia-Huidobro, E., 2000. Educational policies and equity in Chile. In: Reimers, F.
(Ed.),Unequal Schools Unequal Chances. Harvard U. Press, Cambridge (Chapter 7).
Gauri, V., 1998. School Choice in Chile: two Decades of Educational Reform.
University of Pittsburgh Press, Pittsburgh, PA.
A. Mizala, F. Torche / International Journal of Educational Development 32 (2012) 132–144 Gauri, V., Vawda, A., 2003. Vouchers for Basic Education in Developing Countries. A Mizala, A., Romaguera, P., 2001. Factores explicativos de los resultados escolares en Principal-Agent Perspective. World Bank Policy Research Working Paper 3005.
la educacio´n secundaria en Chile.(Explanatory factors of students’ performance Goldhaber, D., Eide, E., 2003. Methodological thoughts on measuring the impact of at the secondary level in Chile). Trimestre Econo´mico 272, 515–549.
private sector competition on the educational market place. Educational Eval- Neal, D., 2002. How vouchers could change the market for education. Journal of uation and Policy Analysis 25 (2), 217–232.
Gonza´lez, P., Mizala, A., Romaguera, P., 2004. Vouchers, inequalities, and the Chilean Nechyba, T.J., 2000. Mobility, targeting, and private-school vouchers. American experience. Occasional Paper Series No. 94 National Center for Study of Privati- zation in Education, Teachers College, Columbia University.
Pallas, A., Entwisle, D., Alexander, K., Stluka, M.F., 1994. Ability group effects: Greene, J.P., Peterson, P., Du, J., 1998. School choice in Milwaukee: a randomized instructional, social, or institutional? Sociology of Education 67, 27–46.
experiment. In: Peterson, P., Hassel, B. (Eds.), Learning from School Choice.
Patrinos, H., 2002. A Review of Demand-Side Financing Initiatives in Education.
Brookings, Washington, DC, pp. 335–356.
Hanushek, E., 1996. A more complete picture of school resource policies. Review of Raudenbush, S., Bryk, A., 1986. A hierarchical model for studying school effects.
Educational Research 66 (3), 397–409.
Sociology of Education 59 (1), 1–17.
Hanushek, E., 1997. Assessing the effects of school resources on student perform- Raudenbush, S., Bryk, A., 1992. Hierarchical Linear Models: applications and Data ance: an update. Educational Evaluation and Policy Analysis 19 (2), 141–164.
Analysis Methods. Sage Publications, Thousand Oaks, CA.
Hauser, R., 1970. Context and consex: a cautionary tale. American Journal of Reschovsky, A., Imazeki, J., 2001. Achieving educational adequacy through school finance reform. Journal of Education Finance 26, 373–396.
Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47 Rouse, C.E., 1998. Private school vouchers and student achievement: an evaluation of the Milwaukee parental choice program. Quarterly Journal of Economics 113, Henig, J., 1994. Rethinking School Choice: the Limits of the Market Metaphor.
Princeton University Press, Princeton.
Rumberger, R., Willms, D., 1992. The impact of racial and ethnic segregation on the Heyneman, S., Loxley, W., 1983. The effect of primary-school quality on academic achievement gap in California high schools. Educational Evaluation and Policy achievement across twenty-nine high- and low-income countries. American Journal of Sociology 88 (6), 1162–1194.
Rumberger, R., Palardy, G., 2004. Multilevel models for school effectiveness re- Howell, W., Peterson, P.E., 2002. The Education Gap: vouchers and Urban Schools.
search. In: Kaplan, D. (Ed.), Handbook of Quantitative Methodology for the Brookings Institute, Washington, DC.
Social Sciences. Sage, Thousand Oaks, CA, pp. 235–258.
Hoxby, C., 2003. School choice and school productivity: could school choice be a tide Rumberger, R., Palardy, G., 2005. Does segregation still matter? The impact of that lifts all boats? In: Hoxby, C. (Ed.), The Economics of School Choice Chicago.
student composition on academic achievement in high school. Teachers College Hsieh, C., Urquiola, M., 2006. The effect of generalized school choice on achievement Sabatini, F., Caceres, F., Cerda, J., 2001. Segregacio´n residencial en las principales and stratification: evidence from Chile’s school voucher program. Journal of ciudades chilenas: tendencias de las tres u´ltimas de´cadas y posibles cursos de accio´n.(Residential Segregation in Chilean Cities: trends over the last three Krueger, A., Zhu, P., 2004. Another look at the New York city school voucher decades and implications). EURE 27, 82.
experiment. American Behavioral Scientist 47, 699–717.
Sapelli, C., Vial, B., 2002. The performance of private and public schools in the Ladd, H., 2002. School vouchers: a critical review. Journal of Economic Perspectives Chilean voucher system. Cuadernos de Economı´a 39 (118), 423–454.
Sapelli, C., Vial, B., 2005. Private vs. public voucher schools in Chile: new evidence on Lara, B., Mizala, A., Repetto, A., 2009. The effectiveness of private-voucher educa- efficiency and peer effects. Working Paper N 289 Instituto de Economı´a, tion: evidence from structural school switches. Working Paper N 263 Center for Applied Economics, University of Chile.
Simce, 2006. Simce Parental Survey of 4th Grade. Ministry of Education, Santiago, Lee, V., Chow-Hoy, T., Burkam, D., Geverdt, D., Smerdon, B., 1998. Sector differences in high school course taking: a private school or Catholic school effect? Seitzer, M., 1995. Furthering our understanding of the effects of educational Sociology of Education 71, 314–335.
programs via a slopes-as-outcomes framework. Educational Evaluation and Lee, V., Bryk, A., 1989. A multilevel model of the social distribution of high school achievement. Sociology of Education 62 (3), 172–192.
Tokman, A., 2002. Is private education better? Evidence from Chile. Working Paper Levin, H., 1998. Educational vouchers effectiveness, choice and costs. Journal of Policy Analysis and Management 17 (3), 373–392.
Torche, F., 2005. Privatization reform and inequality of educational opportunity: the Ma, X., 2008. Within-school gender gaps in reading mathematics, and science case of Chile. Sociology of Education 78, 316–343.
literacy. Comparative Education Review 52 (3), 437–460.
Valenzuela, J.P., Bellei, C., de los Rios, D., 2008. Evolucion de la segregacion socio- McEwan, P., Carnoy, M., 2000. The effectiveness and efficiency of private schools in economica de los estudiantes chilenos y su relacion con el financiamiento Chile’s voucher system. Educational Evaluation and Policy Analysis 22 (3), 213– compartido. (Evolution of Socioeconomic Segregation of Chilean Students, and its Relationship with Shared Financing).Mimeo, Department of Economics McEwan, P., 2001. The effectiveness of public, catholic, and non-religious private schools in Chile’s voucher system. Education Economics 9 (2), 103–128.
Vial, B., 1998. Financiamiento Compartido de la Educacio´n. Cuadernos de Economı´a McEwan, P., 2004. The potential impact of vouchers. Peabody Journal of Education West, E., 1997. Education vouchers in principle and practice: a survey. The World McEwan, P., Urquiola, M., Vegas, E., 2008. School choice stratification, and infor- Bank Research Observer 12 (1), 83–103.
mation on school performance: lessons from Chile. Economia Spring 1–42.
Willms, D., 1985. Catholic-school effects on academic achievement: new evidence Mizala, A., Romaguera, P., Urquiola, M., 2007. Socioeconomic status or noise? from the ‘high school and beyond’ follow-up study. Sociology of Education 58 Tradeoffs in the generation of school quality information. Journal of Develop- Willms, D, Somers, M.A., 2001. Family, classroom, and school effects on children’s Mizala, A., Romaguera, P., 2000. School performance and choice: the Chilean educational outcomes in Latin America. School Effectiveness and School Im- experience. Journal of Human Resources 35 (2), 392–417.

Source: http://www.educacion2020.cl/sites/default/files/estudio_mizala_torche.pdf