Do private schools increase academic achievement? Evidence from France

ABSTRACT

This article investigates the effect of private lower secondary schools on student achievement in France. I use propensity score matching on a large French database to estimate the effect of enrollment in a private school on academic achievement as measured by ninth-grade test scores in three school subjects. I find that private school attendance has a large and significant effect on educational success. Boys' (girls') scores in private school were between 0.193 (0.138) and 0.222 (0.198) standard deviations higher on standardized tests in ninth grade. A series of checks confirm the robustness of these results.

KEYWORDS:

• Private schools
• public schools
• secondary school
• academic achievement
• school choice
• propensity score matching

JEL CLASSIFICATION:

• I20
• I21
• I24

Acknowledgments

I thank Colin Green and two anonymous referee for their comments. I am grateful to Mehtabul Azam, Giovanni Cerulli, Marion Monnet, as well as seminar and conference participants at AFSE, ASSA, and INED for their valuable advice. I also thank Christopher Leichtnam and Paul Reeve for help in preparing the paper for publication.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Matching is one of the methods based on selection on observables. Other popular approaches are regression-adjustment, reweighting, and the doubly robust estimator (Cerulli 2015).

2 For ease of reading, in the remainder of the paper, private schools under contract with the state are referred to simply as ‘private schools’. Independent private schools are excluded from this study.

3 The panel year refer to the year in which students enter sixth-grade.

4 The baccalauréat is the French national exam at the end of secondary school.

5 Priority education in France includes two categories of schools. In 2006, they were classified as RAR (Réseaux ambition réussite), the schools with the most socially and academically disadvantaged category of students, and RRS (Réseaux de réussite scolaire), for schools with less disadvantaged students (Cayouette-Remblière and Moulin 2019). In 2011, RAR became ECLAIR (Ecoles, collèges et lycées pour l'ambition, l'innovation et la réussite (schools for ambition, innovation, and success)); in 2015, they became REP+ (Réseaux d'éducation prioritaire renforcée (reinforced priority education networks)), while the RRS schools were renamed REP (Réseaux d'éducation prioritaire (priority education networks)).

6 Without mentioning motivation by name, Fack and Grenet (2010) refer to a similar criterion as ‘good behavior’ and ‘commitment to the school's values’.

7 The assignment of a teacher to a public school depends on a score that is calculated on the basis of results in competitive recruitment examinations, seniority, personal situation (disability, request for transfer to reunify with a spouse who works elsewhere), or having taught in a priority education school.

8 Furthermore, this avoidance of the public sector in favor of the private sector is not without consequences for the social composition of schools. Thus, from a study of student enrollment in three French urban areas, Boutchénik, Givord, and Monso (2021) show that a higher proportion of the most advantaged families avoid public school catchment areas by sending their children to private school, thus contributing to an increase in school segregation between lower secondary schools.

9 The brevet des collèges is the French national exam at the end of lower secondary education.

10 These exams do not determine access to grade 6. They are used as an assessment measure to detect students in difficulty.

11 Before the 2007 panel, standardized test scores were not available for either sixth or ninth grade. However, using scores obtained during in-classroom assessment within schools would lead to incorrect results as Cayouette-Remblière and Moulin (2019) showed.

12 The variables for father's and mother's socio-occupational categories, monthly family income, family type, and number of siblings were constructed from the 2011 family survey. Although the survey took place during treatment, this is not a problem for PSM insofar these variables cannot be influenced by the treatment. The variable includes permanent and transitory income; the data set does not allow one to be distinguished from the other. In England, Anders et al. (2020) highlighted a distinct and larger effect of permanent over transitory income on private school participation.

13 This practice is common in the literature, and the inclusion of students with missing baseline scores can change the results (Krueger and Zhu 2004). Among these students, students in private schools are slightly overrepresented (25.9% versus 21.4% for students in public schools).

14 I have chosen to limit my sample to students who are enrolled in an academic track, because including the other tracks in the model would create a specific outcome which would no longer be a variable measured at the national level, as are ninth-grade test scores, but one which would instead be strongly dependent on the local context, since the decisions to orient students into different tracks are defined locally by the teaching teams.

15 I include the scores of students who repeated a grade in lower secondary school in the model, but I do not create a specific variable for grade retention during lower secondary education. This is because estimation based on PSM requires choosing explanatory variables that are not influenced by the treatment, whereas grade retention takes place after the assignment of the treatment. Grade retention policies are decided at a local level in the schools themselves (Blanchard and Cayouette-Remblière 2016), and there are disparities in these policies between public and private schools. For more on grade retention in France, I refer the reader to Gary-Bobo, Goussé, and Robin (2016).

16 Students who did not participate in the family surveys do not show a significant difference in terms of private school attendance (15.5% for students in private schools and 15.7% for students in public schools).

17 There are very few private schools in the priority education program that are included in the analysis.

18 For a recent work on aspirations on educational pathways, I refer the reader to Guyon and Huillery (2021).

19 In addition to the 2007 panel of secondary school students, I use the 2007 version of the national database on schools and students (Base Centrale Scolarité) to calculate the proportion of private schools in municipalities of different population sizes (nine categories).

20 However, this does not mean that all students who attend a private elementary school will attend a private lower secondary school. This phenomenon is linked to two elements. The first is the availability of places in lower secondary private schools, and the fact that students, whether or not they were enrolled in an elementary private school, may not be admitted. The second pertains to the proximity of a public secondary school to the parental home and the parents' preference (for reasons of reputation) to send their child there.

21 The implementation of the PSM procedure used here is as specified by Caliendo and Künn (2011), Caliendo, Künn, and Weißenberger (2016), and Caliendo and Tübbicke (2022).

22 This choice is based on the widely discussed gap in educational achievement between boys and girls (Fryer and Levitt 2010; Machin and McNally 2005; Machin and Pekkarinen 2008). Moreover, this practice is commonly used in the literature to analyze differences in the effect of the variable of interest on the outcomes of girls and boys (see Caliendo, Hujer, and Thomsen 2015; Caliendo, Künn, and Weißenberger 2016, for example).

23 This corresponds to a difference in scores between private and public school students ranging from 2.163/100 for girls in French to 4.962/100 for boys in mathematics. Compared to the average score, this difference is quite small for girls in French (average score of 57.257/100), whereas it is relatively large for boys in mathematics (average score of 48.466/100).

24 I compute hidden bias equivalents following DiPrete and Gangl (2004).

25 In France, a distinction is made between three categories of teachers: agrégé teachers, who are recruited through a highly selective competitive examination; the agrégation, certified teachers who are recruited through a less selective competition; and other teachers who are not recruited through competitive examinations. In theory, having more highly qualified teachers could help improve students' academic performance. And yet in France, 13% of public school teachers are agrégés versus only 4% in private schools (author's calculations from MENJ - DEPP 2020).