A cross-national analysis of the relations of school choice and effectiveness differences between private-dependent and public schools

Abstract

We apply propensity score matching to the estimation of differential school effectiveness between the publicly funded private sector and the public sector in a sample of 26 countries. This technique allows us to distinguish between school choice and school effectiveness processes and thus to account for selectivity issues involved in the comparison of the 2 sectors. Concerning school choice, we found 2 patterns: a choice of upwardly mobile parents for private schools and a preference for segregation by (lower) middle-class parents. As regards school effectiveness, our results indicate that, after controlling for selectivity, a substantial advantage in reading achievement remains among students in publicly funded private schools in 10 out of the 26 countries.

Keywords:

• school choice
• school effectiveness
• private-dependent and public schools
• international comparison

Notes

 1. This article is one of the products of the cross-national project “Religious Education in a Multicultural Society: School and Home in Comparative Context”, directed by Emer Smyth (Economic & Social Research Institute, Dublin, Ireland) and financed by the European Commision within the 7th Frame Work (FP7-SSH-2007-1- REMC).

 2. In subsequent studies, we will compare the choice/effectiveness gaps between public and private-independent schools, as well as compare school effects concerning student attitudes on environmental issues and school climate.

 3. For the description of this technique, we use the article of Frisco, Muller, and Frank (2007), who use this technique in their analysis of the effects of parental divorce on children's well-being.

 4. We use nearest neighbour and nearest neighbour with Mahalanobis distance matching.

 5. Normally, this should not be a big problem (see Morgan & Winship, 2007).

 6. The PISA survey does not distinguish between denominational and non-denominational schools.

 7. We could only use the first PISA wave for France, because the public dataset of the second and third PISA waves does not contain valid values for French public and private school indicator. We have to assume that this private-public distinction has become a state secret, too annoying for the secular French Republic to be published.

 8. There were too few charter schools in the USA (which are considered by PISA as private-dependent schools) to be included.

 9. While this admission based on programs is necessary for all French and Norwegian private-dependent schools.

10. This is a negative coefficient in Table 2, because the variable is the student–teacher ratio.

11. Based on country of birth of (one of) the parents outside the country of birth and/or the foreign language used at home.

12. Given the different measurement scales of variables, we multiplied the coefficient by the existing range in the sample to compute a maximum effect; also, the analysis uses the Gower measure and the Ward's method of clustering.

13. We had to delete from the cluster analysis those countries (Argentina, Canada, France, Indonesia, Norway) for which not all parameters are available.

14. Luxembourg is clearly a separate case, which has few in common with the two clusters.

15. Information about the characteristics of the matched treated and control pupils and their parents is available from the first author.