Social segregation in secondary schools: how does England compare with other countries?

Abstract

New evidence is provided about the degree of social segregation in England’s secondary schools, employing a cross‐national perspective. Analysis is based on data for 27 industrialised countries from the 2000 and 2003 rounds of the Programme of International Student Assessment (PISA). We allow for sampling variation in the estimates. England is shown to be a middle‐ranking country, as is the USA. High segregation countries include Austria, Belgium, Germany and Hungary. Low segregation countries include the four Nordic countries and Scotland. In explaining England’s position, we argue that its segregation is mostly accounted for by unevenness in social background in the state school sector. Cross‐country differences in segregation are associated with the prevalence of selective choice of pupils by schools.

Acknowledgements

This research was supported by ESRC grant RES‐000‐22‐0995 ‘Social segregation in UK schools: benchmarking with international comparisons’. We thank two anonymous referees; Lucinda Platt; and seminar participants at Southampton, LSE, the Institute of Education and Nuffield College, Oxford for helpful comments, and Ray Chambers for advice on bootstrapping of standard errors. Jenkins thanks DIW Berlin for their hospitality during work on the paper.

Notes

1. The evidence on peer group effects is reviewed by Vignoles et al. (2000), and other possible impacts of social segregation are described by Allen & Vignoles (2007).

2. See the Appendix to this paper available on the Taylor & Francis website (http://www.informaworld.com/mpp/uploads/254064appendix.pdf) and Micklewright & Schnepf (2006).

3. The prevalence of missing values and our treatment of them (imputation through regression modelling) is described in the Appendix to this paper (see note 2).

4. If there is no segregation, p_i /P = r_i /R in every school, and the geometric mean of the shares in school i in this case is simply p_i /P or, equivalently, r_i /R.

5. This is not to say that total segregation would fall by H _between were private schooling to be abolished. Further unevenness in social background would be introduced into the state sector as parents of ex‐private‐school pupils chose new schools, for example by moving into catchment areas of schools perceived to be of better quality.

6. That is, we repeatedly (400 times) draw samples of schools (with their PISA pupils) with replacement from the actual sample of schools for each country. (The sample size in each case is the same as for the original sample for each country.) We calculate D and H in each of these 400 samples. The standard deviations of these 400 values provide estimated standard errors of our levels of D and H for each country.

7. As a further check on the robustness of results, we compared the segregation curves for England, Scotland and Northern Ireland (see the Appendix to this paper described in note 2). We find that the curve for England lies everywhere outside that for Scotland, indicating that England is the more segregated country according to all ‘segregation curve consistent’ indices of segregation, and not only H. Such indices satisfy the transfers principle (which D does not) and three other desirable properties (organisational, compositional and size invariance).

8. The full set of values of D and H for all countries are given in the Appendix to this paper (see note 2).

9. We define a private school as one in which the principal reports ‘student fees or school charges paid by parents’ to make up 50% or more of the school’s income. See Jenkins et al. (2006) for results with private schools defined on a basis of management.

10. In fact, the within‐group value is a weighted sum of the values for three groups of schools rather than two: private schools, non‐private schools, and a small group of schools for which information on private status is missing. Excluding the schools with missing data on status from the calculations, and using just two groups, made a negligible difference to the results.

11. These two factors are asked about separately but we have combined them into a single category.

12. This does not mean that social segregation at the level of the LEA is no higher in those LEAs that still operate a grammar school system. (On this subject, see Burgess et al., 2004.)

13. We define school type to have seven categories in Germany, seven in Austria and four in Hungary. The three most important school tracks (in terms of numbers of 15 year olds in the PISA samples) are Gymnasium, Realschule and Hauptschule in Germany, Gymnasium, vocational and high vocational in Austria, and grammar, vocational secondary and vocational in Hungary. For Austria and Hungary, information on school type is available for 2003 and decompositions are for that year. The Austrian figure is based on all schools (funding information being missing in 2003), but the 2000 data indicate that only 4% of pupils are in privately funded schools.