Macro-Political Determinants of Educational Inequality between Migrants and Natives in Western Europe

Abstract

This paper analyses how macro-societal conditions determine educational inequality between migrants and natives in the Western EU member states. The education system, with its ability to integrate young immigrants, is often seen as the foundation for successful integration in later curricular stages. By means of linear hierarchical models, this paper demonstrates that EU standards of good practice in integration policy are ineffective in producing equal educational opportunities for immigrant pupils. Educational inequality between migrants and natives is mainly a result of the political-institutional settings. The paper thus rejects the widely held belief that consensus democracies are in fact kinder and gentler. By contrast, the analyses reveal that majoritarian democracies are more effective in equalising educational opportunities between immigrants and natives.

Acknowledgements

For extremely helpful comments and/or the provision of data, we would like to thank Prof. Dr. Markus Freitag, Julian Bernauer, Prof. Dr. Adrian Vatter, Jennifer Shore, Anne-Sophie Fendrich, Prof. Dr. Kerstin Martens, and the researchers at the Bremen International Graduate School of Social Sciences (BIGSSS) and the SFB ‘Transformations of the State’ at the University of Bremen.

Notes

 1. For all analyses in this article we apply the PISA 2006 data provided by the OECD. Source: http://pisa2006.acer.edu.au/downloads.php. Countries included in the analyses and abbreviations: Austria (AUT), Belgium (BEL), Denmark (DNK), Finland (FIN), France (FRA), Germany (DEU), Great Britain (GBR), Greece (GRC), Ireland (IRL), Italy (ITA), Luxembourg (LUX), Netherlands (NDL), Portugal (PRT), Spain (ESP), and Sweden (SWE). Cyprus and Malta are not included since they do not participate in PISA (Programme for International Student Assessment).

 2. In Lithuania and Latvia, pupils’ immigration background does not significantly affect educational achievement when controlling for social background and gender. Moreover, Bulgaria, Hungary, Romania, and Poland all have very low percentages of pupils with immigration backgrounds (<4 per cent). The effects of immigration background on educational achievement in the Eastern EU are therefore not reliable (and moreover not significant).

 3. The European Civic Citizenship and Inclusion Index ( http://www.migpolgroup.com; see Appendix 2) is the forerunner of the Migrant Integration Policy Index (MIPEX): http://www.integrationindex.eu/

 4. Source: http://www.integrationindex.eu/use?tg=All&st=49&ct=37 (accessed 30 June 2011).

 5. We apply hierarchical linear modelling instead of a two-step multi-level analysis, since we tried to account for more than two levels: students, schools, and countries. According to Levels et al. (2008) and Schlicht et al. (2010), it is highly appropriate to at least control for the school level. Moreover, even though two-step models are assumed to be more efficient, they would in this case not solve the trade-off between a large pool of potential macro-political conditions opposed to very few cases on the macro-level.

 6. It is worth mentioning that this approach does not allow us to draw any conclusions about the relative importance of the macro-conditions when individually added to the models. It should be clarified that this study is one important step in a rather inductive research programme. The findings will be highly valuable for deriving more sophisticated theories on the evolution of inequality structures.

 7. Appendix 4 includes the main effects of the macro-political variables on pupils’ mathematical achievement in general.

 8. For measuring education policy we use the data of Schlicht et al. (2010) (Appendix 3). Two of the education policy variables – ‘Pupil-to-teacher-ratio’ and ‘Minimum hours of teaching per year’ – include several missing values. These cases are excluded from the analyses. Due to the small number of cases on the macro level, the results regarding these two variables are limited by uncertainty and should be interpreted with caution. For the cultural origin of the immigrants (Appendix 7) detailed data were only available for 2009. Moreover, the data only include the share of foreign citizens but not EU citizens with a migration background.

 9. Thanks to Julian Bernauer (University of Konstanz, Germany) and Adrian Vatter (University of Bern, Switzerland) for providing us with access to these data.

10. Due to the small number of cases on the macro-societal level, we apply restricted iterative generalised least squares (RIGLS) estimations in all our hierarchical models (Rasbash et al. 2009). RIGLS estimations provide ‘less biased estimates of the variance than IGLS when the number of highest-level units is small’ (Rasbash et al. 2009: 184). These estimates are thus less sensitive to outliers.

11. 27 per cent of the total mathematics achievement variation can be ascribed to the school level and 63 per cent to the individual level.

12. We calculate several models, each containing one of the macro-societal variables. It is not reasonable to simultaneously include the macro-context variables in one model for two reasons: first, due to the small number of units at the country level, it is not possible to integrate all contextual variables plus cross-level interactions into one model. Second, the macro-variables are partly strongly related (see Appendices 8a and 8b).

13. With regard to education policy, the only significant effect also contradicts the hypothesis: the stronger the enrolment in preschool education in a Western EU member state, the higher the degree of educational inequality between migrants and natives (see Appendix 10).

14. The inclusion of ethnic fractionalisation as a macro-effect resulted in a significant negative cross-level interaction (see Appendix 4). However, the marginal effects of individuals’ immigrant background at different levels of ethnic fractionalisation are not significantly different.