Bold choices: how ethnic inequalities in educational attainment are suppressed

Abstract

In this paper, I examine ethnic inequalities in educational attainment in England and Wales. I focus on the two main educational transitions in England and Wales: the transition at age 16, from compulsory to post-compulsory education, and the transition at age 18, from school to university. I take into account the distinction made by Boudon (1974) between ‘primary’ and ‘secondary’ effects, and ask how far overall ethnic inequalities in educational attainment can be attributed to primary or secondary effects. The paper first assesses the extent of gross ethnic inequalities in the chances of making each transition, before asking how far the picture is altered by controlling for social class background. I then determine the relative importance of primary and secondary effects in creating ethnic inequalities in educational attainment. Results show that both primary and secondary effects are operating to produce ethnic inequalities in educational attainment. In general, where ethnic groups are disadvantaged relative to the white majority, this is due to their lower average levels of performance. But conditional on their performance, ethnic minority students are much more likely to choose to make educational transitions, suggesting that if performance effects were eliminated, all ethnic minority groups would be advantaged relative to the white majority.

Keywords:

• ethnic inequality
• primary effects
• secondary effects
• A-level
• university degree
• performance
• choice

Acknowledgements

I would like to thank David Grusky, Jan O. Jonsson, Frida Rudolphi, participants in the Sociology Seminar at Bamberg University, members of the ‘Primary and Secondary Effects’ team in the EU-funded EQUALSOC network, participants at the Stanford RC28 meeting and the two anonymous reviewers for helpful comments on this paper. I am also grateful to the Economic and Social Research Council for funding under the Understanding Population Trends and Processes initiative (RES-163-27-1002).

Notes

1. As four years is a relatively short time period, it is reasonable to suppose that no damage will be done to the results by combining the three cohorts. During the time period covered, only one policy change of note occurred, relating to the Advanced-level (A-level) qualification. In 2000, the A-level was split into two parts. Instead of working towards an A-level qualification over two years (typically), students take an Advanced Subsidiary-level (AS-level) after one year and then complete the A-level by taking A2 examinations (House of Commons, 2003). Although it had been possible to gain an AS-level qualification since 1987, the policy change meant that it was now not possible to gain a full A-level qualification without first gaining an AS-level. If this policy change was to be consequential for our understanding of inequalities in educational attainment, we would expect the policy change to be reflected in different transition rates to A-level education for individuals from the 2000 and 2002 cohorts. Inspection of these transition rates shows that no step change occurred between the 1998 cohort and the later cohorts.

2. Respondents can identify themselves as being of mixed ethnic background either by choosing the ‘Mixed’ ethnic group option (in YCS11) or by identifying themselves as a member of more than one of the listed groups (in all three YCS surveys).

3. The National Statistics Socio-Economic Classification, or NS-SEC, became Britain’s official class schema in 2000, and is in effect an updated version of the Goldthorpe schema (see Rose & Pevalin, 2003 for more details).

4. This is explicitly acknowledged in the Government’s discussion of ‘Functional Skills’, where mathematics and English language, alongside ICT skills, are described as, ‘essential elements … that individuals need to enable them to engage successfully as citizens and progress to further learning or employment’ (DCSF, 2010, p. 2).

5. In addition to the models described above, I also ran a model to test whether the effects of ethnicity differed by gender. The results showed that the ethnicity/gender interaction was only significant for the Bangladeshi group: Bangladeshi girls were significantly more likely to make the transition than Bangladeshi boys. As the interaction effect was insignificant for all other ethnic minority groups, and to preserve power for the later analyses, I conduct pooled analyses from this point forward.

6. The gross z-score for each ethnic group is estimated from an OLS regression of performance on ethnic group. Net z-scores are estimated from a second model, regressing performance on ethnic group and class background; class is held constant at mean values to estimate the net performance score after controlling for class background.

7. The ldecomp package in Stata was used to calculate the actual and counterfactual transition rates. This is described in Buis (2010).

8. Results for the full analyses are available from the author on request.

9. The fairlie package in Stata was used to apply the Fairlie decomposition (see Jann, 2006).

10. Full results available from the author on request.

11. Although the transition rates estimated using the method are rather close to the empirical transition rates, the implied odds ratios are for three groups in the opposite direction from the empirical rates (the Black Caribbean, Other Black, and Other groups). This is because the empirical transition rates for these groups are in fact extremely close to that of the white majority population, and the empirical odds ratios are therefore extremely close to 1.