Explaining social class inequalities in educational achievement in the UK: quantifying the contribution of social class differences in school ‘effectiveness’

Abstract

There are large social class inequalities in educational achievement in the UK. This paper quantifies the contribution of one mechanism to the production of these inequalities: social class differences in school ‘effectiveness’, where ‘effectiveness’ refers to a school’s impact on pupils’ educational achievement (relative to other schools). It builds on the small number of existing studies, whilst overcoming a number of their limitations. It estimates the effectiveness of the (state) schools attended between ages 7/8 and 10/11 by a cohort of children born in the 1990s in Avon, a former county of England, and then compares the effectiveness of the schools attended by children from different social classes. It finds that ‘higher’ social class children attend more effective schools, on average, and that these social class differences in average school effectiveness between ages 7/8 and 10/11 account for 7% of social class differences in average educational achievement age 10/11. This is not a causal analysis, however.

Keywords:

• Social class
• inequality
• school effectiveness
• school segregation

Acknowledgements

This article has benefitted from discussions with Leon Feinstein, Rebecca Allen, Nikos Tzavidis, Anna Vignoles, John Micklewright, and the ALSPAC study team.

Notes

1. An extended version of this section is available on request from the author.

2. For ease of exposition, these biases are described in the context of estimating school effects, but they also arise when estimating the effects of school characteristics.

3. In other words, school fixed, as opposed to random, effects are estimated. Nevertheless, these effects are highly correlated; in a sample similar to that used here, I found that the Pearson’s correlation coefficient between school fixed and random effects was 0.95 (Hobbs, 2007, p. 51).

4. Whilst I have not reviewed the evidence on differential school effectiveness, note that Strand (2010, p. 307) found ‘no evidence of substantial differential effectiveness’ between FSM and non-FSM pupils. Strand (2014b) also discusses the contribution of differential school effectiveness to the production of inequalities in educational achievement between FSM and non-FSM pupils.

5. For children with matched data, the Pearson correlation coefficient between mother’s age at birth and educational achievement age 11 is 0.21–0.22. More generally, mother’s age at first birth is correlated with children’s educational achievement in other studies.

6. For more details on the assessments and variable construction, see, respectively, QCA (2002) and Hobbs (2007, p. 37).

7. For more details on the assessments and variable construction, see, respectively, QCA (1998) and Hobbs (2007, pp. 38–39).

8. The results will be insensitive to the inclusion of the covariates up to age 4½, if they are insensitive to the covariates up to age 4½ per se, and, either they are insensitive to the covariates after age 4½, or they are sensitive to the covariates after age 4½ but the correlation between the covariates up to age 4½ and after age 4½ is weak. Alternatively, they will be sensitive to the inclusion of the covariates up to age 4½, if they are sensitive to the covariates up to age 4½ per se, and/or they are sensitive to the covariates after age 4½ and the correlation between the covariates up to age 4½ and after age 4½ is strong.

9. The results are almost identical if the sample is restricted instead to those in schools with at least two (unweighted) children.

10. While this paper has not discussed policy, section 2 discussed the mechanisms through which social class differences in school effectiveness might arise in the UK. This is a starting point for thinking about policies to eliminate social class differences in school effectiveness.