The Dynamics of School Attainment of England ’ s Ethnic Minorities

1. The Dynamics of School Attainment of England’s Ethnic Minorities Deborah Wilson, Simon Burgess, Adam Briggs February 2006

2. Introduction • Accumulation of human capital is a key to economic success for individuals and communities. • Relative achievement of minority ethnic learners is an on-going cause for concern among policy-makers in the UK. • A lot of evidence for US, rather less for UK. www.bristol.ac.uk/Depts/CMPO

3. In this paper: • We exploit a universe dataset of state school students in England • We document the evolution of attainment for different ethnic groups through school • We explore some factors lying behind relative achievement • Results: www.bristol.ac.uk/Depts/CMPO

4. We confirm some well-known facts for the high stakes exams taken at age 16: • pupils from some ethnic groups achieve considerably lower scores than white pupils on average – pupils with Black Caribbean heritage, other Black heritage or Pakistani ethnicity. • Students with Indian or Chinese ethnicity score much higher than their white peers www.bristol.ac.uk/Depts/CMPO

5. We provide some striking new findings: • All ethnic minority groups make greater progress on average than white students between ages 11 and 16. • Much of this improvement is in the high-stakes exams at the end of compulsory schooling. • For most ethnic groups, this gain relative to white students is pervasive, happening in almost all schools. www.bristol.ac.uk/Depts/CMPO

6. Our analysis addresses some of the usual factors invoked to explain attainment gaps, although these are typically about levels rather than growth in attainment • We consider the roles of poverty, language, school quality, and teacher influence • We analyse attainment gaps at the lower end of the distribution. www.bristol.ac.uk/Depts/CMPO

7. Plan • Literature • Data • Modelling Framework • Results I • Results II • Conclusions www.bristol.ac.uk/Depts/CMPO

8. English School System University, job, … Primary Secondary Age 5 7 11 14 16 18 A levels Tests KS1 KS2 KS3 KS4 = GCSEs End of compulsory schooling This paper www.bristol.ac.uk/Depts/CMPO

9. Data • PLASC/NPD: administrative data from the DfES. All pupils in English state schools; approx 0.5 million in each cohort. • Key Stage (KS) tests: • Cohort 1: KS1 (age 7) in 1998; KS2 (age 11) in 2002. • Cohort 2: KS2 in 1997; KS3 (age 14) in 2000; KS4 = GCSE (age 16) in 2002. • As yet, no single cohort going all the way through www.bristol.ac.uk/Depts/CMPO

10. Data II • PLASC/NPD gives individual characteristics: • Ethnicity • English as an additional language (EAL) • Eligibility for free school meals (FSM) • Gender, age within year • Special educational needs status (SEN) • Home postcode • School attended • Attainment data at each Key Stage • All but attainment data is for 2001/02 only. www.bristol.ac.uk/Depts/CMPO

11. Data III • Pupil home postcode enables us to match in local area data: • Index of multiple deprivation (IMD) • Ward level; 6 components (income; employment; health; education; housing; access to services) • MOSAIC • Postcode level dataset. Categorises each postcode into one of 61 types. www.bristol.ac.uk/Depts/CMPO

12. Data IV • Analysis sample: • study the cohorts as balanced panels – proportion of students with full record is high • track the same group through school without worrying about changing sample composition • Unrepresentative of all students taking tests? No, apart from Black African heritage students • Sample sizes: www.bristol.ac.uk/Depts/CMPO

13. www.bristol.ac.uk/Depts/CMPO

14. Table 2: Summary statistics of Key-stage scores for both cohorts www.bristol.ac.uk/Depts/CMPO

15. Measuring test score gaps • Different distribution of marks at each KS. At KS4 – SD here four times bigger than at KS2. Just using marks – hard to interpret gaps. • We do three things: • We use z scores – normalise each KS# mark separately by its mean and SD (all ethnic groups together). So units are in SD’s. • Use ranks • Discretise KS4 marks as alternative to treating KS2 marks as continuous www.bristol.ac.uk/Depts/CMPO

16. Plan • Literature • Data • Modelling Framework • Results I • Results II • Conclusions www.bristol.ac.uk/Depts/CMPO

17. Modelling Framework • Adopt a human capital approach – test score depends on human capital • hit = ht + SjgjtXij + SlaltZil + Smbmteim • The final term is the myriad influences on human capital can’t measure. • These may be correlated with a pupil’s ethnicity. So the coefficient on an ethnic group dummy summarises the correlation of membership of that ethnic group with these variables, weighted by their impact on human capital. www.bristol.ac.uk/Depts/CMPO

18. Role of schools • In most tables, we don’t focus on schools: • A straightforward interpretation of such variables would require the assumption that pupils are randomly allocated to schools • Interpretation of ethnicity is that it includes both the direct impact of that characteristic on test score, plus the indirect effect on school quality times the impact of that quality on test score. www.bristol.ac.uk/Depts/CMPO

19. Estimation • We estimate: yit = Sgpgt I(ethnic group)i + b1t.genderi + b2t.agei + b3t.FSMi + b4t.SENi + Sn b5nt.I(n’hood)i • We also look at a pupil’s progress over the Key Stages, referred to as value-added: • An individual pupil’s value added from KS2 to KS4 is the difference between her own grade and that average for those with the same KS2 score. www.bristol.ac.uk/Depts/CMPO

20. Plan • Literature • Data • Modelling Framework • Results I • Results II • Conclusions www.bristol.ac.uk/Depts/CMPO

21. Results I • Raw attainment gaps • Conditional attainment gaps • Value Added gaps • Attainment Gaps at Lower Quantiles • Quantifying the gap • Results II • Non-school factors • Systemic schooling factors • Between-school factors • Within-school factors www.bristol.ac.uk/Depts/CMPO

22. Graphical approach: 2000 1998 1997 2002 www.bristol.ac.uk/Depts/CMPO

23. KS Scores by ethnicity www.bristol.ac.uk/Depts/CMPO

24. Using ranks www.bristol.ac.uk/Depts/CMPO

25. Discretising KS4: www.bristol.ac.uk/Depts/CMPO

26. Results I • Raw attainment gaps • Conditional attainment gaps • Value Added gaps • Attainment Gaps at Lower Quantiles • Quantifying the gap • Results II • Non-school factors • Systemic schooling factors • Between-school factors • Within-school factors www.bristol.ac.uk/Depts/CMPO

27. Table 5: Regressions of standardised Key-stage 4 scores for cohort 2 www.bristol.ac.uk/Depts/CMPO

28. Table 6: Regressions of standardised key-stage scores www.bristol.ac.uk/Depts/CMPO

29. Figure 5: ‘Group’-White conditional gaps www.bristol.ac.uk/Depts/CMPO

30. Heterogeneity – matching analysis www.bristol.ac.uk/Depts/CMPO

31. Results I • Raw attainment gaps • Conditional attainment gaps • Value Added gaps • Attainment Gaps at Lower Quantiles • Quantifying the gap • Results II • Non-school factors • Systemic schooling factors • Between-school factors • Within-school factors www.bristol.ac.uk/Depts/CMPO

32. Table 7: Regressions of Key-stage 2 to 4 Value added for cohort 2 www.bristol.ac.uk/Depts/CMPO

33. Table 8: Regressions of value-added www.bristol.ac.uk/Depts/CMPO

34. Results I • Raw attainment gaps • Conditional attainment gaps • Value Added gaps • Attainment Gaps at Lower Quantiles • Quantifying the gap • Results II • Non-school factors • Systemic schooling factors • Between-school factors • Within-school factors www.bristol.ac.uk/Depts/CMPO

35. Z-scores: male, FSM, bottom 20% KS2 and IMD www.bristol.ac.uk/Depts/CMPO

36. Results I • Raw attainment gaps • Conditional attainment gaps • Value Added gaps • Attainment Gaps at Lower Quantiles • Quantifying the gap • Results II • Non-school factors • Systemic schooling factors • Between-school factors • Within-school factors www.bristol.ac.uk/Depts/CMPO

37. Table 10: Predicted Vs actual GCSE attainment by ethnicity www.bristol.ac.uk/Depts/CMPO

38. Results I • Raw attainment gaps • Conditional attainment gaps • Value Added gaps • Attainment Gaps at Lower Quantiles • Quantifying the gap • Results II • Non-school factors • Systemic schooling factors • Between-school factors • Within-school factors www.bristol.ac.uk/Depts/CMPO

39. Statistical factors • Regression to the mean • Split pupils from each ethnic group into gender*FSM*KS2 cells • Designate equivalent white pupils in each sub-cell; track these over subsequent KS’s. • Figure 8 www.bristol.ac.uk/Depts/CMPO

40. Figure 8: Performance of equivalent ‘Group’-White pupils www.bristol.ac.uk/Depts/CMPO

41. Non-school factors • Individual characteristics affect progress? • Language • PLASC records whether English is a pupil’s “mother tongue”, the language spoken at home. • Only two groups with some variation: Black Africans and Indian ethnicity students • Accounts for about a third of the gain for these two groups (Table 12) • Separate analysis of maths, english and science www.bristol.ac.uk/Depts/CMPO

42. Systemic Schooling Factors • Differences in assessment? • No: consistent assessment across KS2 – KS4. • Teacher expectations or bias? • Yes: greater divergence between mark and teacher assessment for some groups (Table 13) • Role of Special Educational Needs (SEN) indicator? • Not conditioning on SEN, same results on progress. www.bristol.ac.uk/Depts/CMPO

43. Between-school factors • School quality • the quality of the teachers, the ethos and leadership of the school, and peer groups • Non-random allocation of pupils to schools • Comparing fixed effects and OLS means comparing average variation within a school, to variation both within and across schools. • Look at London only www.bristol.ac.uk/Depts/CMPO

44. Table 14: School fixed effects vs OLS www.bristol.ac.uk/Depts/CMPO

45. Within-school factors • Differences in school practices? • For each school and for each ethnic group, we ask whether that group has higher mean value added than white students. • Table 15 presents the percentage of schools for which that group improves relative to whites. www.bristol.ac.uk/Depts/CMPO

46. Table 15: Proportion of schools/LEAs where ethnic group progress relative to White pupils is positive www.bristol.ac.uk/Depts/CMPO

47. Hot off the press … Black Caribbean Black African Pakistani Indian Conditional score gaps www.bristol.ac.uk/Depts/CMPO

48. Plan • Literature • Data • Modelling Framework • Results I • Results II • Conclusions www.bristol.ac.uk/Depts/CMPO

49. Conclusions • Minority ethnic groups make better average progress through secondary school than do white students. • These gains are substantial for some groups, only marginal for students of Black Caribbean heritage. • These gains are pervasive for most of the groups. • The gains are particularly marked in the final exams that are crucial for further progress in education or jobs. www.bristol.ac.uk/Depts/CMPO

50. Findings suggest systemic factors: the importance of aspirations and values? • Modood (2005): “Asian trajectory … social mobility by education, self-employment and progression into the professions” • Winder (2004): “familiar immigrant paradigm”: “the children of immigrants, lacking financial capital of their own, devote themselves to the acquisition of knowledge” www.bristol.ac.uk/Depts/CMPO