Transcript view presentation - MS Powerpoint

Investigating Inequalities in Educational
Attainment
Michelle Jackson
Department of Sociology
Nuffield College, Oxford
Inequalities in educational attainment

Development of educational systems during 20th Century as response to
changing economic and occupational structures

Sociological interest in class, ethnic and sex inequalities in educational
attainment

E.g. Class inequalities
 Children of salariat (professional and managerial) background are
around five times more likely to take A-level courses, rather than
taking vocational courses or leaving education than are children of
working class background
 In England&Wales, seems to be little change in extent of class
inequalities over time
Primary and secondary effects

Boudon
 Primary effects – those that result from previous academic
performance (may be present due to range of factors – e.g. genetic,
cultural)
 Secondary effects – those that result from educational choices made
by children

In this project, examining relative importance of primary and secondary
effects in creating inequalities in educational attainment

Following results relate to class inequalities in transition to A-level in
England and Wales (see reference)

Look at the transition to A level at three points in time
 When students are 16 in 1974, 1987, 1996
Data

Use data from National Child Development Study and Youth Cohort Study
 1974: NCDS. Continuing birth cohort study covering all children born
in GB in one week in 1958
 1987 and 1996: YCS. Study commissioned by DEE (now DfES).
Cohorts of young people in England and Wales aged 16 and upwards

Three variables in analyses
 Class background. Father’s (or head of household’s) Goldthorpe class.
Use 3 class simplification: salariat, intermediate+petty bourgeoisie,
working class
 Academic performance. Performance in public examinations in
mathematics and English. Scores attached to grades summed,
inverted, and standardised to be z-scores with mean of 0, s.d. of 1
 Transition to A level
 In NCDS, whether student in education after age of 16
 In YCS, question asking whether studying A/AS levels
Descriptive statistics
Per cent making transition
2: Information on class,
1: Information on
transition&performance
class&transition
Class
Year
Salariat
1974
1987
1996
40
62
76
51
65
77
Intermediate+PB
1974
1987
1996
19
40
49
29
44
52
Working
1974
1987
1996
10
28
37
17
33
40
All
1974
1987
1996
21
43
58
33
48
61
Distinguishing primary and secondary effects

Run binary logistic regression
 Response variable: whether an individual reaches A level education or
not
 Explanatory variable: standardised performance scores (maths and
English scores)

Analyses run separately for each class

Three time points – 1974, 1987, 1996
Graphical representation of regression of
transition to A level work on academic
performance: 1974
1
Data Range
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
3
Academic performance
Salariat
Intermediate
Working Class
Salariat
Intermediate
Working Class
3.5
4
Probability of transition
Proportion of cases
1
Graphical representation of regression of
transition to A level work on academic
performance: 1987
1
Data Range
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
3
Academic performance
Salariat
Intermediate
Working Class
Salariat
Intermediate
Working Class
3.5
4
Probability of transition
Proportion of cases
1
Graphical representation of regression of
transition to A level work on academic
performance: 1996
1
1
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
3
Academic performance
Salariat
Intermediate
Working Class
Salariat
Intermediate
Working Class
3.5
4
Probability of transition
Proportion of cases
Data Range
Primary effects

Differences in performance distributions between three classes

Primary effects clearly operate, with no evidence of general decline
Salariat
Intermediate
Diff SI
1974
0.42
Diff IW
-0.06
0.48
1987
0.39
-0.39
-0.05
0.39
0.81
-0.35
0.30
-0.20
0.59
Diff SW
0.33
0.44
1996
Working
0.74
-0.47
0.27
0.86
Graphical representation of regression of
transition to A level work on academic
performance: 1974
1
Data Range
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
3
Academic performance
Salariat
Intermediate
Working Class
Salariat
Intermediate
Working Class
3.5
4
Probability of transition
Proportion of cases
1
Graphical representation of regression of
transition to A level work on academic
performance: 1987
1
Data Range
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
3
Academic performance
Salariat
Intermediate
Working Class
Salariat
Intermediate
Working Class
3.5
4
Probability of transition
Proportion of cases
1
Graphical representation of regression of
transition to A level work on academic
performance: 1996
1
1
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
3
Academic performance
Salariat
Intermediate
Working Class
Salariat
Intermediate
Working Class
3.5
4
Probability of transition
Proportion of cases
Data Range
Secondary effects

Over time, curves start sharp upward rise at lower levels of performance

Strong class differences in all three periods

Gaps between curves widest at intermediate levels of performance
(around 0). Gaps narrow as move to either extreme of performance range

What is the relative importance of primary and secondary effects?
Integrating

Integral to be evaluated:
abx 

 1
( x   )2 / 2 2  e
dx

abx  
4  2 e

 1  e

4
where μ is the mean of the performance scores and σ the standard
deviation and a is the constant and b the performance coefficient from the
regression model

By calculating integral, can distinguish two components of any class
transition rate
 Can calculate transition rates for each class
 Can carry out counterfactual analyses by combining performance
distribution for one class with transition propensities of another
Results of integrations
Counterfactuals

What would happen if we allowed intermediate and working class children
to maintain their own performance distribution, but to have the same
transition propensities as salariat children?
1974
1987
per cent
1996
Class
Salariat
Actual
49
63
76
Intermediate
Actual
Counterfactual
Difference
28
35
7
43
52
9
50
56
6
Working Class
Actual
Counterfactual
Difference
17
27
10
33
44
11
38
47
9
Odds ratios
1974
1987
1996
Actual
2.49
2.24
3.17
Counterfactual
1.75
1.61
2.49
Actual
1.90
1.56
1.58
Counterfactual
1.51
1.36
1.43
Actual
4.72
3.48
5.03
Counterfactual
2.63
2.18
3.57
Salariat/Intermediate
Intermediate/Working
Salariat/Working
Conclusions and future work

Both primary and secondary effects are important. If we eliminated
secondary effects, there would be substantial impact on class differentials

Will examine later educational transition – school to university

Method can be used to look at other inequalities. Will also examine ethnic
and sex inequalities in educational attainment
Conclusions and future work

Datasets: NCDS, BCS, YCS

Comparative analyses with colleagues from Sweden, France, Germany, the
Netherlands

Policy implications
 Policy which could eliminate primary effects would clearly have great
impact
 However, effects of pre-school intervention likely to wash out later on.
Eliminating secondary effects might be a more plausible policy goal

Reference
 Jackson, M., Erikson, R., Goldthorpe, J. H. and Yaish, M. (forthcoming)
‘Primary and Secondary Effects in Class Differentials in Educational
Attainment: the Transition to A-Level Courses in England and Wales,
Acta Sociologica