Research report

Trajectories of socioeconomic inequalities in health,

behaviours and academic achievement across
childhood and adolescence
Laura D Howe,1 Debbie A Lawlor,1 Carol Propper2
Additional material is
published online only. To view
please visit the journal online
(http://dx.doi.org/10.1136/jech2012-201892).
1

MRC Centre for Causal

Analyses in Translational
Epidemiology, School of Social
and Community Medicine,
University of Bristol, Bristol, UK
2
Centre for Market and Public
Organisation, University of
Bristol, Bristol, UK
Correspondence to
Dr Laura D Howe,
Oakeld House, Oakeld
Grove, Bristol BS8 2BN, UK;
[email protected]
Received 28 August 2012
Revised 7 November 2012
Accepted 25 November 2012
Published Online First
15 January 2013

ABSTRACT
Background Socioeconomic inequalities are a key
policy challenge. Studies to date have not taken a
unied approach to assess how socioeconomic
inequalities in health, behaviour and educational
attainment change as children age.
Methods We examined maternal education inequalities
in multiple offspring health, behavioural and educational
outcomes and how these changed across childhood and
adolescence in the Avon Longitudinal Study of Parents
and Children, a cohort born in 1991/1992 in South-West
England (N=556011 463).
Results Inequalities were observed for some health
measures (blood pressure (BP), height, cholesterol, bone
mineral density (BMD) and fat-mass (females)) but not
in other measures ( parent-assessed child health,
triglycerides, fat-mass (males), high-density lipoproteincholesterol, C reactive protein). The strongest health
inequality was systolic BP (mean difference comparing
highest to lowest maternal education 0.28 SD (95% CI
0.35 to 0.20), approximately 2.6 mm Hg. Wide
inequalities, similar in magnitude to BP, were observed
for behavioural outcomes. Even greater inequalities were
observed for offspring academic achievement (mean
difference comparing highest to lowest maternal
education 1.43 SD (95% CI 1.37 to 1.50), a difference
of 22%). For all behavioural outcomes and some health
indicators, inequality was stable over childhood. For
some outcomes (BP, BMD and most education
outcomes), inequality narrowed as children got older.
Only for height and attainment in English tests was
there evidence of widening inequalities with age.
Conclusions Our results suggest that within this
cohort, maternal education inequalities in offspring
health, behaviour and educational attainment are
established in childhood but do not increase up to
adolescence. Maternal education inequalities in
behaviour and educational attainment were considerably
larger than in health measures.

INTRODUCTION

Open Access
Scan to access more
free content

To cite: Howe LD,

Lawlor DA, Propper C. J
Epidemiol Community Health
2013;67:358364.
358

Socioeconomic inequalities in health remain as one of

the biggest challenges to health policy. An emerging
evidence base demonstrates socioeconomic inequalities in health in childhood and adolescence,1 a relationship between socioeconomic position (SEP) in
childhood and health in adulthood,2 3 and the existence of inequalities in childhood behavioural problems4 5 and educational attainment.5 Thus, it is
possible that inequalities in adult health may be
already determined in childhood and may be accentuated by socioeconomic inequalities in behaviour and
educational attainment in childhood/adolescence.

Some studies show widening of socioeconomic

inequalities in health across childhood.68 However,
this nding is not universal.7 912 While differences in
geographical setting and birth year of study participants may explain some of the differences between
study ndings, methodological factors could also be
important.
Much of the existing literature on changes in
socioeconomic inequalities over childhood uses
cross-sectional studies and examines whether
inequalities differ for children of varying ages.612
Any age effects seen in such analyses could result
from cohort effects, that is, they may be driven by
differences between different birth cohorts rather
than by differences in age. Most studies also focus
on a limited number of health indicators and do
not explore how socioeconomic inequality patterns
are similar (or not) to those of behaviour or educational attainment. Our aim is to compare socioeconomic inequalities during childhood, and how
they change as children age, across multiple
domains of childhood: health, behaviour and educational attainment. The larger the inequalities and
the more they widen, the more likely are they as
precursors of adult health inequalities, and therefore important targets for policies.

METHODS
Data
The Avon Longitudinal Study of Parents and
Children (ALSPAC) is a prospective cohort
study.13 14 Pregnant women resident in one of the
three Bristol-based health districts with an expected
date of delivery between 1 April 1991 and 31
December 1992 were invited to take part. From
the 14 541 women recruited, 13 988 children were
alive at 1 year. Ethical approval was obtained from
the ALSPAC Law and Ethics Committee and the
Local Research Ethics Committees.
We analyse outcomes from as close as possible to
birth/infancy, 7, 9, 11 and 15 years, although exact
ages and number of measurement occasions vary
between outcomes as detailed below.

Socioeconomic position
We use maternal education as the measure of SEP.
Analyses with paternal education, head of household occupational class and family income as the
measure of SEP produced very similar results to
those for maternal education (results available from
authors) and therefore we present associations with
maternal education only.
A questionnaire at 32 weeks of gestation asked
mothers to report their educational attainment,

Howe LD, et al. J Epidemiol Community Health 2013;67:358364. doi:10.1136/jech-2012-201892

Research report
which was categorised as below O-level (ordinary level; examinations taken in different subjects usually at age 1516 at the
completion of legally required school attendance; equivalent to
todays UK General Certicate of Secondary Education), O-level
only, A-level (advanced-level; examinations taken in different
subjects usually at age 18), or university degree and above.

Health indicators
Offspring health indicators are parent-assessed overall child
health (higher scores indicate worse health), height, total body
fat-mass assessed by whole body dual x-ray accelerometry scan,
blood lipids (triglycerides, high-density lipoprotein cholesterol
(HDLc) and serum total cholesterol (hereafter referred to as
cholesterol)), C reactive protein (CRP), systolic and diastolic
blood pressure (SBP and DBP) and total body-less-head bone
mineral density (BMD). Full measurement details are provided
in online supplementary material.

Behavioural outcomes
Caregivers were asked in self-completed questionnaires to rate
their childs behaviour and conduct using the Strengths and
Difculties Questionnaire. Full details are provided in online
supplementary material. Answers are grouped into scores for
total difculties, hyperactivity-inattention, emotional, conduct
and peer problems.

Educational outcomes
We use scores in nationally set examinations as measures of offspring educational attainment in English, mathematics and
science, and an overall summary of these. Full details are provided in the online supplementary material.

Choice of outcomes
We wished to examine a range of outcomes covering different
domains of childhood: health, behaviour and education. We a
priori selected the outcomes from a large range of measurements
available in ALSPAC on the basis of their known associations
with adult disease and key socioeconomic achievements/outcomes. The childhood health indicators track strongly across the
life course1518 and have all been associated with key chronic diseases in adulthood (cardiovascular disease, diabetes and osteoporosis/fractures) that are known to be associated with SEP.1925
Taller height is associated with both socioeconomic and health
advantage.21 Lower levels of fat mass, cholesterol, triglycerides,
CRP, SBP, DBP and higher levels of HDLc are advantageous for
cardiovascular health. Higher levels of BMD are associated with
lower risk of osteoporosis. Parent-assessed overall child health
(higher scores indicating worse health) is similar to a self-report
assessment of general health that in adults is associated with premature mortality and a wide range of chronic diseases and is
strongly socioeconomically patterned.11 26 The behavioural outcomes we analyse (higher scores indicating more difculties) are
associated with future mental-health problems in adults,27 and
educational attainment is predictive of later income, occupation,
health and other adult outcomes.28

We t random-effects linear regression models of each

outcome on a rank score of maternal education, including an
interaction term between maternal education and age (using
ages 0, 7, 9, 11 and 15 years as appropriate for the measurement schedule of the outcome and centred on the rst age of
measurement for a given outcome). We use the Stata command
xtreg, and calculate 95% CIs using robust SEs. This allows for
the clustering of repeat outcome measures within individuals
and also enables us to use all available measures for all eligible
individuals with the outcome measured on at least one occasion
(under a missing at random assumption). The random effects
model produces two results for each outcome: the coefcient
for the rank score of maternal education, which is the slope
index of inequality (SII, mean difference in outcome comparing
highest to lowest maternal education at the rst age of assessment for a given outcome) and the coefcient for the interaction between the rank score of maternal education and age,
which is a measure of change in the SII per year of greater age
(model details in online supplementary material).
Examination of stratum-specic coefcients, together with
statistical tests for interaction, demonstrated evidence of gender
differences in the maternal education-outcome associations for
fat-mass, HDLc, CRP and hyperactivity but not for other outcomes. For these outcomes, analyses are presented separately
for males and females. For other outcomes, analyses are presented with data from males and females combined and with
adjustment for gender. All analyses are adjusted for the childs
exact age (weeks) at outcome measurement.
To facilitate comparisons between outcomes and to allow for
the changing scale of some outcomes over age, all outcomes are
standardised to mean of zero, variance of one by subtracting the
mean and dividing by the SD within age and gender categories.
SIIs are therefore the mean difference in SD units of the
outcome comparing highest with lowest maternal education.

Sensitivity analyses
To explore whether any changes in inequality across time are
linear, we plot the SII at each age. To examine deviation from a
linear association between categories of maternal education and
outcomes, we plot the trajectory of each outcome across each
category. To explore any consequences of missing data, analyses
were repeated using three alternative approaches in addition to
our main analysis using data from all individuals with one or
more measures for a given outcome: (1) restricting to individuals with at least two measures for a given outcome, (2) restricting to individuals with complete data for at all time points for a
given outcome and (3) using multivariate multiple imputation
(details in online supplementary material).

RESULTS
Sample characteristics
Sample sizes varied from 5560 (blood-based outcomes) to
11 463 (height) (table 1). Approximately 23% of participants
had mothers in the lowest educational group (less than O-level)
and 15% had mothers educated to a degree level (table 2).

Socioeconomic inequalities
Statistical analyses
Maternal education data are available for 12 493 ALSPAC participants (89% of those alive at 1 year). For each outcome, analysis is restricted to individuals with data on the outcome for at
least one of our time points.
Total fat-mass, triglycerides and CRP were right-skewed so
natural logs were used in all analyses.

For the offspring health measures, higher maternal education was

associated with taller height, lower fat-mass (females only), lower
cholesterol, SBP and DBP and higher BMD (table 3 and gure 1).
There was little inequality in fat-mass (males only), triglycerides,
HDLc or CRP. Children from higher maternal education families
had slightly worse overall parent-reported health (higher score),
but the CI included the null value. The physical health measures

Howe LD, et al. J Epidemiol Community Health 2013;67:358364. doi:10.1136/jech-2012-201892

359

Research report

360

Table 1

Distributions of outcome measures and sample size at each age

Health indicators
Global parent-assessed health, n (%)
1 (very healthy, no problems)
2 (healthy, few minor problems)
3 or 4 (sometimes quite ill, or almost always unwell)
Mean (SD)
Height (cm), mean (SD)
DXA-assessed fat-mass (kg), median (IQR)

Howe LD, et al. J Epidemiol Community Health 2013;67:358364. doi:10.1136/jech-2012-201892

Cholesterol (mmol/l), mean (SD)

HDL cholesterol (mmol/l), mean (SD)
Triglycerides (mmol/l), median (IQR)
C reactive protein (mg/l), median (IQR)
SBP (mm Hg), mean (SD)
DBP (mm Hg), mean (SD)
BMD (g/cm2), mean (SD)
Behavioural outcomes
Total difficulties score (range 040), mean (SD)
Hyperactivity score (range 010), mean (SD)
Emotional problems score
(range 010), mean (SD)
Conduct difficulties score (range 010), mean (SD)
Peer problems score (range 010), mean (SD)
Educational outcomes
Summary educational attainment measure (%), mean (SD)
English test scores (%), mean (SD)
Mathematics test scores (%), mean (SD)
Science test scores (%), mean (SD)

Birth/infancy

7 years

9 years

11 years

N=9895
4817 (48.7)
4685 (47.4)
393 (4.0)
1.55 (0.57)
N=9655, 50.63 (2.49)

N=7576
4652 (61.4)
2786 (36.8)
138 (1.8)
1.40 (0.53)
N=7474, 125.61 (5.39)

N=7493
4902 (65.4)
2476 (33.0)
115 (1.5)
1.36 (0.51)
N=6971, 139.51 (6.28)
N=6702, 7.18 (4.79, 10.94)

N=6596
4237 (64.2)
2263 (34.3)
96 (1.5)
1.37 (0.51)
N=6510, 150.76 (7.25)
N=6411, 9.78
(6.63, 15.26)

N=7346, 98.86 (9.18)

N=7344, 56.46 (6.63)

N=4674,
N=4674,
N=4674,
N=4674,
N=6292,
N=6292,
N=6702,

4.27 (0.66)
1.40 (0.31)
1.00 (0.76, 1.38)
0.21 (0.11, 0.54)
102.83 (7.71)
60.00 (6.88)
0.90 (0.05)

Total N (at least

one measure)*

15 years

N=6446, 105.49 (9.86)

N=6446, 58.72 (6.53)
N=6411, 0.95 (0.06)

10663

N=4810, 169.26 (8.37)

N=4766, 13.62 (8.28, 19.93)
N=3203,
N=3203,
N=3203,
N=3203,
N=4704,
N=4704,
N=4766,

3.75 (0.64)
1.28 (0.29)
0.75 (0.60, 0.98)
0.39 (0.22, 0.89)
123.12 (10.78)
67.59 (8.75)
1.11 (0.08)

11463
7642
5560
5560
5560
5560
8488
8487
7642

N=7119, 7.09 (4.21)

N=7119, 3.28 (2.30)
N=7119, 1.42 (1.56)

N=6353, 6.24 (4.20)

N=6353, 2.79 (2.17)
N=6353, 1.39 (1.60)

N=5819, 5.92 (4.19)

N=5819, 2.60 (2.13)
N=5819, 1.33 (1.56)

8708
8708
8708

N=7119, 1.52 (1.35)

N=7119, 0.97 (1.28)

N=6353, 1.16 (1.26)

N=6353, 1.00 (1.32)

N=5819, 1.08 (1.26)

N=5819, 0.96 (1.34)

8708
8708

N=10365,
N=10506,
N=10582,
N=10633,

66.16
58.63
64.94
73.95

(15.54)
(15.74)
(21.19)
(15.20)

N=8856,
N=9022,
N=9276,
N=9330,

52.24
46.27
54.80
54.29

(11.20)
(17.26)
(14.62)
(13.53)

10659
10735
10812
10840

*The total sample size for each outcome is the number of individuals with one or more measure at any age. The statistical analysis technique uses all available measures for each individual, as well as the observed changes with age across the
whole dataset, to estimate a full trajectory across all ages for that individual.
BMD, bone mineral density; DBP, diastolic blood pressure; DXA, dual x-ray accelerometry; HDL, high-density lipoprotein; SBP, systolic blood pressure.

Research report
Table 2 Socioeconomic position of individuals included in
analyses of inequalities in fat-mass and comparison with ALSPAC
participants not included in analyses
Participants included
in our analyses in
fat-mass models,
N=7642
Maternal education, N (%)
Less than O-level 1776 (23.2)
O-level
2694 (35.3)
A-level
1995 (26.1)
Degree
1177 (15.4)

Non-included
ALSPAC
participants,
N=4851

1977 (40.8)
1636 (33.7)
808 (16.7)
430 (8.9)

p Value for
comparison

Socioeconomic inequalities were greatest for offspring educational attainment (table 4 and gure 1). Going from lowest to
highest maternal education was associated with an increase in
overall offspring educational attainment of 1.43 SD (95% CI
1.37 to 1.50); this equates to a difference of 22%. Wide socioeconomic inequalities were evident for all three subdomains of
offspring educational attainment, ranging from 1.28 SD (95%
CI 1.21 to 1.34) for mathematics to 1.36 SD (95% CI 1.30 to
1.43) for English.

<0.001

Fat-mass was chosen for illustrative purposes only. It has a sample size in the middle
of our range of sample sizes; N was greatest for analyses of height and smallest for
analysis of blood-based measures. Patterns of differences were similar for each
outcome analysis and excluded set. The p value is derived from a 2 test.
ALSPAC, Avon Longitudinal Study of Parents and Children.

with the greatest inequality were fat-mass (females only) and BP;
children with the highest compared with lowest maternal education
had SBP on average 0.28 SD lower (95% CI 0.35 to 0.20),
which represents a difference of 2.6 mm Hg.
Socioeconomic inequalities in behavioural outcomes were generally of a similar or greater magnitude to inequalities in SBP
(table 4 and gure 1). Higher maternal education was associated
with lower total difculties in the child of 0.39 SD (95% CI
0.47 to 0.31 SD), hyperactivity of 0.35 SD in males (95% CI
0.46 to 0.23) and 0.51 SD in females (95% CI 0.62
to 0.39), conduct problems of 0.26 SD (95% CI 0.34 to
0.18) and peer problems of 0.30 SD (95% CI 0.38 to 0.22).
A 0.39 SD difference for total difculties equates to a higher
overall difculty of 1.7 points out of a possible total 40. There
was no evidence of socioeconomic inequalities in emotional
problems.

Changes in socioeconomic inequalities over childhood

and adolescence
Maternal education differences in offspring cholesterol, SBP,
DBP and BMD narrowed as the children got older (table 3 and
gure 1). For cholesterol, the socioeconomic inequality that was
apparent at age 9 had disappeared by age 15 (gure 1 and
online supplementary gure S1). For BP, maternal education differences were static between ages 7 and 9 years, and then narrowed between 9 and 11 and again between 11 and 15 years
(see online supplementary gure S1). There was evidence of
slight widening of socioeconomic inequalities in height with
age; when the coefcients are examined graphically, this
appeared to be driven by increasing height inequality between
ages 11 and 15 (see online supplementary gure S1). Although
the statistical evidence for changes in maternal education differences in total body fat-mass and CRP was weak for females
(p=0.1 for fat-mass and p=0.1 for CRP, table 3) the graph
showed that for females the inequalities widened slightly with
age (see online supplementary gure S1); in males there were
only weak socioeconomic differences at any age. There was no
evidence for changes with age for socioeconomic inequalities in
overall parent-assessed child health, triglycerides or HDLc;
inequalities in these were weak at all ages.
For most of the behavioural outcomes, there was no evidence
of change in socioeconomic differences as the children got

Table 3 Socioeconomic inequalities in child health measures, and changes in these over childhood

Outcome
Overall
parent-assessed
health
Height
Total fat-mass
Cholesterol
Triglycerides
HDL cholesterol
CRP
SBP
DBP
BMD

Inequality by maternal education at

baseline (first age of assessment)

Change in inequality with each year of

age

SII (95% CI)

Interaction with age

(95% CI)

Age (years) at
first measure

Gender

Combined 10663

0
9
9
9
9
9
9
9
9
7
7
9

Combined 11463 0.15 (0.09 to 0.22)

Male 3799 0.8 (0.20 to 0.04)
Female 3843 0.29 (0.41 to 0.17)
Combined 5560 0.12 (0.23 to 0.02)
Combined 5560 0.02 (0.08 to 0.13)
Male 2803 0.10 (0.24 to 0.04)
Female 2757 0.10 (0.05 to 0.25)
Male 2803 0.03 (0.18 to 0.12)
Female 2757 0.24 (0.39 to 0.08)
Combined 8488 0.28 (0.35 to 0.20)
Combined 8487 0.24 (0.31 to 0.16)
Combined 7642 0.12 (0.03 to 0.20)

0.07 (0.004 to 0.14)

p Value
0.07

<0.001
0.2
<0.001
0.02
0.7
0.2
0.2
0.7
0.003
<0.001
<0.001
0.007

0.01 (0.01 to 0.003)

0.01
0.001
0.01
0.02
0.004
0.01
0.02
0.001
0.03
0.03
0.02
0.01

(0.004 to 0.02)
(0.02 to 0.02)
(0.03 to 0.003)
(0.001 to 0.04)
(0.03 to 0.02)
(0.01 to 0.04)
(0.004 to 0.05)
(0.04 to 0.04)
(0.07 to 0.01)
(0.01 to 0.04)
(0.01 to 0.04)
(0.03 to 0.002)

p Value
0.2

0.002
0.9
0.1
0.06
0.7
0.3
0.09
1.0
0.1
<0.001
0.01
0.02

p for gender
differences
0.8

0.7
0.004
0.8
0.1
0.02
0.002
0.4
0.3
0.2

SII represents the mean difference in SDs of the outcome between the individuals with the hypothetical highest and lowest maternal education at baseline (intercept)that is, at the
first age at which outcomes were assessed. The interaction with age coefficient represents the additional change in SDs of the outcome between the hypothetical highest and lowest
SEP for every 1 year increase in the childs age. 95% CIs are calculated using robust SEs. Results are adjusted for the childs exact age in weeks at the time of outcome measurement
and the childs gender. Fat-mass is additionally adjusted for height and height squared.
BMD, bone mineral density; CRP, C-reactive protein; DBP, diastolic blood pressure; HDL, high density lipoprotein; SBP, systolic blood pressure; SEP, socioeconomic position; SII, slope
index of inequality.

Howe LD, et al. J Epidemiol Community Health 2013;67:358364. doi:10.1136/jech-2012-201892

361

Research report

Figure 1 The SII (slope index of inequality) is plotted against age for each outcome. All outcomes are standardised to have a mean of zero and a
variance of one. The SII is the mean difference in SDs of the outcome between the highest and lowest maternal education. Graphs are presented for
males and females combined m and f if no evidence of gender interactions was found, or separately for males and females for outcomes where
there was evidence of gender differences in the association with maternal education. Abbreviations of outcome names: BMD, bone mineral density;
CRP, C reactive protein; DBP, diastolic blood pressure; HDLc, high-density lipoprotein cholesterol, SBP, systolic blood pressure; SDQ, summary score
of behavioural difculties measured by the strengths and difculties questionnaire, education, summary score of attainment in tests in English,
mathematics and science. This gure is only reproduced in colour in the online version.
older; inequalities in total difculties, hyperactivity and conduct
problems remained similar across childhood (table 4, gure 1
and online supplementary gure S1). Socioeconomic inequalities
in peer problems reduced slightly as children got older, with

each increasing year closing the gap by on average 0.03 SD

(95% CI 0.01 to 0.06). Overall there was little evidence of
socioeconomic inequalities in emotional problems, but this is
the only behavioural outcome for which maternal education

Table 4 Socioeconomic inequalities in behavioural measures and educational attainment, and changes in these over childhood
Age (years)
at first
measure
Gender
Behavioural difficulties
Total difficulties
Hyperactivity
Conduct problems
Peer problems
Emotional problems
Educational attainment
Overall attainment
English test scores
Mathematics test scores
Science test scores

Inequality by maternal education at

baseline (first age of assessment)

Change in inequality with each year

of age

SII (95% CI)

p Value

Interaction with age

(95% CI)

<0.001
<0.001
<0.001
<0.001
<0.001
0.3

0.004
0.003
0.01
0.01
0.03
0.03

(0.02 to 0.03)
(0.03 to 0.03)
(0.02 to 0.04)
(0.04 to 0.01)
(0.01 to 0.06)
(0.05 to 0.003)

0.7
0.9
0.4
0.3
0.02
0.023

0.2
0.4
0.2

<0.001
<0.001
<0.001
<0.001

0.09
0.04
0.05
0.18

(0.11 to 0.08)
(0.02 to 0.05)
(0.07 to 0.03)
(0.20 to 0.16)

<0.001
<0.001
<0.001
<0.001

0.9
0.7
0.6
0.9

7
7
7
7
7
7

Combined
Male
Female
Combined
Combined
Combined

8708
4443
4265
8708
8708
8708

0.39
0.35
0.51
0.26
0.30
0.05

11
11
11
11

Combined
Combined
Combined
Combined

10659
10735
10812
10840

1.43
1.36
1.28
1.34

(0.47 to
(0.46 to
(0.62 to
(0.34 to
(0.38 to
(0.13 to
(1.37
(1.30
(1.21
(1.28

0.31)
0.23)
0.39)
0.18)
0.22)
0.04)

to 1.50)
to 1.43)
to 1.34)
to 1.41)

p Value

p for gender
differences

0.7
0.1

SII represents the mean difference in SDs of the outcome between the individuals with the hypothetical highest and lowest SEP maternal education at baseline (intercept)that is, at
the first age at which outcomes were assessed. The interaction with age coefficient represents the additional change in SDs of the outcome between the hypothetical highest and
lowest SEP for every 1 year increase in the childs age. 95% CIs are calculated using robust SEs. Results are adjusted for the childs exact age in weeks at the time of outcome
measurement and the childs gender.
SEP, socioeconomic position; SII, slope index of inequality.

362

Howe LD, et al. J Epidemiol Community Health 2013;67:358364. doi:10.1136/jech-2012-201892

Research report
differences widened as children get older (table 4, gure 1 and
online supplementary gure S1).
Maternal education inequalities in offspring educational
attainment mostly decreased in magnitude as the children aged.
With each year, maternal educational differences in overall offspring educational attainment decreased by 0.09 SD (95% CI
0.11 to 0.08) (table 4, gure 1 and online supplementary
gure S1). Socioeconomic inequalities in offspring attainment in
mathematics and science tests also narrowed as children got
older, with the most narrowing for science scores. There was a
slight widening in socioeconomic inequalities in offspring
English test scores. However, for all offspring educational outcomes, socioeconomic inequalities remained large at all ages.

Sensitivity analyses
For most offspring outcomes, the differences in outcomes
between maternal education categories and how these changes
over time are linear across the four SEP categories (see online
supplementary gure S2). One notable exception, however, is
cholesterol. Levels of cholesterol increase with age for the two
higher maternal education categories, but remain stable for the
lower two categories. Maternal education differences in peer
problems appear to be driven by differences between the lowest
maternal education category (<O-level) and all other categories,
whereas socioeconomic inequalities for fat mass in males are
driven by differences between the highest maternal education
category (degree) and all others. Emotional problems decrease
with age for the highest maternal education category (degree)
but increase for all other categories. Our results and conclusions
were robust to sensitivity analysis for different ways of treating
missing data (see online supplementary table S1).

DISCUSSION
We have assessed childhood socioeconomic inequalities and how
these change with increasing age, for a range of health, behavioural
and educational outcomes within a birth cohort from the UK, born
in 1991/1992. We nd strong socioeconomic inequalities in most
outcomes but, in contrast to some hypotheses,6 little evidence that
inequality widens as children get older. Socioeconomic inequalities
in many of the outcomes we study remained stable over childhood
and adolescence, and inequalities in other outcomes narrowed.
Only maternal education differences in offspring height and educational attainment in English widened as children got older but,
even for these, the change with age was small. The lack of widening
of socioeconomic inequalities in offspring educational attainment
is in contrast to previous research, which demonstrated that children from wealthy families who scored poorly in early educational
tests tended to catch up, whereas children from poorer families did
not catch up,29 although recent research suggests this nding may
be at least partially due to regression to the mean, and that once
appropriate analysis techniques are used inequalities in educational
attainment may be large but fairly stable across childhood.30

Differences in magnitudes of socioeconomic inequalities

between health, behavioural and educational outcomes
in childhood
We utilised standardised measures across a range of outcomes
for a single cohort. This allows comparison of socioeconomic
inequalities across outcomes. Socioeconomic inequalities were
by far greatest for offspring educational attainment compared
with other outcomes. Strong socioeconomic inequalities were
also observed for behavioural outcomes. Maternal education differences in offspring health indicators were absent or less pronounced, and sometimes decreased as children got older.

The weaker socioeconomic inequalities in health, compared

with behavioural or educational outcomes, may be due to SEP
having an immediate and direct effect on behaviour and education in childhood but a lagged effect on physical health. Effects
on health indicators may emerge later in life, in part through
indirect pathways related to the impact of behavioural difculties and education on lifestyle behaviours that impact on health.
The differences in patterns of change in socioeconomic
inequality between types of outcome could hint at underlying
mechanisms and life course models. For example, socioeconomic
inequalities in behavioural outcomes were stable across childhood and adolescence suggesting that these are established very
early in life and that school factors may not succeed in narrowing
these differences. The narrowing in BP and BMD socioeconomic
inequalities is consistent with the theory that there is equalisation
of health during adolescence,11 although support for this is tempered by the fact we did not observe this pattern for other outcomes. For BP and BMD, we need postadolescent measurements
in this cohort to determine whether these socioeconomic
inequalities re-emerge after this period.

Study strengths and limitations

The key strengths of this study are its large sample size, the use of
repeat measurements of outcomes across childhood and the comparison of socioeconomic inequalities across a range of different
outcomes. As is common in prospective birth cohort there has been
attrition over time, leading to a reduced sample size and a more
socioeconomically advantaged group of participants at older ages,
which may lead to underestimation of the extent of socioeconomic
inequalities.31 However, our analysis strategy enabled us to use all
available outcome data for all participants with at least one
outcome measure, therefore minimising the chance that this has
affected our results. Our sensitivity analysis indicated that our
results were not affected by selection bias. However, for outcomes
with only two measurements, changes in inequalities over childhood should be interpreted with caution and require replication in
other studies. Our conclusions are limited by the fact that we do
not have early measurements on all outcomes, and also do not
have outcome measures beyond adolescence and so we cannot see
how these socioeconomic inequalities in childhood go on to inuence inequalities in adulthood; as the ALSPAC participants move
into adulthood we will be able to explore this. Our sample was predominantly of white European origin. We are thus unable to
examine ethnic inequalities in health, and our ndings may not
generalise to those from other ethnic backgrounds. The nature of
socioeconomic inequalities and how they change over childhood
may be context specic, differing between countries and generations; the main study demonstrating widening of health inequalities
across childhood was conducted in the USA,6 whereas another
study that did not nd evidence of widening socioeconomic
inequalities over childhood used data from the UK.10 Thus, the
generalisability of our ndings needs to be assessed by replicating
the research in other cohorts. We have assessed inequalities on the
mean difference scale, that is, absolute inequalities; examining relative differences (ratios) can sometimes produce different patterns,32
so our result cannot be extrapolated to draw conclusions about
relative inequalities. The relationships between maternal education
and child/adolescent outcomes observed in this analysis will be
mediated through many different pathways, for example, parental
health, environmental exposures, quality of schooling, parenting
techniques; exploration of these mediating pathways would
provide insight into the socioeconomic inequalities and potential
interventions that could alleviate them, but this is beyond the scope
of the current analysis.

Howe LD, et al. J Epidemiol Community Health 2013;67:358364. doi:10.1136/jech-2012-201892

363

Research report
CONCLUSION AND IMPLICATIONS

REFERENCES

We sought to compare socioeconomic inequalities in outcomes

across multiple domains of childhood as precursors to possible
inequalities in adult health. We observed socioeconomic inequalities in a range of health, behavioural and educational outcomes
in childhood. Associations were strongest for educational and
behavioural outcomes and considerably weaker for health indicators. Up to age 15 years we found no strong evidence that
inequalities increased with increasing age across childhood.
Thus, while childhood socioeconomic inequalities in health and
health-related characteristics might be important determinants
of adult health inequalities, our results suggest that such
inequalities are established in early childhood but do not
increase markedly up to adolescence.

1
2
3

4
5

6
7
8
9

What is already known on this subject

10

Low socioeconomic status is associated with worse child

health, behaviour and educational attainment.
Some studies suggest that socioeconomic inequalities in
health widen as children get older.
Studies to date have not taken a unied approach to
examine how socioeconomic inequalities in multiple
domains of childhood (health, behaviour and educational
attainment) change with age.

11
12
13

14
15

16
17

What this study adds

18

Our results suggest that health, behavioural and educational

inequalities are established in childhood but do not increase
up to adolescence.
Inequalities in behaviour and education were considerably
larger than in health measures.

19

20

21
22

Contributors The study was designed by CP and LDH. LDH, DAL and CP designed
the analysis strategy. LDH carried out statistical analysis and wrote the rst draft of
the manuscript. DAL and CP made critical edits and revisions to the manuscript, and
all authors read and approved the nal version of the manuscript.
Funding This work was supported by a grant from the UK Economic and Social
Research Council (RES-060-23-0011), which paid the salary for LDH. The UK
Medical Research Council (MRC), the Wellcome Trust and the University of Bristol
provide core funding support for ALSPAC. The UK MRC (G0600705) and the
University of Bristol provide core funding for the MRC Centre of Causal Analyses in
Translational Epidemiology. The views expressed in this paper are those of the
authors and not necessarily those of any funding body or others whose support is
acknowledged. The funders had no role in study design, data collection and
analysis, decision to publish, or preparation of the manuscript.

23

24
25
26
27
28

Competing interests None.

Ethics approval ALSPAC law and ethics committee.

29

Provenance and peer review Not commissioned; externally peer reviewed.

30

Data sharing statement The data access policy for the ALSPAC cohort is
published on its website: http://www.bristol.ac.uk/alspac
Open Access This is an Open Access article distributed in accordance with the Creative
Commons Attribution Non Commercial (CC BY-NC 3.0) license, which permits others to
distribute, remix, adapt, build upon this work non-commercially, and license their
derivative works on different terms, provided the original work is properly cited and the
use is non-commercial. See: http://creativecommons.org/licenses/by-nc/3.0/

364

31
32

Adler NE, Ostrove JM. Socioeconomic status and health: what we know and what
we dont. Ann N Y Acad Sci 1999;896:315.
Banks J, Marmot M, Oldeld Z, et al. Disease and disadvantage in the United
States and in England. JAMA 2006;295:203745.
Lawlor DA, Sterne JAC, Tynelius P, et al. Association of childhood socioeconomic
position with cause-specic mortality in a prospective record linkage study of
1,839,384 individuals. Am J Epidemiol 2006;164:90715.
Ermisch J. Origins of social immobility and inequality: parenting and early child
development. Nat Inst Econ Rev 2008;205:6271.
Sacker A, Schoon I, Bartley M. Social inequality in educational achievement and
psychosocial adjustment throughout childhood: magnitude and mechnisms. Soc Sci
Med 2002;55:86380.
Case A, Lubotsky D, Paxson C. Economic status and health in childhood: the
origins of the gradient. Am Econ Rev 2002;92:130834.
Chen E, Matthews KA, Boyce WT. Socioeconomic differences in childrens health: how
and why do these relationships change with age? Psychol Bull 2002;128:295329.
Currie J, Stabile M. Socioeconomic status and child health: why is the relationship
stronger for older children? Am Econ Rev 2003;93:181323.
Chen E, Martin AD, Matthews KA. Socioeconomic status and health: do gradients
differ within childhood and adolescence? Soc Sci Med 2006;62:216170.
Currie A, Shields MA, Price SW. The child health/family income gradient: evidence
from England. J Health Econ 2007;26:21332.
West P. Health inequalities in the early years: is there equalisation in youth? Soc Sci
Med 1997;44:83358.
West P, Sweeting H. Evidence on equalisation in health in youth from the West of
Scotland. Soc Sci Med 2004;59:1327.
Boyd A, Golding J, Macleod J, et al. Cohort prole: the Children of the 90sthe
index offspring of the Avon Longitudinal Study of Parents and Children. Int J
Epidemiol 2012.
Fraser A, Macdonald-Wallis C, Tilling K, et al. Cohort prole: the Avon Longitudinal
Study of Parents and Children: ALSPAC mothers cohort. Int J Epidemiol 2012.
Shear CL, Burke GL, Freedman DS, et al. Value of childhood blood pressure
measurements and family history in predicting future blood pressure status: results
from 8 years of follow-up in the Bogalusa Heart Study. Pediatrics 1986;77:8629.
Chen X, Wang Y. Tracking of blood pressure from childhood to adulthood: a
systematic review and meta-regression analysis. Circulation 2008;117:317180.
Kalkwarf HJ, Gilsanz V, Lappe JM, et al. Tracking of bone mass and density during
childhood and adolescence. J Clin Endocrinol Metab 2010;95:16908.
Singh AS, Mulder C, Twisk JW, et al. Tracking of childhood overweight into
adulthood: a systematic review of the literature. Obes Rev 2008;9:47488.
Davey Smith G, Hart C, Upton M, et al. Height and risk of death among men and
women: aetiological implications of associations with cardiorespiratory disease and
cancer mortality. J Epidemiol Community Health 2000;54:97103.
Lawlor DA, Ebrahim S, Davey Smith G. The association between components of
adult height and type II diabetes and insulin resistance: British Womens Heart and
Health Study. Diabetologia 2002;45:1097106.
Batty GD, Shipley MJ, Gunnell D, et al. Height, wealth, and health: an overview
with new data from three longitudinal studies. Econ Hum Biol 2009;7:13752.
Owen CG, Whincup PH, Orfei L, et al. Is body mass index before middle age related
to coronary heart disease risk in later life? Evidence from observational studies. Int J
Obes 2009;33:86677.
Oren A, Vos LE, Uiterwaal CSPM, et al. Cardiovascular risk factors and increased
carotid intima-media thickness in healthy young adults: the Atherosclerosis Risk in
Young Adults (ARYA) Study. Arch Intern Med 2003;163:178792.
McCarron P, Davey Smith G, Okasha M, et al. Blood pressure in young adulthood
and mortality from cardiovascular disease. Lancet 2000;355:14301.
Law MR, Wald NJ, Meade TW. Strategies for prevention of osteoporosis and hip
fracture. BMJ 1991;303:4539.
Idler EL, Benyamini Y. Self-rated health and mortality: a review of twenty-seven
community studies. J Health Soc Behav 1997;38:2137.
McLeod JD, Kaiser K. Childhood emotional and behavioural problems and
educational attainment. Am Sociol Rev 2004;69:63658.
Ofce for National Statistics. Proportional effect on earnings of a degree level
qualication: by sex and degree subject, 19932001: Social Trends 34. 2011. UK.
18-4-2011.
Feinstein L. Inequality in the early cognitive development of British children in the
1970 cohort. Economica 2003;70:7397.
Jerrim J, Vignoles A. Social mobility, regression to the mean and the cognitive
development of high ability children from disadvantaged homes. J R Stat Soc Series
A 2012. ePub ahead of Print. doi: 10.1111/j.1467-985X.2012.01072.x
Howe LD, Tilling K, Galobardes B, et al. Loss to follow-up in cohort studies: bias in
estimates of socioeconomic inequalities. Epidemiol 2013;24:19.
Lynch J, Davey Smith G, Harper S, et al. Explaining the social gradient in coronary
heart disease: comparing relative and absolute risk approaches. J Epidemiol
Community Health 2006;60:43641.

Howe LD, et al. J Epidemiol Community Health 2013;67:358364. doi:10.1136/jech-2012-201892