Content uploaded by Alice Sullivan
Author content
All content in this area was uploaded by Alice Sullivan on Feb 23, 2016
Content may be subject to copyright.
Reading for pleasure and attainment in vocabulary and
mathematics
Dr. Alice Sullivan (corresponding author)*
Director, 1970 British Cohort Study
Centre for Longitudinal Studies
Mr. Matt Brown
Survey Manager, 1970 British Cohort Study
Centre for Longitudinal Studies
*Contact details:
Centre for Longitudinal Studies
Institute of Education, University of London
55-59 Gordon Square
LONDON
WH1H 0NU
Tel: 020 76126661
Email: a.sullivan@ioe.ac.uk
Acknowledgements: We are grateful to the BCS70 cohort members for their
contribution over many years. Thanks to Tarek Mostafa for providing the analysis in
Appendix A1. Thanks to Dick Wiggins for comments on an earlier draft, and to two
anonymous reviewers.
1
Reading for pleasure and progress in vocabulary and
mathematics
Abstract
This paper examines inequalities in attainment in vocabulary and mathematics at age 16 for a
nationally representative cohort of people born in Britain in 1970 (the 1970 British Cohort
Study). Our analytical sample is n=3, 583 cohort members who completed vocabulary and
mathematics tests at age 16. We explore whether inequalities due to childhood social
background are similar across the linguistic and mathematical domains, or whether they
differ, and to what extent these inequalities are driven by families’ social class position,
parents’ education and home reading resources. We examine the role of children’s own
reading for pleasure controlling for all these background factors. As reading can be seen as an
indicator of ‘cultural capital’, we also test the influence of an alternative indicator of cultural
capital, playing a musical instrument. Our longitudinal analysis addresses the question of the
extent to which differences in attainment are determined by age 10; and which factors are
linked to a growth in differentials during adolescence. We show that childhood reading is
linked to substantial cognitive progress between the ages of 10 and 16, whereas playing an
instrument is not. Reading is most strongly linked to progress in vocabulary, with a weaker,
but still substantial link to progress in maths. Strikingly, reading for pleasure is more strongly
linked than parental education to cognitive progress in adolescence.
2
Keywords: Reading; longitudinal; BCS70; vocabulary; mathematics. Introduction
Persistent socio-economic inequalities in educational attainment and cognitive scores have
been documented by many studies over the years, and the explanation of these social
inequalities is one of the central problems within the sociology of education (Halsey et al.
1980). Debate continues regarding the mechanisms via which privileged families ensure the
educational and subsequent occupational success of their children, and the relative
importance of economic and cultural resources in determining class differentials in
educational outcomes.
Sociologists have put forward both cultural and economic explanations for educational
inequalities. The most prominent theory emphasising the importance of cultural resources,
Bourdieu’s theory of cultural reproduction (Bourdieu & Passeron [1977] 1990) has generated
much debate about the nature of cultural capital. Interpretations can be divided into the
‘status seeking’ and ‘information processing’ views (Ganzeboom 1982). According to the
status seeking view, children who display cultural attributes that are valued by the school are
rewarded by teachers with high attainment. In contrast, the ‘information processing’ view
suggests that particular cultural activities actually foster intellectual development. We have
argued that reading is distinctive because it develops linguistic ability and wider knowledge
(Sullivan 2002; 2007). Reading differs in this way from other cultural activities such as
music or going to galleries and museums (often termed ‘beaux arts’ participation). We know
that the home reading culture is important for children’s early cognitive scores (Byford et al.
2012). Past studies have also found that books in the home and children’s reading behaviour,
but not ‘beaux arts’ participation, help to explain social differentials in children’s outcomes
(De Graaf et al. 2000; Sullivan 2001) (Cheung & Andersen 2003; Georg 2004; Jaeger 2011;
Sullivan 2001).
The role of language has been neglected in most empirical studies of cultural reproduction,
yet Bourdieu’s discussion of cultural capital emphasises the importance of language as the
key to success in school. "Obvious in the literary disciplines but more subtle in the sciences,
the ability to manipulate academic language remains the principal factor in success in
examinations" (Bourdieu et al. 1994) (p.21). Hirsch (1983) also emphasises vocabulary,
though from a different perspective, on the grounds that knowledge of words is both an
adjunct to knowledge of concepts and assists further learning. It has been argued that reading
must be a particularly important driver of vocabulary development, given the paucity of
vocabulary used in speech compared to books, even comparing children’s books to adult
speech (Cunningham & Stanovich 1998).
Children pick up styles of speech, vocabulary and forms of reasoning simply by hearing their
parents talk, and also pick up reading habits through seeing their parents read, and having
reading materials readily available in the home (Sullivan 2007). We may expect this passive
cultural transmission to be most important in the case of linguistic skill, since studies have
found huge differences in the number of different words that children are exposed to in
middle and working class homes (Hart & Rinsley 1995). Vocabulary is transmitted within the
home almost constantly, without any conscious effort. In contrast, while parents may well
3
seek to promote their children’s success in subjects such as mathematics, this must typically
be done consciously, with discrete time set aside for the task. It is arguable therefore that we
should expect wider disparities according to parental cultural resources in children’s
linguistic ability than in their abilities in mathematics. On the other hand if reading ability
and vocabulary are central to further learning in all disciplines, we may expect the influences
on vocabulary development to also drive attainment in other subjects including mathematics.
Evidence for this view is provided by a recent study of monozygotic twins (Ritchie et al.
2014 in press) showing that learning to read well leads to improved general intelligence. In
this paper, we are able to address whether the socio-economic and cultural factors driving
vocabulary are essentially the same as those driving mathematics attainment, or whether they
differ.
Studies focussing on children’s reading have faced challenges in unpacking the reciprocal
relationship between ability and participation in reading (Cunningham & Stanovich 1998).
We know that children who read a lot perform well in school, but are they bright because they
read, or is it simply that they read because they are bright? Reviews of the literature find
extensive evidence for an association between reading frequency and reading attainment
(Twist et al. 2007), but note the difficulty in establishing whether reading frequency actually
leads to improved attainment in the absence of compelling longitudinal evidence (Clark & De
Zoysa 2012; Clark & Rumbold 2006; Department for Education 2012; Department for
Education Education Standards Research Team 2012). While some longitudinal studies on
reading exist, they have typically been small scale, covered relatively short periods, and
lacked controls for socio-economic background (Taylor et al. 1990).
The growth in cognitive inequalities according to socioeconomic status during childhood has
been established by analyses of the British cohort studies of 1946, 1958, 1970 and 2000
(Douglas 1964; Feinstein 2003; 2004; Fogelman & Goldstein 1976; Sullivan et al. 2013), but
for the 1970 cohort this has only so far been examined up to age 10. Here we analyse the
extent to which social inequalities in the 1970 cohort continued to grow during adolescence,
and which childhood resources drive this growth. Specifically, what is the role of reading for
pleasure during childhood and adolescence? As far as we are aware, ours is the first large
study to take a longitudinal approach to reading and cognitive development during
adolescence.
Research questions
In this paper, we focus on the potential role of both parents’ and children’s reading in
explaining differentials in cognitive test scores at age 16 for a cohort of children born in
1970. The BCS70 is a large, nationally representative, longitudinal birth cohort study,
containing rich measures of both cognition and home background, which provides some
strong advantages in tackling these questions.
4
We address the following questions.
1. Are inequalities due to parental social background similar across the domains of
vocabulary and mathematics, or do they differ? We hypothesize that parental
education (which is typically seen as reflecting cultural and cognitive resources) but
not occupational social class (which is an indicator of economic position) may be
more strongly linked to vocabulary than to maths attainment.
2. Are inequalities due to parental social background explained by parental reading
environment, behaviour and ability? We hypothesise that the link between parental
education and children’s test scores is more likely to be accounted for in this way than
the link between parental social class and children’s test scores.
3. Is the influence of parental reading environment, behaviour and ability explained by
children’s own reading behaviour?
4. Which factors are linked to changing test scores between the ages of 10 and 16? In
particular, is the child’s own reading linked to cognitive progress? We hypothesise
that children’s reading should be more strongly linked than children’s music
participation to cognitive progress. We also hypothesise that childhood reading should
be more strongly linked to progress in vocabulary than to progress in maths, since
reading directly exposes children to new vocabulary. Nevertheless, we expect that
reading will develop mathematical abilities to some degree, since reading ability
facilitates learning in all subjects.
Data and variables
The 1970 British Cohort Study (BCS70) follows the lives of more than 17,000 people born in
England, Scotland and Wales in a single week of 1970 (Elliott & Shepherd 2006). Over the
course of cohort members’ lives, the BCS70 has collected information on health, physical,
educational and social development, and economic circumstances among other factors. Since
the birth survey in 1970, there have been eight surveys (or ‘waves’) at ages 5, 10, 16, 26, 30,
34, 38 and 42. An understanding of the educational progress of this cohort during their
childhood is vital to understanding their later life course trajectories.
The 1970 cohort study is rich in cognitive test scores throughout the early years, and the early
test scores (up to age ten) have been analysed extensively, including influential work by
Feinstein (2003; 2004). The cognitive scores at age ten have also been shown to be important
for adult outcomes, including in employment (Breen & Goldthorpe 2001) and health (Batty et
al. 2007). To date, almost no research has been carried out using the age 16 test scores
(though see (Duncan et al. 2012)), due to a lack of awareness of the existence of these scores
among the research community, and because the arithmetic dataset was not deposited until
2008. We aim to encourage wider use of these variables by researchers.
5
Because we exploit data from all of the childhood waves of the study, including the age 16
wave, the problem of missing data must be addressed. The age 16 survey employed sixteen
separate survey instruments, and unfortunately coincided with a teachers’ strike which
affected the completion of those instruments, including cognitive tests, that were
administered via schools (Dodgeon 2008). This led to substantial instrument non-response.
Nevertheless, the 1986 sample is more representative in terms of the birth characteristics of
the sample in 1970 than any other wave of the study excluding the birth wave (Mostafa &
Wiggins 2014). Appendix A1 shows logit response models for response to both the arithmetic
and vocabulary tests in terms of the birth characteristics of the 1970 sample. Respondents
who took both tests were no more highly selected than the 1986 sample as a whole. Levels of
missing data for the variables used in our analysis are provided in table 1. As list-wise
deletion was not a practical option, we use multiple imputation to ‘fill-in’ values of any
missing items in the variables selected for our analysis adopting Schafer’s algorithm under
the assumption of ‘missing at random’ (MAR). All reported analyses are averaged across
twenty replicates based upon Rubin’s Rule for the efficiency of estimation under a reported
degree of missingness across the whole data of just under 20% (Little & Rubin 1987). The
analytical sample consists of the 3, 583 cohort members who completed the age 16
vocabulary and maths tests.
It is important to acknowledge that people’s levels of motivation and compliance, as well as
potential stereotype-threat (Croizet & Claire 1998; Spencer et al. 1999) will affect their
scores in cognitive tests. We also acknowledge that multiple-choice tests do not capture the
full range of academic skills, and girls tend to fare worse in multiple-choice tests than in
other forms of assessment (Gipps & Murphy 1994). We do not interpret the tests used here as
providing an estimation of innate intelligence. They are simply tests of attainment based on
the capability and motivation to complete a particular task under given conditions.
Analytical Strategy
We will first present a descriptive analysis of children’s attainment trajectories. Subsequently
we will model mathematics and vocabulary scores at age 16 as a function of family
characteristics, childhood reading behaviour and children’s prior attainment scores. This will
be presented as a series of multivariate linear regression models. The outcome and
explanatory variables to be used in these analyses are described below.
Outcome variables: Mathematics and vocabulary at age 16
Influences on cognitive scores and changes in these scores may be expected to differ
according to the nature of the assessment, and the demands the assessment makes on
processing capacity or problem solving as opposed to knowledge (Richards & Sacker 2003).
Here we examine attainment scores in vocabulary and maths. Vocabulary reflects linguistic
competence, which we expect to be developed within the home and through reading rather
than primarily through schoolwork. An advantage for our purposes of the vocabulary test
6
used here is that it is purely a test of linguistic competence, with no verbal reasoning element.
The vocabulary test is purely dependent on recall, while the maths test can be seen as a test of
problem solving, although of course background knowledge is important here too.
Maths attainment was assessed using the Applied Psychology Unit (UPU) Arithmetic test - a
30 minute assessment comprising 60 multiple choice items covering arithmetic, probabilities
and area (Closs & Hutchings 1976; Dodgeon 2008). Vocabulary was assessed using a 75 item
multiple choice test. Table 1 shows descriptive statistics for the mathematics and vocabulary
tests.
TABLE 1
Explanatory variables
This section outlines the predictors to be used in the regression models presented in the
results section of this paper. The descriptives for the variables in our regression analyses are
shown in tables 2 and 3.
TABLE 2
TABLE 3
Model 1: Socio-economic background, sex and siblings
Social class is coded according to the NS-SEC schema (Goldthorpe 1997) at age ten (in
1980). NS-SEC is an occupational schema, and determines class position in terms of
employment relations. It reflects not just income, but longer term economic security, stability
and prospects, as reflected in a person’s labour market position. It also reflects power in terms
of relationships of authority, control and autonomy within the workplace (Goldthorpe &
McKnight 2006). We adopt a ‘dominance’ approach to household social class, taking the
mother or father’s occupation whichever is higher, using the three category classificationi.
Parents’ education is coded as the highest qualification of the mother or father (whichever of
the two is higher).
Position in the birth order is a well established predictor of child outcomes (Nisbet 1953).
Sex is included to test whether development in maths and vocabulary differed between boys
and girls.
Model 2: Model 1 + Home reading culture
We are able to provide a relatively thorough operationalisation of the home reading culture
compared to many previous studies. We include not just reading to the child, but also parental
reading behaviour, newspapers (broadsheet or tabloid) in the home and parental reading
ability.
7
When the cohort members were aged 5, mothers were asked on how many days of the last 7
the child had been read to. In 1986, the mother was asked whether she and her husband read
books. Seeing parents reading may affect children’s attitudes to reading, and parents’ reading
habits are also likely to be positively linked to parents’ reading ability. Mothers were also
asked which newspapers were in the home and were thus available for the teenager to read.
We are able to differentiate between broadsheet and tabloid households and those who did not
have newspapers at home. The prose style of tabloid (then as now) was simpler and geared
towards a lower reading age and smaller vocabulary than the broadsheets. During the 1980s,
newspaper readership was high, and the type of newspaper read was (as it remains) a strong
cultural identifier (Chan & Goldthorpe 2007).
In the absence of a reading assessment for parents, we rely on a self-reported measure of
reading difficulties. Mothers were asked whether they or their husband had reading
difficulties, either when learning to read or currently (maternal self-completion 1986).
Positive responses to these items are low, with 5% of mothers admitting to any difficulties for
themselves, and 4% for their husbands. Item non-response was high on these questions (6%
for mothers and 9% for fathers) and, in preliminary analyses, was found to be significantly
associated with poor performance in both assessments suggesting that those with reading
difficulties may have been embarrassed or reluctant to report this. We acknowledge also that
subjective reporting of difficulties tends to be much lower than actual tested difficulties
(Bynner & Parsons 2006). Here we use a binary variable which indicates whether the mother
reported that either she herself or the father had difficulties with reading. Although parental
reading ability and habits and reading materials available in the home were captured when the
cohort member was age 16, we consider these to be variables which would be unlikely to be
subject to significant change during the preceding years of the cohort member’s life and
therefore do not see it as problematic to treat these variables as predictors of outcomes at age
16. We acknowledge the drawback that we have no measure of books in the home, a variable
which has been shown to be a powerful predictor of children’s educational attainment (Chiu
& Chow 2010; Evans et al. 2010).
Model 3: Model 2 + Cohort member’s own reading and playing a musical instrument
The 1980 self-completion pupil questionnaire includes items on reading books and going to
the library. The 1986 cohort member self-completion questionnaires contained items on
reading books and newspapers. Book reading declined between the ages of ten and 16. Some
difference may be due to the earlier variable being reported by mothers while the later
variable is self-reported, but it is also likely that there was a genuine decline in reading
among teenagers, perhaps partly due to a lack of availability and promotion of suitable books
for this age group. For example, libraries during the 1980s typically devoted very little space
to books aimed at adolescents. This decline in reading for pleasure as children get older is in
line with previous research (Clark & Rumbold 2006).
We also include variables indicating whether the cohort member played a musical instrument
at age 5 and age 10. Playing an instrument would not be expected to directly develop skills in
maths or vocabulary, but can be seen as a more general indicator of the cultural climate of the
8
home, and arguably represents ‘status seeking’ behaviour, in the sense that parents who
encourage their children to play an instrument are complying with a strong social norm
within the educated middle-classes.
Model 4: Model 3 + Cognitive tests at five and ten
The cohort members took age-appropriate tests at age five and ten. These are included in our
final model, to assess cognitive progress between the ages of 10 and 16. Descriptive statistics
for these tests are shown in table 4. We use Principal Components Analysis (PCA) to extract a
single scale for cognition at each age (5 and 10). Full details of the tests used and the PCA
analysis are provided in Parsons (2014).
TABLE 4
Age five tests
The cohort members took five tests designed to capture verbal and non-verbal skills:
Copying designs (Rutter et al. 1970); English picture vocabulary (Brimer & Dunn 1962);
Human figure drawing (draw-a-man) (Goodenough 1926; Harris 1963); Complete a profile;
and Schonell graded reading (Golding 1975).
Age ten tests
The eight tests taken at age ten were: Shortened Edinburgh reading test (Godfrey Thompson
Unit 1978); Pictorial language comprehension test; Friendly maths test; Spelling (dictation
task); British Ability Scales (BAS) (Elliott et al. 1979; Hill 2005): Two verbal subscales
(word definitions and word similarities) and two non-verbal subscales (digit recall and
matrices) (Butler et al. 1980).
Results
We begin with a simple graphical presentation of children’s trajectories over time, before
presenting the results of our regression analyses.
Attainment trajectories
First we look at children’s progress in vocabulary and mathematics in terms of their
percentile rankings. Figure 1 shows vocabulary trajectories between the ages of five and 16
comparing frequent readers to infrequent readers and the children of at least one graduate to
those whose parents had no qualifications. Young people who read books most frequently at
10 and 16 (20%) are classified as high readers. Those who were in the lowest book reading
category at both ages or the lowest at one and second lowest at the other (21%) were
categorised as low readers.
9
FIGURE 1
High readers started out in the 57th percentile at age 5, and increased their position by 5
percentiles by age 10 and a further 7 percentiles by age 16, meaning that by age 16, the high
readers had caught up with the children of graduates. The least frequent readers saw a decline
of 8 percentiles over the same period.
The children of both graduates and parents with no qualifications maintained a roughly
constant position between the ages of 5 and 10. Interestingly, the children of parents with no
qualifications maintained a roughly constant position between 5 and 16, with a slight increase
in their mean rank at 10 and a slight decline by 16. However, the gap between the children of
graduates and the children of the unqualified grew during the secondary school years, as the
children of graduates improved their position.
FIGURE 2
Turning to children’s maths trajectories, we must first note that there was no maths test at age
5, so we use a non-verbal score (Copying Designs) as a baseline. The children of graduates
improve their position between 5 and 16 by 3 percentiles (as opposed to 5 for vocabulary).
The position of the children of parents with no qualifications barely changed between 5 and
16. The gap between frequent and infrequent readers remained fairly constant between 5 and
16.
In summary, these charts suggest that parents’ education and children’s reading make a
difference to children’s progress in both vocabulary and maths during the secondary school
years. However, the role of children’s reading appears to be particularly important for
vocabulary development. In the analyses to follow, we test whether these descriptive results
hold given an extensive set of control variables.
Multivariate Regression Analysis
We present a multivariate regression analysis (also known as multivariate general linear
model or multivariate response model), which treats the outcome variables jointly, and hence
uses only those cases with data for both test scores at age 16. This allows us to compare
model fit across the two outcomes, and to examine differences in the predictors across the
two outcomes. The dependent variables are treated as percentage test scores, in order to make
the coefficients interpretable as percentage point differences. We have run parallel analyses
(available on request) treating the dependent variables as standardised z scores, which did not
change the results in any substantial way.
TABLE 5
Table 5 shows the multivariate regression analysis results. In Model 1 we control only for sex
and family background. Model 1 shows no link between gender and either maths or
10
vocabulary scores. Parents’ qualifications are significantly linked to both test scores, with the
children of more highly educated parents achieving higher scores. Children who had a parent
with a degree scored around 13 percentage points higher than those whose parents had no
qualifications in both vocabulary and maths. The advantage associated with managerial and
professional occupations (‘the salariat’) compared to the manual/routine class was
considerably smaller, 4 percentage points for vocabulary and 5 for maths. Older siblings are
negative for both vocabulary and maths, with a 2 percentage point disadvantage for each
additional older sibling.
In model 1, we can also note that model fit is weak for both maths and vocabulary, but fit is
stronger in the case of vocabulary (R2 = 0.15) than for maths (0.10), suggesting that ascribed
social characteristics are more important predictors of variability in vocabulary than in maths.
In model 2, we introduce variables related to the home reading climate. How often the child
was read to at age five is significantly positive for both vocabulary and maths, with an
advantage of 1.2 and 0.9 percentage points respectively for each additional day the child was
read to in the last week. Parents’ reading books in their spare time is also significantly
positive forvocabulary (an advantage of 1.6 percentage points), but not for maths. Parents’
self-reported reading problems are significantly negative for both scores (-2.9 percentage
points for vocabulary, -4.9 for maths). Having broadsheet newspapers in the home is linked to
higher scores on both vocabulary (4.9) and maths (4.4), while tabloids in the home were not
significant compared to having no national newspapers in the home.
In this model, the coefficients for both parental education and socialsocial class are somewhat
reduced, but remain consistently statistically significant. In model 2, the improvement in
model fit is greatest for vocabulary (R2 increases from 0.15 to 0.20, compared to 0.10 to 0.12
for maths) suggesting that the home reading climate accounts for more of the variability in
vocabulary than for maths.
Model 3 introduces the child’s own reading behaviour. Book reading at 10 and 16 and
newspaper reading at 16 are powerfully linked to both vocabulary and maths. Reading often
at age 10 is linked to an advantage of 13 percentage points in vocabulary and 14 in maths.
Reading more than weekly at 16 is linked to a further advantage of 7 percentage points in
vocabulary and 3 in maths, while reading a newspaper more than weekly at age 16 is
associated with an additional 5 percentage points in vocabulary and 7 in maths. This model
also includes playing a musical instrument at 10 and 16. Playing an instrument is linked to
test scores in this model, but much less strongly than reading is. Playing an instrument at 10
is linked to an advantage of only 1 percentage point in vocabulary and 2 percentage points in
maths. Playing an instrument at 16 is not linked to a statistically significant advantage in
maths, but is linked to a 1.5 percentage point advantage in vocabulary.
Interestingly, gender becomes significant for both subjects in this model, with positive
coefficients for boys. This suggests that, while boys’ absolute performance was not different
from girls’ in either maths or vocabulary, boys performed at higher levels than girls for any
11
given level of recreational reading – in other words, boys performed as well as girls, despite
not reading as much as girls. The influence of variables reflecting the parents’ reading culture
is reduced in this model, suggesting that the influence of parents’ reading culture is somewhat
explained by the child’s own reading. However, the negative link between tabloid newspapers
and vocabulary (a 1 percentage point reduction in scores) becomes statistically significant in
this model.
Model 3 also shows a substantially improved model fit for vocabulary (from R2 = 0.2 to
0.32), with a smaller improvement for maths (0.12 to 0.18), suggesting that the child’s own
reading is most important for vocabulary development. In particular, the cohort member’s
reading at age 16 was more strongly linked to vocabulary than to mathematics attainment.
Model four introduces the cohort member’s test scores at the ages of five and ten. Principal
Components Analysis (PCA) is used to extract a single standardised Z score for cognition at
each age (5 and 10).. The inclusion of the age five and ten test scores in the model means that
it becomes a model of how far the predictions of model 3 had already been established by age
ten, and how far they continued to be reflected in changes in the child’s test scores between
age ten and age 16. Essentially it is a model of progress, with the proviso that the tests taken
at ages five and ten were not the same as those taken at 16, although vocabulary was
measured at both 5 and 10, and maths was also measured at age 10. Coefficients in this model
could be biased if measurement error in the cognitive tests at ages 5 and 10 are linked to other
variables of interest (Jerrim & Vignoles 2013). Therefore, we have attempted to minimise the
risk of spurious results due to measurement error in any given test by including the full set of
test scores at both ages in our analysis.
An increase of 1 standard deviation in cognitive scores at age 5 was associated with an
increase of 1 percentage point in vocabulary, but no significant gain for maths. The age 10
score was more highly predictive of attainment at 16, as would be expected. A one standard
deviation increase in age 10 scores was linked to gains of 9 percentage points in vocabulary
and 13 percentage points in maths. Model fit improves substantially, to R2 = 0.61 for
vocabulary and 0.49 for maths.
Many variables that were significant in model 3 become non-significant or marginally
significant in model 4 because they are linked to absolute attainment in the test scores at age
16, but not to progress between ten and 16. Parents’ education remains significant, but the
coefficient is much reduced in size (the advantage due to a parental degree falls from 8 to 2.5
percentage points for vocabulary and from 9 to 2 percentage points for maths). Social class is
non-significant in this model. The negative influence of elder siblings remains significant for
both maths and vocabulary, with a one percentage point disadvantage in progress in maths
and vocabulary for each additional elder sibling.
The influence of the home reading environment is much reduced, although broadsheet
newspapers in the home and having been read to at age 5 remain significantly positive for
vocabulary, though not for maths. Overall though, the influence of the home reading
12
environment has been largely mediated by the combination of the children’s own reading
behaviour and by their early attainment scores.
Importantly, the coefficients for cohort member’s own reading behaviour, including reading
books and newspapers, remain large and highly statistically significant in model 4. This
suggests that it is not just the case that academically able children read more, but that leisure
reading is linked to greater cognitive progress during the teenage years. This is in marked
contrast to playing a musical instrument, which becomes non-significant in this model, apart
from a borderline significant negative relationship between playing an instrument at age 10
and vocabulary development.
Taking the three variables reflecting childhood reading together, being in the top categories
for all three adds up to a gain of 10 percentage points in vocabulary and 5 percentage points
in maths. This compares to a difference associated with a parental degree of 2.5 percentage
points for vocabulary and 2 percentage points for maths. In other words, the influence of
reading on growth in academic attainment scores is substantial. Given that parental education
has typically been found to be the strongest determinant of children’s educational and
cognitive outcomes, the fact that childhood reading matters more than parental education for
cognitive progress during the secondary school years is striking.
Finally, in additional analysis (available on request) we tested whether there was an
interaction between childhood reading and parents’ education. This was to assess whether the
cognitive return to reading differed according to parents’ educational level. We found that
there was a positive interaction for reading at age 16 (but not age 10) only for vocabulary.
This differential may suggest that 16 year olds with more educated parents had access to
higher quality reading materials, but unfortunately we lack the data to test this hypothesis
directly.
Conclusions
Our initial descriptive analyses showed that differentials in vocabulary scores between the
children of graduates and the children of parents with no qualifications increased sharply
between the ages of ten and 16, but differentials for maths were stable. However, our
subsequent regression analyses showed that differentials in test scores at age 16 due to
parental social background did not vary greatly across the domains of mathematics and
vocabulary. We hypothesised that parental education would be more strongly linked to
vocabulary than to maths scores, but this was not the case. Rather, parental education was
much more strongly linked than parental social class to both vocabulary and maths scores.
This finding broadly supports Bourdieu’s emphasis on cultural resources, suggesting that they
matter more than economic resources, at least for cognitive outcomes.
The home reading culture, including reading to the child, parents’ reading books and
newspapers, and parental reading problems, was linked to children’s test scores, and this had
a role in mediating the influence of parents’ education, and also to some extent in mediating
13
parents’ social class. In order to interpret the influence of the home reading culture on
children’s outcomes, it is important to acknowledge that elements of this culture are likely to
be strongly related to one another. For example, a mother who struggles with reading is likely
to struggle to read to her child, and unlikely to read in her leisure time. Given the prevalence
of adult literacy problems in Britain, (National Audit Office 2008) this is likely to be an
important aspect of the educational disadvantage suffered by many children. Eight per cent of
the BCS70 cohort were assessed as having poor basic skills in literacy at age 34, with strong
evidence of intergenerational transmission of poor literacy and numeracy to their own
children (Bynner & Parsons 2006)
Children’s own reading behaviour was strongly linked to test scores in maths and vocabulary,
and this accounted for some of the influence of parents’ reading. Our findings support other
work suggesting that children’s leisure reading is important for educational attainment and
social mobility (Taylor 2011), and suggest that the mechanism for this is increased
development of cognitive skills. Once we controlled for the child’s test scores at age five and
ten, the influence of the child’s own reading remained large and statistically significant,
suggesting that the positive link between leisure reading and cognitive outcomes is not purely
due to more able children being more likely to read a lot, but that reading for pleasure is
actually linked to increased progress over time., We found a stronger link between reading
and progress in vocabulary development than in mathematics. This is in line with our
hypothesis, based on the grounds that books directly expose readers to new words, and
therefore reading should influence vocabulary directly, whereas the influence on attainment
in other areas would be indirect, via the fact that improved reading ability improves an
individual’s ability to learn across the whole curriculum. We expected this indirect influence
of reading on progress in mathematics to be weaker than the hypothesised direct effect on
vocabulary. The differential influence of reading on the two cognitive scores can therefore be
seen as tentatively supporting a causal interpretation of the role of reading in vocabulary
development. As a further check on the interpretation of the effect of reading, we included
playing a musical instrument in our models. While it could be argued that reading simply
proxies a particular cultural milieu within the family, or child characteristics such as diligence
or concentration, we would argue that the lack of a significant association between playing a
musical instrument (which should also proxy such characteristics) and cognitive progress
does not accord with this view. In summary, although statistical regression analyses can never
prove causality, we suggest that the combined evidence of temporal ordering and differential
effects in line with theory makes a causal interpretation of the role of reading more plausible
than the alternative explanations in this case.
From a policy perspective, our findings strongly back the need to support and encourage
children’s reading in their leisure time, especially given that the available evidence on trends
over time suggests that children’s reading for pleasure has declined in recent years (Clark &
Rumbold 2006). Reasons for this decline in reading for pleasure may include increases in
competing demands on young people’s time, including homework, organised activities and
the internet. However, there is scope for new technologies to be exploited to provide greater
access to books. Supporting young children in becoming confident readers is clearly
14
necessary but not sufficient to achieve the goal of encouraging reading for pleasure
throughout childhood and adolescence. Schools need to foster a love of reading as well as
teaching reading as a skill. In light of the decline in leisure reading between the ages of ten
and 16, our findings suggest the particular need to support teenagers’ reading. We would also
argue that supporting reading for pleasure among disadvantaged children could potentially
provide a powerful tool in closing education attainment gaps (Connelly et al. 2014). Further
research is needed on effective approaches to promoting reading for pleasure, but it seems
clear that library services both within and outside schools are necessary to promoting reading,
and sharing knowledge with individual young people about books they may enjoy, especially
for those children who do not live in homes that are lined with books. The lack of library
provision in many British schools is a cause for concern (APPG 2014). By definition, no one
can be forced to read for pleasure, but practices such as silent reading periods in school may
help to establish reading as a habit to be enjoyed.
In this paper, we have attempted to shed new light on educational inequalities and cultural
reproduction processes by examining the role of reading in attainment trajectories over time,
and the distinctive link between cultural resources and vocabulary. In future work, we intend
to assess the role of cognition in determining educational attainment and life chances,
including the question of whether vocabulary plays a distinctive role in determining
educational attainment and social reproduction and mobility.
15
Bibliography
APPG (2014) The beating heart of the school: Improving educational attainment through
school libraries and librarians (London, Libraries All Party Parliamentary Group).
Batty, G. D., Deary, I. J., Schoon, I. & Gale, C. R. (2007) Mental ability across childhood in
relation to risk factors for premature mortality in early life: the 1970 British Cohort Study,
Journal of Epidemiology and Community Health, 61, 977-1003.
Bourdieu, P., Passeron, J.-C. & Saint-Martin, M. (1994) Academic Discourse (Cambridge,
Polity Press).
Bourdieu, P. & Passeron, J. C. ([1977] 1990) Reproduction in education, society and culture
(London ; Beverly Hills, Sage Publications).
Breen, R. & Goldthorpe, J. H. (2001) Class, mobility and merit - The experience of two
British birth cohorts, European Sociological Review, 17(2), 81-101.
Brimer, M. A. & Dunn, L. M. (1962) English Picture Vocabulary TestEducational Evaluation
Enterprises).
Butler, B., Despotidou, S. & Shepherd, P. (1980) 1970 British Cohort Study (BCS70): Ten
Year Follow-up (London, Social Statistics Research Unit, City University).
Byford, M., Kuh, D. & Richards, M. (2012) Parenting practices and intergenerational
associations in cognitive ability, International Journal of Epidemiology, 41, 263-272.
Bynner, J. & Parsons, S. (2006) New light on literacy and numeracy (London, NRDC
Research Report).
Chan, T. W. & Goldthorpe, J. H. (2007) Social Status and Newspaper Readership, American
Journal of Sociology, 112(4), 1095-1134.
Cheung, S.-Y. & Andersen, R. (2003) Time to read: Family resources and educational
outcomes in Britain, Journal of Comparative Family Studies, 34(3), 413-433.
Chiu, M. M. & Chow, B. W. Y. (2010) Culture, motivation and reading achievement: High
school students in 41 countries, Learning and Individual Differences, 20, 579-592.
Clark, C. & De Zoysa, S. (2012) Mapping the interrelationships of reading enjoyment,
attitudes, behaviour and attainment (London, National Literacy Trust).
Clark, C. & Rumbold, K. (2006) Reading for pleasure: A research overview (London,
National Literacy Trust).
Closs, S. J. & Hutchings, M. J. (1976) APU arithmetic test (London, Hodder and Stoughton).
Connelly, R., Sullivan, A. & Jerrim, J. (2014) Primary and secondary education and poverty:
Anti poverty strategy review (London, Joseph Rowntree Foundation).
Croizet, J.-C. & Claire, T. (1998) Extending the concept of stereotype threat to social class:
The intellectual underperformance of students from low socioeconomic backgrounds,
Personality and Social Psychology Bulletin, 24(6), 588-594.
Cunningham, A. E. & Stanovich, K. E. (1998) What reading does for the mind, American
Educator, 22(1-2), 8-15.
De Graaf, N. D., De Graaf, P. M. & Kraaykamp, G. (2000) Parental Cultural Capital and
Educational Attainment in the Netherlands: A Refinement of the Cultural Capital Perspective,
Sociology of Education, 73, 92-111.
Department for Education (2012) Research Evidence on Reading for Pleasure (London, DfE
http://www.education.gov.uk/schools/teachingandlearning/pedagogy/b00192950/encouragin
g-reading-for-pleasure/what-the-research-says-on-reading-for-pleasure).
Department for Education Education Standards Research Team (2012) Research Evidence
on Reading for Pleasure (London, DfE
http://www.education.gov.uk/schools/teachingandlearning/pedagogy/b00192950/encouragin
g-reading-for-pleasure/what-the-research-says-on-reading-for-pleasure).
Dodgeon, B. (2008) Guide to the Dataset: BCS70 16 year follow up: APU arithmetic test
(London, Centre for Longitudinal Studies).
Douglas, J. W. B. (1964) The Home and the School (London, MacGibbon and Kee).
Duncan, G. J., Bergman, L., Duckworth, K., Kokko, K., Lyyra, A.-L., Metzger, M., Pulkkinen,
L. & Simonton, S. (2012) The Role of Child Skills and Behaviors in the Intergenerational
16
Transmission of Inequality: A Cross-National Study, in: J. Ermisch, M. Jäntti & T. Smeeding
(Eds) From Parents to Children: The Intergenerational Transmission of Advantage (New
York, Russell Sage Foundation).
Elliott, C. D., Murray, D. J. & Pearson, L. S. (1979) British Ability Scales (Slough, NFER).
Elliott, J. & Shepherd, P. (2006) Cohort Profile: 1970 British birth cohort (BCS70),
International Journal of Epidemiology, 35(4), 836-843.
Evans, M. D. R., Kelley, J., Sikora, J. & Treiman, D. J. (2010) Family scholarly culture and
educational success: Books and schooling in 27 nations, Research in social stratification
and mobility, 28, 171-197.
Feinstein, L. (2003) Inequality in the Early Cognitive Development of British Children in the
1970 Cohort, Economica, 70(1), 73-97.
Feinstein, L. (2004) Mobility in pupils' cognitive attainment during school life, Oxford Review
of Economic Policy, 20(2), 213-229.
Fogelman, K. R. & Goldstein, H. (1976) Social Factors Associated with Changes in
Educational Attainment between 7 and 11 Years of Age, Educational Studies, 2, 95-109.
Ganzeboom, H. (1982) Explaining Differential Participation in High-Cultural Activities - A
Confrontation of Information-Processing and Status-Seeking Theories, in: W. Raub (Ed)
Theoretical Models and Empirical Analyses: Contributions to the Explanation of Individual
Actions and Collective Phenomena (Utrecht, E.S. Publications), 186-205.
Georg, W. (2004) Cultural capital and social inequality in the life course, European
Sociological Review, 20(4), 333-344.
Gipps, C. & Murphy, P. (1994) A fair test? : assessment, achievement and equity (Milton
Keynes, Open University Press).
Godfrey Thompson Unit (1978) Edinburgh Reading Test (Sevenoaks, Hodder and
Stoughton).
Golding, J. (1975) The 1970 Birth Cohort 5-Year Follow-up: Guide to the dataset (University
of Bristol, Institute of Child Health).
Goldthorpe, J. & McKnight, A. (2006) The economic basis of social class, in: S. Morgan, D.
B. Grusky & G. S. Fields (Eds) Mobility and Inequality: Frontiers of research from sociology
and economics (Stanford, Stanford University Press).
Goldthorpe, J. H. (1997) The 'Goldthorpe' class schema: some observations on conceptual
and operational issues in relation to the ESRC review of government social classifications,
in: D. Rose & K. O'Reilly (Eds) Constructing Classes: Towards a New Social Classification
for the UK (Swindon, ESRC/ONS).
Goodenough, F. L. (1926) The measurement of intelligence by drawings (New York, World
Book Company).
Halsey, A. H., Heath, A. & Ridge, J. (1980) Origins and Destinations: family, class, and
education in modern Britain (Oxford, OUP).
Harris, D. B. (1963) Children's drawings as measures of intellectual maturity (New York,
Harcourt, Brace and World).
Hart, B. & Rinsley, T. R. (1995) Meaningful differences in the everyday experiences of young
American children (Baltimore, M.D., Paul H. Brookes).
Hill, V. (2005) Through the past darkly: A review of the British Ability Scales Second
Edition, Child and Adolescent Mental Health, 10, 87-98.
Hirsch, E. D. (1983) Cultural Literacy, The American Scholar, 52(2), 159-169.
Jaeger, M. M. (2011) Does Cultural Capital Really Affect Academic Achievement? New
evidence from combined sibling and panel data, Sociology of Education, 84(4), 281-298.
Jerrim, J. & Vignoles, V. (2013) Social mobility, regression to the mean and the cognitive
development of high ability children from disadvantaged homes, Journal of the Royal
Statistical Society (Series A), 176(4), 887-906.
Lareau, A. (2003) Unequal childhoods : class, race, and family life (Berkeley, University of
California Press).
Little, R. J. A. & Rubin, D. B. (1987) Statistical analysis with incomplete data, (New York,
Wiley).
17
Mostafa, T. & Wiggins, D. (2014) Handling attrition and non-response in the 1970 British
Cohort Study, CLS Working Paper, 2014/2.
National Audit Office (2008) Skills for Life: Progress in improving adult literacy and numeracy
(London, House of Commons Public Accounts Committee).
Nisbet, J. (1953) Family Environment and Intelligence, Eugenics Review, XLV, 31-42.
Parsons, S. (2014) Childhood cognition in the 1970 British Cohort Study, CLS Data Note.
Richards, M. & Sacker, A. (2003) Lifetime antecedents of cognitive reserve, Journal of
clinical and experimental neuropsychology, 25(5), 614-624.
Ritchie, S. J., Bates, T. C. & Plomin, R. (2014 in press) Does Learning to Read Improve
Intelligence? A Longitudinal Multivariate Analysis in Identical Twins From Age 7 to 16, Child
development.
Rutter, M., Tizard, J. & Whitmore, K. (1970) Education, Health and Behaviour (London,
Longmans).
Spencer, S. J., Steele, C. M. & Quinn, D. M. (1999) Stereotype threat and women's math
performance, Journal of experimental social psychology, 35(1), 4-28.
Sullivan, A. (2001) Cultural Capital and Educational Attainment, Sociology, 35(4), 893-912.
Sullivan, A. (2002) Bourdieu and Education: How Useful is Bourdieu's Theory for
Researchers?, Netherlands' Journal of Social Sciences, 38(2), 144-166.
Sullivan, A. (2007) Cultural Capital, Cultural Knowledge, and Ability, Sociological Research
Online, 12(6).
Sullivan, A., Ketende, S. & Joshi, H. (2013) Social class and inequalities in early cognitive
scores, Sociology, 47(6), 1187-1206.
Taylor, B. M., Frye, B. J. & Maruyama, G. M. (1990) Time spent reading and reading growth,
American Educational Research Journal, 27(2), 351-362.
Taylor, M. (2011) Life course outcomes of cultural practices -
instrumental benefits?, paper presented at British Sociological Association annual
conference (London School of Economics, London, England.
Twist, L., Schagan, I. & Hodgson, C. (2007) Progress in International Reading Literacy
Study (PIRLS): Reader and Reading National Report for England 2006 (Slough, NFER and
DCSF).
18
Tables and Figures
Table 1: Age 16 Arithmetic and Vocabulary Scores – Percentage points
N Min. Max. Mean Std.
Deviation
Arithmetic 3583 0 100 61.6 19.4
Vocabulary 3583 0 96 53.5 14.9
Table 2: Descriptive statistics for categorical regressors
Imputed
%
Original
N
%
missing
Child sex
Male 46.1 1653
Female 53.9 1930
Missing -0 0
Parental social class (NS-SEC)
Managerial / professional 29.9 1044
Intermediate 28.1 835
Routine and manual 38.5 1322
Not working 3.6 128
Missing -254 7.1
Parental highest qualification (Age 10)
No qualifications 27.1 972
Vocational/Other 19.1 684
O-Levels/A-levels 34.4 1113
Degree+ 19.3 692
Missing -122 3.4
Parental reading (Age 16 survey)
At least one parent reads books 88.7 2127
Missing -1040 29.0
Parental reading problems (Maternal
report - Age 16 survey)
At least one parent has reading problem 9.6 345
Missing -970 27.1
Reading materials available in home
(Age 16)
Tabloid paper 43.6 1562
Broadsheet paper 11.8 423
Missing -970 27.1
How often reads books (Age 10)
Often 54.9 1962
Sometimes 39.8 1034
Never or hardly ever 5.3 161
Missing -426 11.9
How often reads newspapers (Age 16)
Rarely or never 11.6 414
Less than once a week 7.1 256
19
Once a week 20.6 562
More than once a week 60.7 2130
Missing -221 6.2
How often reads books (Age 16)
Rarely or never 33.3 1192
Less than once a week 20.6 736
Once a week 16.9 472
More than once a week 29.2 967
Missing -216 6.0
Plays musical instrument
Age 10 46.5 1662
Missing -436
Age 16 24.0 861
Missing -249 6.9
Table 3: Descriptive statistics for continuous regressors
Imputed
mean
Original
N
%
missing
Birth order 1.8 3163 11.7
Number of days child read to in past
week (age 5) 4.5 2872 19.8
20
Table 4: Age 5 and 10 test scores – Imputed where missing
Original
NMin. Max. Mean. Std.
Deviation
%
missing
Age 5
Copying designs 2968 0 8 4.9 1.8 17.2%
English picture
vocabulary 2775 0 62 38.1 12.0 22.6%
Human figure
drawing 2931 1 22 10.5 2.8 18.2%
Complete-a-profile 2876 0 16 6.9 3.6 19.7%
Schonell graded
reading 2962 0 49 1.7 4.0 17.3%
Age 10
Edinburgh Reading
Test 2690 0 64 39.7 13.0 24.9%
Pictorial Language
Score 2904 24 100 63.0 9.6 19.0%
Friendly Maths Test 2690 3 72 45.9 11.2 24.9%
Spelling score 2880 0 56 36.3 9.7 19.6%
BAS word definitions 2675 0 29 10.8 4.7 25.3%
BAS word
similarities 2662 0 20 12.3 2.3 25.7%
BAS digit recall score 2671 7 34 22.5 3.8 25.5%
BAS Matrices 2666 0 28 16.2 4.9 25.6%
21
Table 5: Multivariate regression analysis of vocabulary and mathematics % scores (n for all models
=3, 583)
Model 1 Model 2 Model 3 Model 4
Maths Vocab Maths Vocab Maths Vocab Maths Vocab
Intercept 57.7** 50.9** 54.6** 46.3** 35.6** 29.6** 57.6** 46.3**
Sex (Male) 0.6 -0.1 0.8 0.2 3.4** 2.9** 0.4 0.6
Social class (Ref =
Manual/Routine)
Managerial / professional 4.8** 3.9** 3.6** 2.6** 3.2** 2.1** 0.4 -0.1
Intermediate 2.2** 2.0** 1.8* 1.5** 1.7* 1.4** 0.2 0.3
Not working 0.6 0.8 1.3 1.4 1.7 1.2 0.5 0.5
Highest parental qual (Ref=None)
Vocational/Other 3.9** 2.8** 3.4** 2.2** 3.1** 2.0** 0.9 0.3
O-Levels/A-levels 5.7** 5.6** 4.7** 4.5** 4.2** 3.9** 0.6 1.2**
Degree+ 13.2** 12.6** 10.9** 9.7** 9.4** 8.2** 2.0* 2.5**
Birthorder -2.1** -2.2** -1.8** -1.9** -1.6** -1.6** -0.8** -1.0**
Age 5 - number of days child read
to in past week 0.9** 1.2** 0.5** 0.9** -0.2 0.3**
Age 16 - Parent(s) read books 1.3 1.6* 0.5 0.7 -0.9 -0.4
Age 16 - Parental reading
problems -4.9** -2.9** -4.1** -2.2** -1.2 0.0
Age 16 - Tabloid newspaper in
home 0.2 -0.1 -0.9 -0.9* -0.3 -0.4
Age 16 - Broadsheet newspaper in
home 4.4** 4.9** 3.5** 3.7** 1.0 1.8**
Age 10 - How often reads books
(Ref = Never or hardly ever)
Often 13.9** 12.7** 2.6* 4.2**
Sometimes 10.3** 7.8** 3.4** 2.6**
Age 16 - How often reads
newspapers (Ref = Rarely or
never)
Less than once a week 3.3* 2.0* 1.5 0.5
Once a week 3.2** 1.7* 2.1* 0.8
More than once a week 6.9** 4.9** 2.8** 1.8**
Age 16 - How often reads books
(Ref = Rarely or never)
Less than once a week 2.9** 2.7** 0.7 1.0*
Once a week 1.6 3.1** 0.4 2.2**
More than once a week 3.0** 6.5** -0.4 4.0**
Age 10 - Plays musical instrument 2.3** 1.1* 0.1 -0.7
Age 16 - Plays musical instrument 1.1 1.5** -0.3 0.4
Age 5 ability score 0.5 1.0**
Age 10 ability score 12.6** 9.2**
Adjusted R20.10 0.15 0.12 0.20 0.18 0.31 0.48 0.61
22
Figure 1: Vocabulary trajectories
N=3583, data imputed where missing.
Age 5: English Picture Vocabulary; Age 10: the vocabulary subscale of the Pictorial
Language Score; Age 16: Vocabulary score.
23
Figure 2: Mathematics trajectories
N=3583, data imputed where missing.
Age 5: Copying Designs; Age 10: Friendly Maths test; Age 16: Arithmetic test.
24
Appendix Table A1: Odds ratios of Logit response models for BCS70 1986 school tests*
Gender (reference: Men)
Women 1.37*** (0.057)
Marital status (reference: Single)
Married 1.90*** (0.265)
Mother lives in London in 1970 (reference: not in London)
In London 0.60*** (0.043)
Parity (reference: 0)
1 0.83*** (0.042)
2 0.75*** (0.05)
3+ 0.55*** (0.045)
Lactation (reference: attempted)
No 0.94 (0.042)
Mother’s age at Delivery (reference: less than 20)
20-24 1.25*(0.107)
25-29 1.35*** (0.122)
30-34 1.61*** (0.163)
35 or more 1.51*** (0.187)
Mother’s age at completion of education (reference: 14 or less)
15 1.2 (0.133)
16 1.38** (0.162)
17 1.27 (0.166)
18 + 1.29 (0.166)
Father’s social class (reference: SC 1)
SC2 1 (0.104)
SC3 non-manual 1 (0.108)
SC3 manual 0.84 (0.086)
SC4 0.76*(0.087)
SC5 0.55*** (0.079)
Other 0.82 (0.116)
Father’s age at completion of education (reference: 14 or less)
15 0.99 (0.095)
16 1.04 (0.112)
17 1.31*(0.16)
18 or more 1.02 (0.116)
N15270
pseudo R20.024
Exponentiated coefficients; Standard errors in parentheses, * p < 0.05, ** p < 0.01, *** p < 0.001
*With thanks to Tarek Mostafa. This logit response model can be compared to identical models for each wave of
BCS70 in Mostafa and Wiggins (2014).
25
26
27
i We treat ‘not working’ as a separate category as NS-SEC is not available in BCS70 prior to 1980,
so non-workers could not be categorised according to their previous NS-SEC.
