Do comprehensive schools reduce social mobility?
Vikki Boliver*
Oxford Network for Social Inequality Research,
Department of Sociology and Nuffield College,
University of Oxford
[email protected]
and
Adam Swift
Centre for the Study of Social Justice,
Department of Politics and International Relations and Balliol College,
University of Oxford
[email protected]
Published in the British Journal of Sociology 62(1) 2011
Do comprehensive schools reduce social mobility?
Abstract
This paper investigates the claim that the shift from a selective to a
comprehensive school system had a deleterious effect on social mobility in
Great Britain. Using data from the National Child Development Study, we
compare the chances, for both class and income mobility, of those who
attended different kinds of school. Where media attention focuses
exclusively on the chances for upward mobility of those children from lowly
origins who were (or would have been) judged worthy of selection into a
grammar school, we offer more rounded analyses. We match respondents in a
way that helps us to distinguish those inequalities in mobility chances
that are due to differences between children from those due to differences
between the schools they attended; we look at the effects of school system
on the mobility chances of all children, not merely those from less
advantaged origins; and we compare comprehensive- and selective-system
schools, not merely comprehensive and grammar schools. After matching, we
find, first, that going to a grammar school rather than a comprehensive
does not make low-origin children more likely to be upwardly mobile but it
helps them move further if they are; second, that grammar schools do not
benefit working-class children, in terms of class mobility, more than they
benefit service-class children, but, in terms of income mobility, such
schools benefit low-income children somewhat more than they benefit higher-
income children - that benefit relating only to rather modest and limited
movements within the income distribution. Finally, however, the selective
system as a whole yields no mobility advantage of any kind to children from
any particular origins: any assistance to low-origin children provided by
grammar schools is cancelled out by the hindrance suffered by those who
attended secondary moderns. Overall, our findings suggest that
comprehensive schools were as good for mobility as the selective schools
they replaced.
Keywords: Social mobility; comprehensive, grammar, selective schools.
Do comprehensive schools reduce social mobility?[i]
Introduction
In 2005, a study carried out by economists at the London School of
Economics for the Sutton Trust became one of the most widely cited pieces
of social-scientific research in recent years. Focusing on movement
between income quartiles, it reported that 'intergenerational mobility fell
markedly over time in Britain, with there being less mobility for a cohort
of people born in 1970 compared to a cohort born in 1958' (Blanden et al.
2005a: 2).[ii] So extensively was the study reported and discussed that its
results quickly became the conventional wisdom: British society had become
less fluid, opportunities more unequal, social justice more remote.[iii]
The LSE research has attracted a good deal of criticism from sociologists
disputing its claim to have revealed a reduction in mobility.[iv] Here we
put to one side the debate about whether the alleged reduction in social
mobility was real or artificial and focus rather on a striking further
feature of the extensive media coverage: the widespread reporting of that
study as finding not merely that mobility had decreased but, also, why it
had. The culprit, apparently, was the comprehensive school. What had
changed between the two cohorts, and was the alleged cause of the reduction
in mobility, was the substantial comprehensivisation of the UK secondary
school system, as selective schools (grammar schools and secondary moderns)
were transformed into non-selective comprehensives. For the Sunday Times,
'A report from the London School of Economics (LSE) published last month
showed that the decline of grammar schools had helped deepen class
divisions, effectively kicking the ladder away from bright children.' (31
July 2005). On the same day Nick Cohen's opinion column in The Observer
mentioned the LSE research and told his readers that '…40 years of
comprehensives have left Britain a sclerotic society where parents' money
matters more than a child's talent.' (31 July 2005). Then Tim Luckhurst
informed the Times' readers that 'A Sutton Trust study for the London
School of Economics proves that comprehensives damage social mobility.' (9
August 2005).
This media coverage was particularly remarkable for the fact that the study
in question in fact made no mention of grammar or comprehensive schools!
The research is said to 'highlight how the relationship between family
income and children's higher education attainment has grown between cohorts
completing education in the late 1970s and the late 1990s' (Blanden, Gregg
and Machin 2005a: 13-4). But its authors offer no empirical evidence, nor
even any theoretical speculation, specifically about the effects on social
mobility of the shift from a selective to a comprehensive secondary school
system. The Daily Mail cites Professor Stephen Machin, one of the study's
authors, as saying: 'Of course, the grammar school system was perceived at
the time as being very elitist and a force for not being very good for
social mobility. It's rather ironic that it's actually turned out that some
kids from low income backgrounds did benefit from that system. And probably
that system got more people through from the bottom end of the system than
we currently have today' (25 April 2005). But no claim of that kind appears
in the paper and, as Jo Blanden, another of the study's authors, clarified
in a letter published by the Belfast Times (20 May 2005), 'We [the study's
authors] show that the educational outcomes and earnings of poor children
have declined relative to those better off when we compare those born in
1970 with those born in 1958. However, we do not directly assess the
extent to which this is a consequence of the decline in selective schooling
over the period. There were many changes in education and other policies
which could be responsible for the changes we find, of which the abolition
of grammar schools is just one.'
The LSE study may have done nothing either way, to bolster or cast doubt on
the claim that the move away from a selective state secondary system had
deleterious consequences for social mobility, but that claim remains
prevalent and influential. It is that claim which this paper aims to
explore, analysing one of the datasets used by the LSE study: the National
Child Development Survey (NCDS), which has tracked all children born in
Great Britain in a particular week in 1958. To our knowledge, the analyses
presented in this paper represent the first large-scale quantitative study
to focus on the implications for social mobility of the shift from a
selective to a comprehensive schooling system in Britain.[v]
It is striking to what extent criticism of the demise of educational
selection has tended to focus specifically on the suggestion that grammar
schools were ladders facilitating the upward mobility of academically able
children born into less advantaged social origins. These children in
particular, it is alleged, are the ones who have been failed by the
comprehensive school system.[vi] To be sure, there might be reasons to
attach particular weight to the mobility prospects of those children:
perhaps long-run economic productivity depends especially on realising the
talents of the relatively few but especially able children whose natural
abilities would otherwise go to waste. And one might speculate that the
talents of more able children are likely to be developed less fully in
comprehensive schools because of the possibility of a levelling effect of
mixed ability learning environments, or because of differences in school
ethos and curriculum, teacher expectations or teaching quality.[vii] These
are indeed important issues but they are by no means the only ones that
need to be addressed if we are to assess the relationship between school
system and social mobility in the round. Three points are crucial if we are
to avoid an unhelpfully limited analysis.
First, we cannot tell whether children who went to grammar schools did
better, in mobility terms, than they would have done had they gone to
comprehensives unless we take into account the characteristics of children
that affect the likelihood of their attending any particular type of
school. The whole point of the selective school system was to sort
children into those judged more or less able, and we know that other
characteristics of children (social origin, gender, region) also made a
difference to the type of school they attended. In order to identify the
distinct contribution that school type makes to children's mobility
chances, therefore, we need to 'match' respondents educated in grammar and
secondary modern schools to respondents who, given their relevant
attributes, were as likely to have attended these types of school but in
fact attended a comprehensive.
Second, those who value social mobility as a key ingredient of a socially
just society want a society in which mobility chances are more equitably
distributed between those born into different social origins. Reductions in
social mobility are bad, at least in part, because they mean an increase in
the extent to which children's social origins influence their destinations,
a tightening of the association between where children start out and where
they end up. But clearly one cannot investigate the distribution of
mobility chances by focusing on the chances enjoyed, or suffered, by
children from lowly social origins alone. We want to know how school type
affects children's mobility chances in comparison with those of others –
the impact on their relative mobility chances – so we must consider the
effect of school type on all children in the system, not just on those of
lowly social origins.
Third, even if we were interested solely in the upward social mobility of
children from lowly social origins, we could not assess the impact of
comprehensivization simply by comparing pupils attending comprehensives
with those attending grammar schools. A selective school system includes
grammar schools, into which those judged more academically able were
selected, but it also includes secondary moderns, the schools attended by
those judged lower in the distribution of academic ability. So we need to
look not only at the mobility chances of low origin children attending
grammar schools, but also at those of such children attending secondary
moderns, and compare both, taken together, to the chances of similar
children who attended comprehensives.[viii]
Data and methods
The data used in our analyses are taken from the National Childhood
Development Study (NCDS), a longitudinal study of all children born in
Britain in the week from 3rd to 9th of March 1958.[ix] To date the study
has collected data from respondents' parents and/or respondents themselves
at ages 0, 7, 11, 16, 23, 33, 42, 46 and 50. Our working sample comprises
all cases with non-missing values on all relevant variables.[x]
Descriptive statistics are presented in Table AI of the appendix.
The NCDS is the 'before' in the LSE study, the cohort that allegedly
enjoyed higher levels of social mobility than their successors born in
1970. Unfortunately, we are not able to analyse data for the 1970 cohort as
well because information on type of school attended is available for too
few respondents for a reliable statistical analysis to be carried out,
owing to the fact that schools data for this cohort was collected at a time
of industrial action, resulting in a much-reduced response rate (Dodgeon et
al. undated: 13). While looking at data for the 1958 cohort alone could be
considered problematic - because any effect on social mobility of attending
a comprehensive school might be different early on in the
comprehensivization process than later on when the comprehensive system was
better established – it is important to recognise that an equivalent
problem applies to data for the 1970 cohort: schools that continued as
grammar or secondary moderns when most others had become comprehensives
might not be representative of such schools in the heyday of the selective
system (Clifford and Heath 1984; Heath 1984).
One obvious issue concerns the children attending the various types of
school: as the 1958 cohort was entering secondary schools, very few
comprehensive schools were operating in LEAs organised on purely
comprehensive lines (Crook et al. 1999: 14; Maurin and McNally 2009: 11),
and even those that were would be likely to have experienced a 'creaming
off' of potential pupils by grammar schools in other, partially or fully
selective, LEAs (Heath and Jacobs 1999).[xi] In such circumstances, there
are bound to be worries about the extent to which some so-called
'comprehensives' did in fact educate a mixed ability, socially
representative body of pupils. This is one of the advantages of our
'matching' analyses, which attempt to correct for the fact that pupils are
not allocated at random to grammar, secondary modern and comprehensive
schools, but are instead more or less likely to attend these different
kinds of school depending on their attributes, including their measured
ability. Given this matched analysis, we believe that our analysis
provides a fair test of the issue at stake. If, as is claimed, the
introduction of comprehensive schools had a dampening effect on social
mobility, we would expect that effect to be apparent, from our data, in the
differential mobility chances of those educated, simultaneously, in these
different types of school.
We capture respondents' mobility between social origins and destinations in
two ways: firstly, following the LSE study, by exploring mobility between
income quartiles, and, secondly, by looking at mobility between social
classes, understood as groupings of occupations – a measure of social
position more commonly employed by sociologists. Income mobility and class
mobility refer to very different social phenomena and are likely to involve
different mechanisms. Class boundaries, which demarcate different kinds of
occupation on the basis of employment relationships, are more salient to
social actors than are the relatively arbitrary cut-off points
distinguishing quartiles in the distribution of income, and so access to
the service class is likely to involve a greater degree of social closure
than is access to the higher income categories. By looking at both income
and class mobility, we allow the possibility of finding that school type
plays a different role with respect to each.
Respondents' income origins are measured with reference to family income
when cohort members were aged 16. Several different components of net
family income are recorded in the NCDS data as a series of categorical
variables, and we follow the LSE researchers in assigning to each of these
components a value equal to the midpoint for similar families in the 1974
Family Expenditure Survey.[xii] We then sum these values to obtain a
single continuous measure of family income, and divide the resulting income
distribution into income quartiles. For respondents' income destinations,
we again follow the LSE study in using the NCDS's continuous measure of
respondents' gross income at age 33. As with income origin, the income
destination distribution is divided into quartiles. Respondents' class
origins are measured with reference to the occupational class of their
fathers when respondents were aged 16. Three classes are distinguished
corresponding to the reduced-form Goldthorpe schema: the service class (I,
II and IVa), the intermediate class (III, IVb-c and V) and the working
class (VI and VII). Respondents' class destinations are measured with
reference to the respondent's own occupational class at age 33. As with
class origin, three class destination categories are distinguished:
service, intermediate and working.
The type of school respondents attended is established from information on
the designation of their schools as comprehensive, grammar or secondary
modern in 1969, when respondents were aged 11. This information was
collected by a survey of Head Teachers which recorded not only the
designation of the school at the time of the survey, in 1974, when
respondents were aged 16, but also, for schools that were comprehensives at
that time, the year in which the school was established as a comprehensive,
and its designation prior to comprehensivisation. We use this information
to distinguish between comprehensive, grammar and secondary modern schools,
and to distinguish selective-system schools as a whole (that is, grammar
and secondary modern schools combined).[xiii]
The first step of our analysis involves the estimation of a series of
binary logistic regression models in which the dependent variable
distinguishes between respondents who reached a particular destination (for
example, the highest earning income quartile) and those who did not.[xiv]
In these models we use interaction terms to explore how the chances of
reaching each destination vary by type of school attended for respondents
from different origins.
In the second step, we attempt to correct for the fact the allocation of
pupils to different types of school is not random. We make this correction
by using a propensity score matching technique to estimate respondents'
probabilities of attending grammar and secondary modern schools, rather
than comprehensives, given their values on a range of attributes likely to
affect the type of school attended, specifically their social origins,
their gender, their region of residence at age 11, and their measured
ability at age 11.[xv] We then repeat our binary logistic regression
analysis controlling for these propensity scores.[xvi] By this means, we
are able to minimize (as far as is possible) any biases in our findings
resulting from the non-random allocation of individuals to different types
of school on the basis of these observed characteristics.[xvii]
Results
Table I presents the results of our binary logistic regression analyses in
relation to income mobility; Table II presents them in relation to class
mobility. The tables report, in the form of odds ratios, the chances of
reaching each of the destination categories from a given origin, for those
who attended grammar, secondary modern and selective-system schools, by
comparison with those who attended comprehensives. The cells on the main
diagonal of the table report the comparative odds of immobility for those
who attended different types of school, while the cells to the left and
right of the main diagonal report the comparative odds of downward mobility
and upward mobility respectively. Asterisks indicate that an origin-
specific odds ratio is significantly different from 1 (i.e. that absolute
mobility chances differ significantly between school types), while bold
type indicates that an origin-specific odds ratio is significantly
different from the corresponding odds ratio for those from the lowest
social origin category (i.e. that relative mobility chances differ
significantly between school types). Panel A reports results without
adjustment for selection bias, while Panel B reports results controlling
for selection bias according to social origin, gender, region of residence
and measured ability, by means of the propensity score matching methodology
outlined above.
Income mobility
Beginning with the absolute income mobility chances associated with
attending a grammar school, we see from the first cell of Table I, Panel A,
that, for those from the lowest quartile, the odds of immobility are not
significantly different for those who attended grammar schools than for
those who attended comprehensives (0.8 to 1). The off-diagonal cells in
the same row, however, indicate that, for those from the lowest quartile,
attending a grammar school rather than a comprehensive is associated with
lower odds of reaching the second quartile (0.4* to 1) and with higher odds
of reaching the third and fourth quartiles (1.7* to 1 and 2.0* to 1
respectively). As far as absolute mobility chances are concerned, low-
income origin pupils appear to reap an absolute mobility advantage from
attending grammar schools rather than comprehensives. Importantly,
however, this absolute mobility advantage is in relation to how far up the
income ladder any mobility is likely to take them, rather than whether or
not they are likely to be mobile at all.
[Table I about here]
Turning to consider relative chances of income mobility, we also see from
Panel A of Table I that, while those from the lowest quartile are more
likely to reach the highest quartile if they attended a grammar rather than
a comprehensive school, this is also the case for those from higher income
origins. For those who originated in the very highest quartile, for
example, attending a grammar school rather than a comprehensive is
associated with odds of ending up in that quartile that are essentially as
good as the corresponding odds for those from the lowest quartile (1.8* to
1 compared to 2.0* to 1). That said, those from the lowest quartile do
appear to reap a relative mobility advantage from attending a grammar
school rather than a comprehensive specifically when it comes to the
comparative chances of avoiding the second-lowest quartile, and of reaching
the second-highest quartile. For example, compared to those with origins
in the highest quartile, those who originated in the lowest quartile are
less likely to end up in the second-lowest quartile (0.4* to 1 compared to
1.0 to 1) and are more likely to end up in the second-highest quartile
(1.7* to 1 compared to 0.8 to 1). Attending a grammar school rather than a
comprehensive appears to confer on those from the lowest quartile something
of a relative mobility advantage, as well as an absolute one, with respect
to how far up the income distribution any mobility is likely to take them.
Our analyses so far simply compare the mobility chances of those who
attended different types of school without correcting for the fact that
these different types of school tended to serve different kinds of pupil.
Panel B of Table I addresses this issue by reporting results that match
individuals on key attributes relevant to selection into different types of
school. We see from Panel B that, for those from the lowest quartile, the
apparent absolute and relative mobility advantage to attending a grammar
school rather than a comprehensive is notably reduced after matching on the
propensity score. In relation to absolute mobility chances, for those from
the lowest quartile, attending a grammar school rather than a comprehensive
continues to be associated with equivalent odds of remaining in the lowest
quartile (1.0 to 1), with lower odds of reaching the second-lowest quartile
(0.5* to 1 in Panel B, compared to 0.4* to 1 in Panel A), and by
implication with higher odds of reaching the third or fourth quartiles.
However, the odds ratios for the third and fourth destination quartiles are
now substantially smaller than they were prior to matching, and are no
longer individually statistically significant. Nevertheless, in relation
to relative mobility chances, after matching, those who originated in the
lowest quartile continue to be less likely to end up in the second-lowest
quartile, and more likely to reach the second-highest quartile, compared to
those with origins in the highest quartile. These matched results suggest
that low-income origin individuals do reap a genuine grammar school
advantage with respect to both their absolute and relative mobility
chances, albeit only a modest one once we match individuals on variables
relevant to selection into different types of school.
Of course, as we argued earlier, grammar school pupils made up just one
part of the selective school system. To determine whether the selective
system as a whole is better than a comprehensive system for the mobility
chances of those from low income origins, we need to consider those chances
for grammar and secondary modern school pupils combined. When we do this,
we see from the figures in row three of Table I, Panel A, that, in relation
to absolute mobility chances, those from the lowest income origins are not
significantly more likely to reach any of the various income destination
categories if they attended selective-system schools rather than
comprehensives. Similarly, the relative mobility chances of individuals
from different income origins are not significantly different for those
attending a selective-system school compared to those who attended a
comprehensive. Moreover, Panel B of Table I indicates that, after
matching, selective-system schools considered as a whole continue to confer
no absolute or relative mobility advantage of any kind in comparison with
comprehensive schools.
Class mobility
We now turn to the results of the equivalent analysis for class mobility,
presented in Table II. Beginning with the absolute mobility chances
associated with attending a grammar school, we see from the main diagonal
cell in the first row of Table II, Panel A, that, for those from working-
class origins, the odds of immobility are significantly lower for those who
attended grammar schools than for those who attended comprehensives (0.5*
to 1). The off-diagonal cells in the same row also indicate that, for
those from working-class origins, attending a grammar school rather than a
comprehensive is associated with essentially equivalent odds of reaching
the intermediate class (0.9 to 1), but with higher odds of reaching the
service class (2.0* to 1). As far as absolute mobility chances are
concerned, working-class origin pupils appear to reap a mobility advantage
from attending grammar schools rather than comprehensives.
[Table II about here]
Turning to consider the relative mobility chances of those from different
class origins, we see that, while those from working-class origins are more
likely to attain service-class destinations if they attended a grammar
rather than a comprehensive school, this is also the case for those from
higher-class origins. For those with service-class origins, for example,
attending a grammar school rather than a comprehensive is associated with
odds of ending up in a service-class destination that are essentially as
good as the corresponding odds for those from working-class origins (1.7*
to 1 compared to 2.0* to 1). Moreover, for those from service-class
origins, attending a grammar school rather than a comprehensive is
associated with odds of avoiding a working-class destination that are
considerably better than the corresponding odds for those from working-
class origins (0.3* to 1 compared to 0.5* to 1). Thus, mobility chances are
if anything less equitably distributed, as between those from different
class origins, among those educated in grammar schools than among those
educated in comprehensives.
Turning to Panel B of Table II, we see that controlling for selection into
different types of school by means of matching on the propensity score
reduces the absolute mobility advantage that comes from attending a grammar
school rather than a comprehensive. Indeed, for those from working-class
origins, attending a grammar school rather than a comprehensive is no
longer associated with significantly lower odds of remaining in the working
class (0.9 to 1) – in other words, after matching their odds of immobility
are not significantly different for those who attended grammar schools than
for those who attended comprehensives. Grammar school attendance continues
to be associated with higher odds of reaching the service class but to a
lesser extent than before (1.4* to 1 in Panel B, compared to 2.0* to 1 in
Panel A). In respect of absolute mobility chances, then, those from
working-class origins reap a genuine benefit from attending a grammar
school rather than a comprehensive. At the same time, however, after
matching on the propensity score, attending a grammar school continues to
offer those from service-class origins a real degree of protection from
long-range downward mobility into the working class (0.5* to 1), although
this particular odds ratio is not significantly different from the
corresponding odds ratio for those from working-class origins. Overall,
then, grammar schools appear to be no more equitable than comprehensive
schools in terms of their effects on the distribution of mobility chances
between relevantly similar pupils from different class origins.
As before, however, in order to compare the selective school system with
the comprehensive, we need to consider pupils educated in grammar and
secondary modern schools in combination. When we do this, we see from the
third row of figures in Panel B of Table II that those from working-class
origins are not significantly more likely to reach any of the various class
destination categories if they attended selective-system schools rather
than comprehensives: the absolute mobility advantage enjoyed by working-
class children who attended grammar schools rather than comprehensives
seems to be cancelled out by the absolute mobility disadvantage suffered by
those who attended secondary moderns. In terms of relative mobility
chances, those from working-class origins appear to benefit less from
attending a selective-system school rather than a comprehensive, compared
to those from service-class origins, specifically when it comes to the
chances of avoiding working-class destinations (1 to 1 compared to 0.7* to
1), but this disparity disappears after matching. Once we compare like
with like, as with income, selective-system schools considered as a whole
appear to confer no significant absolute or relative class mobility
advantage of any kind on anybody.
Conclusion
Our aim in this paper has been to assess the much promulgated view that the
shift from a selective to a comprehensive school system in the UK had a
damaging effect on social mobility. Although we approach this issue in a
cross-sectional rather than a longitudinal manner, we believe our
investigation of what happened to the cohort of children born in 1958 to be
a fair test. If comprehensivisation was indeed bad for social mobility,
then presumably pupils in our sample attending comprehensives are less
likely to be socially mobile than those educated in selective schools, so
one would expect to find the implied differences in the mobility chances of
those born in the same year and educated, simultaneously, in these
different types of school.
Where much media discussion of the issue focuses exclusively on grammar
school pupils from lowly origins, with many bemoaning the introduction of
the comprehensive school as depriving academically able children of a
crucial ladder of opportunity, our analysis provides a more rounded
approach, in three ways.
First, we must be careful not to attribute to differences between schools
what are really differences between children. Even if children who went to
grammar schools did experience more upward mobility than their counterparts
in other kinds of school, that could be because they were different kinds
of children, and those same children might have done just as well in other
types of school. We thus need to control for children's attributes,
especially their measured ability. Second, if our interest in social
mobility derives from a concern with equality of opportunity between those
born into unequal starting points, then we must be concerned with the
relative chances of those born into those different origins; we cannot
confine our attention to those from less advantaged backgrounds but must
look also at the effect of school type on those from more advantaged
starting points. Third, the proper comparison, if we are considering the
relative merits of comprehensive and selective secondary education, is not
between comprehensives and grammar schools but between comprehensives, on
the one hand, and grammar schools and secondary modern schools, on the
other. We must compare school systems, not merely individual types of
school within them. Thus extended from a one-sided and incomplete analysis
to the full picture, we find little support for the claim that
comprehensive schools had a negative effect on their pupils' mobility
chances, whether those are conceived as chances of movement between income
quartiles or between social classes.
As far as the absolute mobility chances of low-origin children are
concerned, we find that, for both low-income and working-class pupils,
those who attended grammar schools did gain some kinds of advantage
compared to those schooled in comprehensives. Since grammar schools, unlike
comprehensives, were selecting those judged more able, one would expect any
such advantage to be reduced when we control for children's attributes, and
so it is. Nonetheless, some advantage persists even after matching,
suggesting that school type really did make a difference. Most interesting,
however, is the kind of difference that grammar schools made: once we
control for children's attributes, we find that attending a grammar school
as opposed to a comprehensive does nothing to increase low-origin
children's chances of being upwardly mobile. In other words, after
matching, low-income and working-class children who go to grammar schools
are just as likely to be immobile as their counterparts in comprehensives.
The effect of grammar schools is specifically to increase somewhat the
extent of the mobility experienced by those who do move up.
The previous finding was broadly the same for both income and class
mobility. However, when we widen our scope to include the effects of school
type on those from the full range of origins - thus enabling us to assess
the significance of school type for children's relative chances of mobility
– we find a modest difference between the two. As far as class mobility is
concerned, after controlling for children's attributes we find that,
compared to comprehensives, grammar schools confer no more advantage on
working-class children than they do on those from service-class origins.
Going to a grammar school protects the latter against downward mobility as
much as it assists the former in terms of upward movement. For income
mobility, on the other hand, and again comparing relevantly similar
children, we find that attending a grammar school rather than a
comprehensive yields something of a greater benefit to low-income children
than to high-income ones. Grammar schools help low-origin children in terms
of movement up the income distribution more than they protect high-origin
children against downward movement, though it is important to emphasise
that the relative advantage enjoyed by low-income children relates
specifically to the comparative chances of avoiding the second-lowest
quartile, and of reaching the second-highest quartile - they enjoy no
advantage with respect to the chances of reaching the top and avoiding the
bottom quartiles.[xviii]
But the previous comparisons, despite commanding so much media attention,
tell less than half the story. Rather than focusing on grammar schools, and
on the children judged worthy of selection into them, any proper assessment
of the effect of comprehensivization on social mobility must look at all
children, comparing comprehensives with selective-system schools as a
whole. Once we include secondary moderns in the analysis, our findings are
simple and unambiguous. Comparing relevantly similar children, selective-
system schools were no more conducive to mobility, whether upward or
downward, whether between income quartiles or class categories, than were
comprehensives. Any mobility advantage accruing to children from low-income
or working-class origins who attended grammar schools was cancelled out by
an equivalent mobility disadvantage suffered by those who went to secondary
moderns. In short, matching on relevant attributes, considering all types
of social mobility (and immobility), and keeping in view all types of
school, our findings suggest that comprehensive schools were as good for
social mobility as the selective schools they replaced.
Appendix
"Table AI. Descriptive statistics for working sample of NCDS respondents "
" "Models for income "Models for class "
" "mobility "mobility "
"Variable name "N "% "N "% "
"Income origin " " " " "
" 4th quartile "785 "23.5 "- "- "
" 3rd quartile "886 "26.6 "- "- "
" 2nd quartile "885 "26.5 "- "- "
" 1st quartile "779 "23.4 "- "- "
"Income destination " " " " "
" 4th quartile "823 "24.7 "- "- "
" 3rd quartile "827 "24.8 "- "- "
" 2nd quartile "827 "24.8 "- "- "
" 1st quartile "858 "25.7 "- "- "
"Class origin " " " " "
" Service class "- "- "1394 "29.5 "
" Intermediate class "- "- "1167 "24.7 "
" Working class "- "- "2167 "45.8 "
"Class destination " " " " "
" Service class "- "- "1701 "36.0 "
" Intermediate class "- "- "1687 "35.7 "
" Working class "- "- "1340 "28.3 "
"School type " " " " "
" Comprehensive "1138 "34.1 "1600 "33.8 "
" Grammar "868 "26.0 "1190 "25.2 "
" Secondary Modern "1329 "39.9 "1938 "41.0 "
"Gender " " " " "
" Male "1615 "48.4 "2312 "48.9 "
" Female "1720 "51.6 "2416 "51.1 "
"General ability test "Mean = 42.6"SD = 18.3 "Mean = 42.4"SD = 18.0 "
"score at 11 " " " " "
Note: Income origin and destination quartiles are not exactly 25% because
the quartiles were determined for the entire sample of respondents with non-
missing income data at age 33 and not just for the working sub-sample
(those with non-missing data on income origin, income destination and
school type).
"Table AII. Results of logistic regression models predicting selection into "
"a grammar, secondary modern, or selective-system school rather than a "
"comprehensive "
" "Models for income "Models for class "
" "Grammar "Secondary"Selective"Grammar "Secondary"Selective"
" " "Modern "-system " "Modern "-system "
"Class origin (ref = Working class) " " " " "
" Service class " " " "1.4* "1.2 "1.1 "
" Intermediate " " " "1.2 "1.0 "1.0 "
"class " " " " " " "
"Income origin (ref = 1st " " " " " "
"quartile) " " " " " "
" 4th quartile "1.4* "0.8 "1.0 " " " "
" 3rd quartile "1.4* "0.9 "1.0 " " " "
" 2nd quartile "1.1 "0.8 "0.9 " " " "
"Female "1.0 "1.1 "1.0 "0.9 "1.0 "1.0 "
"General ability "1.08* "0.99* "1.02* "1.07* "0.99 "1.02* "
"Region of residence (ref =" " " " " "
"NK) " " " " " "
" North "0.9 "0.9 "1.0 "1.1 "7.1* "1.5 "
" North West "1.5 "0.8 "1.0 "1.2 "5.4* "1.2 "
" E&W Riding "0.9 "0.4* "0.5* "5.1* "2.7 "0.7 "
" North Midlands"1.0 "1.0 "1.0 "0.9 "6.3* "1.3 "
" Midlands "1.4 "1.1 "1.3 "1.0 "7.0* "1.4 "
" East "3.0 "1.0 "1.1 "1.1 "7.0* "1.5 "
" South East "1.5 "1.1 "1.2 "1.5 "8.8* "1.9* "
" South "~ "~ "~ "~ "~ "~ "
" South West "0.9 "0.8 "0.8 "0.8* "5.1* "1.1 "
" Wales "0.3* "0.2* "0.2* "0.3* "1.0 "0.3* "
" Scotland "0.2* "0.1* "0.2* "0.2* "0.7 "0.2* "
"Interactions " " " " " " "
" East X gen. "1.0 " " " " " "
"ability " " " " " " "
" E&W Riding X gen. " " "1.0* " " "
"ability " " " " " "
" Female X service class " " "1.2 " " "
" South X gen. ability " " " "1.0* " "
" Service class X gen. " " " "1.0 " "
"ability " " " " " "
" South X service class " " " " "3.1* "
" " " " " " " "
"Chi-square "583.2 "317.6 "417.2 "800.7 "530.1 "648.3 "
"statistic " " " " " " "
"N "1,814 "2,211 "2,985 "2,539 "3,172 "4,252 "
Note: Figures in the table are odds ratios. Sample sizes are reduced as a
result of the exclusion of cases with missing values for gender and general
ability. Asterisks indicate statistically significant differences in the
odds for the category of interest as compared to the reference category (p.
< 0.05). General ability is measured on a scale ranging from zero to
eighty. Odds ratios are not reported for the South because this variable
was dropped from the model due to collinearity with other variables.
"Table AIII. Descriptive statistics for blocks balanced on the "
"propensity score "
" " "Models for income "Models for class "
" " "Propensity"Number of "Propensity"Number of "
" " "score "cases "score "cases "
"Block 1 "Comprehensive ".06 (.03) "228 ".10 (.05) "592 "
" "Grammar ".06 (.02) "19 ".11 (.05) "83 "
"Block 2 "Comprehensive ".15 (.03) "220 ".29 (.06) "399 "
" "Grammar ".16 (.03) "40 ".30 (.06) "157 "
"Block 3 "Comprehensive ".30 (.06) "258 ".44 (.03) "182 "
" "Grammar ".29 (.06) "93 ".45 (.03) "102 "
"Block 4 "Comprehensive ".50 (.06) "193 ".55 (.03) "114 "
" "Grammar ".51 (.06) "167 ".55 (.03) "132 "
"Block 5 "Comprehensive ".65 (.03) "63 ".69 (.06) "146 "
" "Grammar ".65 (.03) "138 ".71 (.06) "381 "
"Block 6 "Comprehensive ".74 (.03) "42 ".83 (.02) "22 "
" "Grammar ".75 (.03) "181 ".85 (.03) "225 "
"Block 7 "Comprehensive ".84 (.03) "26 " " "
" "Grammar ".85 (.03) "136 " " "
"Block 1 "Comprehensive ".18 (.01) "221 ".14 (.01) "79 "
" "Secondary ".18 (.02) "36 ".14 (.01) "8 "
" "modern " " " " "
"Block 2 "Comprehensive ".26 (.06) "192 ".16 (.01) "226 "
" "Secondary ".27 (.06) "83 ".16 (.01) "29 "
" "modern " " " " "
"Block 3 "Comprehensive ".53 (.06) "194 ".18 (.01) "109 "
" "Secondary ".55 (.06) "224 ".19 (.01) "46 "
" "modern " " " " "
"Block 4 "Comprehensive ".65 (.03) "431 ".25 (.07) "140 "
" "Secondary ".66 (.03) "828 ".25 (.07) "40 "
" "modern " " " " "
"Block 5 "Comprehensive " " ".44 (.03) "113 "
" "Secondary " " ".44 (.02) "101 "
" "modern " " " " "
"Block 6 "Comprehensive " " ".57 (.02) "153 "
" "Secondary " " ".57 (.02) "222 "
" "modern " " " " "
"Block 7 "Comprehensive " " ".63 (.01) "256 "
" "Secondary " " ".63 (.01) "388 "
" "modern " " " " "
"Block 8 "Comprehensive " " ".67 (.02) "246 "
" "Secondary " " ".67 (.02) "578 "
" "modern " " " " "
"Block 9 "Comprehensive " " ".72 (.02) "126 "
" "Secondary " " ".73 (.02) "285 "
" "modern " " " " "
"Block 10"Comprehensive " " ".82 (.02) "6 "
" "Secondary " " ".83 (.02) "16 "
" "modern " " " " "
"Block 1 "Comprehensive ".19 (.00) "7 ".18 (.01) "23 "
" "Selective ".19 (.00) "4 ".19 (.01) "17 "
" "system " " " " "
"Block 2 "Comprehensive ".30 (.05) "351 ".29 (.05) "470 "
" "Selective ".30 (.06) "145 ".29 (.05) "174 "
" "system " " " " "
"Block 3 "Comprehensive ".51 (.06) "104 ".52 (.06) "155 "
" "Selective ".52 (.06) "118 ".53 (.06) "186 "
" "system " " " " "
"Block 4 "Comprehensive ".66 (.03) "196 ".66 (.03) "290 "
" "Selective ".66 (.03) "373 ".66 (.03) "542 "
" "system " " " " "
"Block 5 "Comprehensive ".75 (.03) "314 ".75 (.03) "407 "
" "Selective ".75 (.03) "929 ".75 (.03) "1187 "
" "system " " " " "
"Block 6 "Comprehensive ".82 (.02) "67 ".82 (.01) "100 "
" "Selective ".82 (.02) "376 ".82 (.01) "478 "
" "system " " " " "
"Block 7 "Comprehensive " " ".87 (.01) "13 "
" "Selective " " ".87 (.01) "164 "
" "system " " " " "
"Block 8 "Comprehensive " " ".91 (.00) "1 "
" "Selective " " ".91 (.01) "45 "
" "system " " " " "
Bibliography
Atkinson, A., Gregg, P. and McConnell, B. 2006 'The Result of 11 Plus
Selection: An Investigation into Opportunities and Outcomes for Pupils in
Selective LEAs', CMPO Working Paper 06/150, University of Bristol.
Becker, S.O. and Ichino, A. 2002 'Estimation of average treatment effects
based on propensity scores', The Stata Journal 2(4): 358-377.
Blanden, J., Goodman, A., Gregg, P. and Machin, S. 2004 'Changes in
Intergenerational Mobility in Britain' in M. Corak (ed.) Generational
Income Mobility in North America and Europe, Cambridge University Press.
Blanden, J. 2005 Essays on Intergenerational Mobility and Its Variation
over Time, Place and Family Structure. PhD Thesis, University of London.
Blanden, J., Gregg, P. and Machin, S. 2005a 'Intergenerational Mobility in
Europe and North America. A report supported by the Sutton Trust', Centre
for Economic Performance, London School of Economics.
Blanden, J., Gregg, P. and Machin, S. 2005b 'Educational Inequality and
Intergenerational Mobility' in S. Machin and A. Vignoles (eds) What's the
Good of Education? Princeton University Press.
Blanden, J., Gregg, P. and Macmillan, L. 2008 'Intergenerational
Persistence in Income and Social Class: The Impact of Increased
Inequality', CMPO Working Paper 08/195. University of Bristol.
Breen, R. and Yaish, M. 2006 'Testing the Breen-Goldthorpe Model of
Educational Decision-Making' in S.L. Morgan, D.B. Grusky and G.S. Fields
(eds.) Mobility and Inequality: frontiers of research in sociology and
economics. Stanford CA: Stanford University Press.
Clifford, P. and Heath, A. 1984 'Selection Does Make a Difference', Oxford
Review of Education 10(1): 85-97.
Coe, R., Jones, K., Searly, J., Kokotsaki, D., Mohd Kisnin, A. and Skinner,
P. 2008 Evidence on the effects of selective educational systems. A report
for the Sutton Trust. CEM Centre, Durham University.
Crook, D.R., Power, S. and Whitty, G. 1999 The Grammar School Question,
London, Institute for Education.
Crook, D. 2002 'Local Authorities and Comprehensivisation in England and
Wales, 1944-1974', Oxford Review of Education 28(2&3): 247-260.
Dodgeon, B., Shepherd, P., Butler, N. and Johnson, J. Undated. 1970 British
Cohort Study Sixteen-year Survey: Head Teacher Questionnaire. Guide to
data available at the UK Data Archive. London: Centre for Longitudinal
Studies, Institute of Education, University of London.
Erikson, R. and Goldthorpe, J.H. 2010 'Has Social Mobility in Britain
Decreased? Reconciling Divergent Findings on Income and Class Mobility',
British Journal of Sociology 61(2): 211-230.
Galindo-Rueda, F. and Vignoles, A. 2005a 'The heterogeneous effect of
selection in secondary schools: Understanding the changing role of
ability', The Journal of Human Resources 40(2): 335-353.
Galindo-Rueda, F. and Vignoles, A. 2005b 'The Declining Relative Importance
of Ability in Predicting Educational Attainment', Centre for the Economics
of Education Discussion Paper.
Goldthorpe, J. and Jackson, M. 2007 'Intergenerational class mobility in
contemporary Britain: political concerns and empirical findings', British
Journal of Sociology 58(4): 525-546.
Goldthorpe, J. and Mills, C. 2008 'Trends in Intergenerational Class
Mobility in Modern Britain: Evidence from National Surveys, 1972-2005',
National Institute Economic Review 205: 83-100.
Goodhart, D. 2008 'More Mobile Than We Think', Prospect Magazine 20
December 2008.
Gorard, S. 2008 'A re-consideration of rates of 'social mobility' in
Britain: or why research impact is not always a good thing', British
Journal of Sociology of Education 29(3): 317-324.
Gray, J., McPherson, A.F., and Raffe, D. 1983 Reconstructions of Secondary
Education: Theory, Myth and Practice since the War. London: Routledge and
Kegan Paul.
Gray, J., Jesson, D. and Jones, B. 1984 'Predicting differences in
examination results between local education authorities: Does school
organisation matter?', Oxford Review of Education 10(1): 69-74.
Halsey, A.H., Heath, A.F. and Ridge, J.M. 1980 Origins and Destinations:
Family, Class and Education in Modern Britain. Oxford: Clarendon Press.
Hawkes, D. and Plewis, I. 2006 'Modelling Non-response in the National
Child Development Study', Journal of the Royal Statistical Society, Series
A, 169: 479-91.
Heath, A. 1984 'In Defence of Comprehensive Schools', Oxford Review of
Education 10(1): 115-123.
Heath, A. and Jacobs, S. 1999 'Comprehensive Reform in Britain', in
Leschinsky, A. and Mayer, K-U. (eds) The Comprehensive School Experiment
Revisited: Evidence from Western Europe. Frankfurt: Peter Lang.
Jesson, D. 2000 'The Comparative Evaluation of GCSE Value-Added Performance
by Type of School and LEA', Discussion Papers in Economics 2000/52, Centre
for Performance Evaluation and Resource Management, University of York.
Kerckhoff, A.C., Fogelman, K., Crook, D. and Reeder, D. 1996 Going
Comprehensive in England and Wales: A Study of Uneven Change. London:
Woburn Press.
Marks, J., Cox, C. and Pomian-Srzednicki, M. 1983 Standards in English
Schools. London: National Council for Educational Standards.
Maurin, E. and McNally, S. 2009 'The Consequences of Ability Tracking for
Future Outcomes', Centre for the Economics of Education, London School of
Economics.
Nathan, G. 1999 A Review of Sample Attrition and Representativeness in
Three Longitudinal Surveys. London: Governmental Statistical Service
Methodology Series, 13.
Pischke, J-S. and Manning, A. 2006 'Comprehensive versus Selective
Schooling in England in Wales: What Do We Know?', Discussion Paper No.
12176, National Bureau of Economic Research.
Rosenbaum, P.R. and Rubin, D.B. 1983 'The central role of the propensity
score in observational studies for causal effects', Biometrika, 70(1): 41-
55.
Steedman, J. 1983 Examination Results in Selective and Non-Selective
Schools. London: National Children's Bureau.
"Table I. Odds of reaching each income destination from each income "
"origin if attended a grammar, secondary modern or selective system "
"school rather than a comprehensive "
"Panel A. No controls "
" "1st "2nd "3rd "4th "
" "quartile "quartile "quartile "quartile "
" "destination"destination"destination"destination"
"1st (lowest) quartile " " " " "
"origin " " " " "
" Grammar school "0.8 "0.4* "1.7* "2.0* "
" Secondary modern "1.2 "0.9 "0.8 "1.1 "
"school " " " " "
" Selective system "1.1 "0.7 "1.0 "1.4 "
"school " " " " "
"2nd quartile origin " " " " "
" Grammar school "0.7* "0.8 "0.8 "2.3* "
" Secondary modern "1.4 "1.3 "0.8 "1.0 "
"school " " " " "
" Selective system "0.9 "1.1 "0.8 "1.3 "
"school " " " " "
"3rd quartile origin " " " " "
" Grammar school "0.8 "0.7 "1.1 "1.5* "
" Secondary modern "1.3 "1.0 "0.9 "0.9 "
"school " " " " "
" Selective system "1.1 "0.9 "1.0 "1.1 "
"school " " " " "
"4th (highest) quartile " " " " "
"origin " " " " "
" Grammar school "0.5* "1.0 "0.8 "1.8* "
" Secondary modern "1.3 "1.2 "0.6* "1.0 "
"school " " " " "
" Selective system "0.9 "1.1 "0.7* "1.4* "
"school " " " " "
"Panel B. Controlling for propensity scores "
" "1st "2nd "3rd "4th "
" "quartile "quartile "quartile "quartile "
" "destination"destination"destination"destination"
"1st (lowest) quartile " " " " "
"origin " " " " "
" Grammar school "1.0 "0.5* "1.4 "1.3 "
" Secondary modern "1.2 "1.0 "0.8 "1.1 "
"school " " " " "
" Selective system "1.1 "0.8 "1.0 "1.2 "
"school " " " " "
"2nd quartile origin " " " " "
" Grammar school "0.9 "1.1 "0.8 "1.5 "
" Secondary modern "0.9 "1.3 "0.9 "1.0 "
"school " " " " "
" Selective system "0.9 "1.2 "0.9 "1.2 "
"school " " " " "
"3rd quartile origin " " " " "
" Grammar school "1.0 "1.1 "0.9 "1.1 "
" Secondary modern "1.0 "1.1 "1.0 "0.9 "
"school " " " " "
" Selective system "1.0 "1.1 "1.0 "0.9 "
"school " " " " "
"4th (highest) quartile " " " " "
"origin " " " " "
" Grammar school "0.8 "1.5 "0.7 "1.1 "
" Secondary modern "1.2 "1.3 "0.7 "1.0 "
"school " " " " "
" Selective system "1.0 "1.3 "0.7 "1.1 "
"school " " " " "
Note: Figures in the table are odds ratios. Asterisks indicate a
statistically significant difference in the odds of reaching the stated
destination from the stated origin in comparison with those who attended
comprehensive schools (p. < 0.05). Bold type indicates a statistically
significantly difference between the odds ratio concerned and the
corresponding odds ratio for those from the lowest income origin quartile
(p. < 0.05).
"Table II. Odds of reaching each class destination from each class "
"origin if attended a grammar, secondary modern or selective system "
"school rather than a comprehensive "
"Panel A. No controls "
" "Working class "Intermediate "Service class "
" "destination "class "destination "
" " "destination " "
"Working class origin " " " "
" Grammar "0.5* "0.9 "2.0* "
" Secondary modern "1.2* "1.1 "0.7* "
" Selective system "1.0 "1.0 "1.0 "
"Intermediate class " " " "
"origin " " " "
" Grammar "0.6* "0.6* "2.3* "
" Secondary modern "1.2 "0.9 "1.0 "
" Selective system "1.0 "0.8* "1.3* "
"Service class origin " " " "
" Grammar "0.3* "1.0 "1.7* "
" Secondary modern "1.2 "1.5* "0.6* "
" Selective system "0.7* "1.2 "1.1 "
"Panel B. Controlling for propensity scores "
" "Working class "Intermediate "Service class "
" "destination "class "destination "
" " "destination " "
"Working class origin " " " "
" Grammar "0.9 "0.9 "1.4* "
" Secondary modern "1.3* "1.0 "0.7* "
" Selective system "1.2 "0.9 "0.9 "
"Intermediate class " " " "
"origin " " " "
" Grammar "1.0 "0.7 "1.4 "
" Secondary modern "1.3 "0.7* "1.1 "
" Selective system "1.2 "0.7* "1.2 "
"Service class origin " " " "
" Grammar "0.5* "1.2 "1.0 "
" Secondary modern "1.2 "1.3 "0.7* "
" Selective system "1.0 "1.2 "0.8 "
Note: Figures in the table are odds ratios. Asterisks indicate that the
odds of reaching the stated destination from the stated origin are
statistically significantly different from the corresponding odds for those
who attended comprehensive schools (p. < 0.05). Bold type indicates that
the odds ratio is statistically significantly different from the
corresponding odds ratio for those from working class origins (p. < 0.05).
-----------------------
Notes
[i]. For prompting the paper, we thank Tim Wilson. For feedback on earlier
drafts of this paper, we are grateful to Jo Blanden, John Goldthorpe,
Stephen Gorard, A.H. Halsey, Michelle Jackson, Susan McNally, Colin Mills,
Malcolm Parkes, Alice Sullivan, Anna Vignoles, Anna Zimdars and five
anonymous referees. We also thank the participants at the 2009 British
Educational Research Association conference; the Sociology Department
seminar series at the University of Oxford; and the Quantitative Social
Science Department seminar series at the Institute of Education. Vikki
Boliver gratefully acknowledges the support of a British Academy
Postdoctoral Research Fellowship which she held in the Sociology
Department, University of Oxford.
[ii]. See also Blanden et al. (2004); Blanden (2005); and Blanden et al.
(2005b).
[iii]. According to David Goodhart, editor of Prospect Magazine, 'This
slender analysis has, arguably, had more influence on public debate than
any academic paper of the past 20 years. Every commentator and politician
who 'knows' that mobility has fallen off a cliff in recent years is almost
certainly basing his or her assumption on the Sutton Trust report'
(December 2008, p.46).
[iv]. For the finding that there was no change in class mobility (as
opposed to income mobility) between the two cohorts, see Goldthorpe and
Jackson (2007) and Goldthorpe and Mills (2008); for critique of the income
measures used in the LSE study, see Erikson and Goldthorpe (2010); for work
attributing the difference in findings to an increase in within-class
income differences, see Blanden, Gregg and Macmillan (2008); and for a
sharply critical discussion of the methodology used by the LSE team, and a
reanalysis of the same data yielding different results, see Gorard (2008).
[v]. There is, of course, a substantial literature on the implications of
selective and comprehensive schooling for educational outcomes rather than
social mobility. For a useful review of these studies, see also Coe et al.
(2008).
[vi]. To take two recent examples, Max Hastings in the Daily Mail (10
February 2009) writes that: 'In the Fifties, a golden age of opportunity,
almost 40 per cent of those born to parents in the lowest social income
groups grew up to join higher earners. By 1970 and ever since, only one-
third achieved this. It cannot be coincidence that, in between, Harold
Wilson's government abolished grammar schools.' While in the Daily
Telegraph (30 July 2009) Jeff Randall criticizes Conservative leader David
Cameron for lacking 'the will to admit that grammar schools did more for
working-class children than a thousand free school meals'.
[vii]. For critical discussion of such issues, see Gray et al. (1983).
[viii]. Of course, much research on educational attainment has recognised
the importance of widening the focus beyond the achievements of grammar
school students to encompass the effects of selection overall. Most studies
report no substantive differences between comprehensive- and selective-
system schools (Steedman 1983; Gray et al. 1984; Kerckhoff et al. 1996;
Heath and Jacobs 1999; Galindo-Rueda and Vignoles 2005a; Atkinson et al.
2006). This could be, as some studies suggest, because two countervailing
processes tend to cancel each other out: grammar school pupils outperform
their comprehensively educated counterparts, while the latter do better
than those educated in secondary modern schools (Marks et al. 1983; Jesson
2000). On the other hand, the findings of two recent studies suggest that
students' test scores at age 16 tend to be higher in comprehensives than in
selective system schools (Jesson 2000, Maurin and McNally 2009: 20).
Further research suggests that comprehensives improve achievement for lower
ability students (Gray et al. 1983; Kerckhoff et al. 1996) while selective
system schools improve achievement for higher ability students (Galindo-
Rueda and Vignoles 2005a, 2005b); but see Maurin and McNally (2009).
[ix]. We restrict our sample to those who attended state schools. While the
availability of independent schools might make a difference to educational
choices in a way that would interact with the comprehensive versus
selective issue (e.g. if parents in comprehensive local education
authorities were more likely than those in selective ones to choose an
independent school for their children), our controlling for children's
attributes means that we are properly comparing like with like among those
in state schools.
[x]. Our use of the NCDS, as with any longitudinal data source, raises
questions about bias due to sample attrition and other forms of non-
response (Gorard 2008). However, a number of studies investigating sample
attrition in the NCDS give grounds for confidence that the sample continues
to be representative of the original population (Nathan 1999; Hawkes and
Plewis 2006). Nevertheless, the sub-samples we use for our analyses are
relatively small: 3,335 cases, with cells ranging in size from 21 to 126
for our analyses of income mobility, and 4,728 cases, with cells ranging in
size from 33 to 424, for our analyses of class mobility. As a result, some
of our estimates are likely to have comparatively wide confidence
intervals, one implication of which is fewer statistically significant
results.
[xi]. A further issue concerns the heterogeneity, at this time, of our
category of 'comprehensives'. This groups together not only purpose-built
comprehensives and amalgamations but also former secondary moderns and
grammar schools, which will surely have differed with respect to ethos,
teaching staff, and the like. The fact that secondary modern schools tended
to become comprehensives earlier than grammar schools might bias our
findings, if, as one might expect, former secondary moderns were worse, in
terms of their effects on mobility, than former grammar schools. If
anything, then, our findings might be expected to do comprehensive schools
a disservice.
[xii]. This procedure is described in Blanden (2005). We thank Jo Blanden
for providing us with the syntax needed to replicate the income measures
used in the LSE study.
[xiii]. Nearly forty percent of respondents in our working sample who
attended selective system schools were grammar school pupils. In contrast,
in the heyday of the selective system, only around a quarter of all pupils
nationally attended grammar schools (Halsey et al. 1980; Crook et al. 1999;
Maurin and McNally 2009). This 'over-representation' of grammar school
pupils in our sample partly reflects the fact that grammar schools were
slower than secondary modern schools to take on comprehensive school status
(Crook 2002). The findings we report for selective-system schools use the
40:60 ratio of grammar schools to secondary modern schools found in the
data. However, we have checked the sensitivity of our findings for
selective-system schools as a whole to this over-representation of grammar
school pupils in our sample by re-running all our models on data weighted
to simulate a 25:75 ratio of grammar school to secondary modern school
pupils. The results (available on request) show that our findings are
robust to this alternative specification of the ratio.
[xiv]. We use binary models in preference to ordinal or multinomial models
because we wish to estimate the odds of obtaining a given destination
instead of any other destination, rather than the odds of obtaining a given
destination instead of some specific destination arbitrarily chosen to be
the reference category.
[xv]. The models used to generate the propensity scores are reported in
Table AII of the appendix.
[xvi]. The propensity scores are included in our binary logistic regression
models as a series of categorical variables or 'blocks' generated using the
pscore command in Stata (see Becker and Ichino 2002). Each block represents
a specific range of values on the propensity score, and each block is
'balanced' in the sense that, within a block, the grammar/secondary modern
school population does not differ significantly from the comprehensive
school population with regard to its mean propensity score; its social
origins, gender, and region distributions; and its mean general ability
test score. Within each block there is 'common support'; that is to say,
each block contains both grammar/secondary modern school cases and
comprehensive school cases, making it possible to compare the two kinds of
schools across the entire range of propensity scores. For a more formal
explanation of the statistical theory behind the propensity score matching
approach, see Rosenbaum and Ruben (1983). For some descriptive statistics
for our propensity score matched sample, see Table AIII of the appendix.
[xvii]. In a recent paper, Pischke and Manning (2006) question the adequacy
of attempts made in previous studies to eliminate selection bias when
modelling the effects of school type on educational attainment. The
authors argue that neither a value-added approach, which controls for the
prior attainments of pupils in the conventional way, nor an instrumental
variable approach, such as using political party in control of the Local
Education Authority to capture the likelihood of early comprehensivisation,
is likely to control adequately for selection bias. They show that even
under value-added and instrumental variable model specifications, type of
secondary school attended remains a strong predictor of pupils' attainments
prior to entering secondary school, a finding which they argue must be due
in large part to the continued presence of selection bias. However we
specify our models, we can be confident that we have not succeeded in
entirely eliminating selection effects. Still, we believe that our
analyses control for several of the most important sources of selection
into different types of school, and report our results for the propensity
score matched sample with the caveat that we claim only to have eliminated
selection bias on the basis of the pupil attributes we have named.
[xviii]. Part of the explanation for this modest difference may be that
income quartiles are somewhat arbitrary divisions, lacking the social
salience and significance of our class categories which are defined on the
basis of occupational characteristics; this is especially true for the two
middle quartiles. The mechanisms through which individuals come to occupy
different kinds of job, on the one hand, and different positions in the
distribution of income, on the other, are somewhat different. It is
plausible that the type of school attended - which itself has social
salience, is attended by status considerations, and would significantly
influence a pupil's range of personal contacts (for example, via 'old boys'
networks) - should play a somewhat different role in each case. Another
explanation would appeal to more general defects of 'one-shot' income
measures of income of the kind in the NCDS data. For the claim that class
position is a better proxy for longer-term income prospects than such
measures, see Erikson and Goldthorpe (2010).
