Routes of success

Text-only Preview

Rod Bond and Peter Saunders
Routes of success: in uences on the occupational
attainment of young British males
Using data from the National Child Development Study, the paper develops a
complex path model predicting the occupational grade achieved by 4,298
employed British males at age 33. Most British social mobility research has been
based in the ‘class structurationist’ tradition, and the paper begins by comparing
this with the ‘status attainment’ tradition, which is more common in the USA.
The class structurationist approach has rarely analysed the factors in uencing
individual occupational attainment, and those working in this tradition in Britain
have often assumed that people from working-class origins fare worse on average
than those from the middle class because of factors associated with their class dis-
advantage rather than any difference in individual characteristics such as ability
or ambition. Status attainment research, however, has generally found that indi-
vidual ability and motivation are the key factors in uencing occupational attain-
ment, and that class origins count for comparatively little. Using various measures
of class origins, parental support, quali cations, and individual ability and am-
bition, the paper goes on to develop a linear structural equations model which
achieves a good t to the data. The model demonstrates that individual ability is
by far the strongest in uence on occupational achievement, that motivation is
also important, and that factors like class background and parental support, while
signi cant, are relatively much weaker. The paper concludes that occupational
selection in Britain appears to take place largely on meritocratic principles.
KEYWORDS: Social mobility; status attainment; meritocracy; Britain; social
class; IQ
In a recent speech, the British Prime Minister, Tony Blair, declared ‘The
Britain of the elite is over. The new Britain is a meritocracy where we break
down the barriers of class, religion, race and culture’ (Daily Telegraph, 25
October 1997: 6). Like John Major before him, Mr. Blair is wedded to a
vision of Britain as a ‘meritocracy’.
The de ning feature of a meritocratic society is that occupational selec-
tion is based solely on individual ability and effort. In a meritocracy, social
origins are irrelevant in determining social destinations (Young 1958). At
British Journal of Sociology Vol. 50 No. 2 ( June 1999) pp. 217–249
ISSN 0007–1315 © London School of Economics 1999

Rod Bond and Peter Saunders
rst glance, contemporary Britain seems to fall a long way short of this ideal,
for we know from many years of social mobility research that social class
origins are still signi cantly associated with social class destinations in this
country. In a 1972 survey of ten thousand men, John Goldthorpe found
that those from ‘service-class’ backgrounds (i.e. the sons of professional,
managerial and higher administrative grade fathers) were about four times
more likely to end up in service-class positions than those who were born
and raised in working-class homes. This disparity may have lessened some-
what since then – in a 1983 follow-up, Goldthorpe found that it had fallen
to around 3: 1 (Goldthorpe 1987), and in research on a 1958 birth cohort,
one of us has reported that those born to class I/II parents were just over
twice as likely to end up in the middle class as those born to class IV/V
parents (Saunders 1997) – but there clearly is a persisting association
between class origins and class destinations.
Sociologists (including Goldthorpe) have generally assumed that much
or all of this association is explained by the social advantages and dis-
advantages owing from the operation of the class system itself. This argu-
ment, which we term the ‘SAD thesis’, has a long sociological lineage, for
down the years, many class-based factors have been identi ed as signi cant
in generating middle-class ‘overachievement’ and working-class ‘under-
achievement’. These include the different endowments of ‘cultural capital’
in middle-class and working-class homes (Bourdieu 1974), the susceptibil-
ity of working-class boys to anti-school peer group pressures (Willis 1977),
the supposed bias in the educational system favouring middle-class linguis-
tic codes (Bernstein 1965), the operation of streaming and setting as dis-
guised mechanisms of social selection within the education system
(Abraham 1995), and the impact of social and physical deprivation in the
home, such as lack of parental support and ambition, or physical over-
crowding (Douglas et al. 1968).
There is, however, a different but equally plausible explanation for why
those from middle-class origins tend to outperform those from lower-class
origins. It is possible that on average, middle-class children have higher
ability and/or exhibit a higher level of effort. This alternative explanation,
which is consistent with the operation of a ‘meritocracy’, would suggest that
individuals selected for positions on the basis of their ability and hard work
in one generation tend to produce children with similar levels of ability and
motivation who will then be selected for similar occupational positions in
the next generation. Seen in this way, the association between class origins
and class destinations is a function of the mediating effect of class differ-
ences in individual ability and motivation which are to some extent trans-
mitted from one generation to the next.
Of course, the SAD thesis and the meritocracy thesis are not necessarily
mutually exclusive. It is quite possible that both social advantages and dis-
advantages and individual ability and motivation play a part in in uencing
where people end up in the class system. An obvious problem, however, is
that the different measures associated with each of these two theses are

Routes of success
closely interconnected, and this can make it dif cult to disentangle their
relative effects.
It was in order to unpack such complex associations that Blau and
Duncan (1967) pioneered the use of path models in their classic work on
social mobility in the USA. Using regression-based path coef cients, they
sought to identify the relative importance of different factors in explaining
how individuals come to attain different occupational positions (an
approach which has come to be known as the ‘status attainment’ tradition in
social mobility research). This tradition of work contrasts sharply with that
developed in Britain from the 1970s onwards by researchers such as
Goldthorpe, where the principal concern was to assess the extent to which
social mobility contributes to or undermines the structural cohesion of
different social classes over time. This focus has little to do with status attain-
ment, but is rather addressed to the problem of ‘class structuration’, and
those working in this tradition have tended to be critical of the methodol-
ogy employed in the status attainment approach.
These two approaches in social mobility research have tended to differ, not
only in the questions they seek to address, but also in the way they measure
key concepts and the methods they employ to analyse their evidence. The
status attainment approach has generally taken as its dependent variable
some measure of occupational status organized as a graded hierarchy, and
has tried to predict individuals’ positions on this hierarchy with reference
to various attributes such as their socio-economic background (e.g. parents’
occupation and education levels), their education and quali cations, and
their measured intellectual ability. This has entailed the construction of
path models in which the standardized effects on occupational status of the
various independent and mediating variables are computed, and the overall
model t is assessed according to the proportion of variance in occu-
pational status that is explained by all the variables in the model.
In their original work, for example, Blau and Duncan (1967) managed
to explain more than one-third of the variance in individuals’ occupational
status scores with reference to their IQ scores, their education and their
socio-economic background, and they showed that individual ability
(measured by IQ) outweighed the in uence of social origins or education
in in uencing occupational attainment. Subsequent work on status attain-
ment, both in Britain and America, has tended to con rm (a) that indi-
vidual ability, measured by IQ, is an important in uence on occupational
attainment, (b) that factors such as class origin also exert an independent,
though generally less powerful, in uence, and (c) that much of the vari-
ance in occupational destinations nevertheless remains unexplained by
either individual or class background variables (we review some of the
British work below).

Rod Bond and Peter Saunders
In contrast with the status attainment tradition, the class structuration
approach rejects the use of occupational scales, preferring instead to use a
categorical class schema such as that devised by Goldthorpe himself. Rather
than ranging individuals on a continuous scale of occupational prestige,
this approach focuses on the structured system of social relations which
nds expression in clusters of occupations which are said to constitute
common class positions. In place of occupational status, the key concept is
that of social class understood in neo-Weberian terms as a common market
situation in which life chances are determined by ownership of property
and/or the exercise of authority and autonomy in the workplace. Clear
boundary lines are drawn between the various classes, and social mobility
is de ned not as movement along a continuous scale, but as movement
between discrete class categories (speci cally, into and out of the ‘service
class’). This approach has paid little attention to the issue of why and how
particular individuals end up where they do in this system, and it has there-
fore had no use for path models.1 Instead, it has used loglinear models,
based on computation of odds ratios, to investigate the relative degree of
stability or movement between different class categories while controlling
for changes over time in the proportions of the population occupying each
category. In this way, Goldthorpe and others have been able to demonstrate
that relative social mobility rates appear fairly constant over time and do
not differ markedly between different advanced societies.
For thirty years, these two traditions of work in social mobility have fol-
lowed distinct trajectories. Addressing different questions, they have devel-
oped different methods and have employed different concepts and
different theories. In principle, they are not incompatible, and some
research (such as Hope’s 1984 work on a Scottish data set, or Ishida’s 1993
comparative analysis of social mobility in Japan, Britain and the USA) has
utilized both approaches. More often, however, they have been seen as com-
peting, even as mutually exclusive, and since the 1970s, the status attain-
ment approach in particular has been roundly attacked for what its critics
see as its fundamental methodological and theoretical aws.
Although most work in the status attainment tradition has been carried out
in the USA, a number of studies have been completed on British data.
Kerckhoff (1974) used the 1946 birth cohort study originally analysed by
Douglas et al. (1968) to analyse educational attainment in England and to
compare it with the USA. He developed models which explained between
one-third and one-half of the variance and which showed similar results for
the two countries, with ability in uencing children’s success much more
strongly than father’s education, father’s occupation or family size.
Psacharopoulos (1977) used the General Household Survey to produce a
path model predicting earnings of employed adult males. Again, the model

Routes of success
accounted for around one-third of the variance in the dependent variable,
and again it showed that class background (measured by father’s occu-
pation) was a relatively weak predictor variable, although in this study,
ability (measured indirectly by exam results and type of schooling rather
than by IQ) seemed to add little explanatory power once years of school-
ing had been entered into the model.2
Work in this tradition continued in the 1980s with Mayhew and
Rosewell’s 1981 analysis based on Goldthorpe’s Nuf eld data, and with
Hope’s 1984 Scottish study. Both of these took the Hope–Goldthorpe occu-
pational prestige scale as their dependent variable, and both found that
education and parental background had a signi cant, but not large, effect.
Mayhew and Rosewell concluded that ‘other factors’ such as individual
motivation and effort might prove more important if they were included in
future models, and Hope’s study (based on data from the Scottish mental
survey of 1947) showed that this was indeed the case. Ability (measured by
IQ) was, he found, much more important than class background in in u-
encing occupational attainment, and individual effort (assessed by teacher
ratings) was also found to have a signi cant effect. Material deprivation
during childhood had no effect at all once social class was taken into
account. Hope concluded that around 60 per cent of the variance in social
mobility is explained by ability alone, and he argued that the selective
system of education operating after the war had brought Scotland close to
the achievement of meritocracy (or what he called a ‘meritelective’ system).
Research demonstrating the importance of individual ability and effort
in in uencing occupational attainment has continued to be published in
the 1990s. Kerckhoff (1990) analysed data from the fourth sweep of the
National Child Development Study (NCDS) and showed that ability and
socio-economic background both had signi cant effects on quali cations
gained at school, and that these quali cations were the crucial in uence
on the status of the rst job entered after leaving school. He also showed
that upward mobility between the rst job and the job held at age 23 (the
cut-off point for sweep 4 of this continuing panel study) was associated with
IQ and teacher ratings while at school, thus demonstrating that ability con-
tinues to exert an in uence on occupational attainment even after entry
into the labour market. Meanwhile, Ishida (1993) has analysed the Nuf eld
mobility data using path models and shows that education and class back-
ground (especially family income) both in uence occupational attainment
in Britain, the former slightly more than the latter. He suggests that class
background affects educational attainment independently of IQ differ-
ences, but since his data set has no measure of ability or cognitive skills, this
claim amounts to little more than speculation.
It will be clear from this brief review that the status attainment approach,
pioneered in the USA a third of a century ago by Blau and Duncan, has suc-
cessfully been employed in a number of different studies in Britain, and
that when it has been used, it has tended to point to the importance of
ability and effort in in uencing class destinations. In other words, most of

Rod Bond and Peter Saunders
the work in the status attainment tradition seems to indicate that occu-
pational selection in contemporary Britain is to a considerable degree mer-
itocratic (although all studies also nd that class origins do still have some
effect). These ndings have not, however, had much impact on mainstream
sociological writing about social mobility in Britain which has been domi-
nated since the 1970s by the class structuration approach.
This approach (e.g. Goldthorpe 1987, Marshall et al. 1988) has never
gathered information on the intellectual ability or level of motivation of
individuals who move up or down the class system,3 and it has therefore
never been in a position to evaluate the competing claims of the SAD and
meritocracy theses. Nevertheless, key gures in this tradition have con-
tinued to deny the validity of the meritocracy thesis. Emphasizing the dis-
parities in the chances of children from different backgrounds achieving
occupational success, they assert that the differences must be due to the
operation of class barriers favouring the children of the middle class and
blocking the children of the working class. As Payne (1987) has noted, this
is an assertion which accords with the assumptions and ideological preju-
dices of most British sociologists, and this probably explains why these
unsubstantiated claims have rarely been challenged. Instead, they are
simply recycled uncritically in sociological textbooks on class and inequal-
ity (see, for example, Scase 1992).
There are three main reasons why research ndings in the status attain-
ment tradition have failed to dent British sociology’s faith in the SAD thesis,
despite mounting evidence that individual ability and hard work are more
important in uences on where people end up in the occupational system
than their class origins.
The rst is that path models always leave much unexplained, for the pro-
portion of variance in occupational status explained by these models rarely
approaches 50 per cent. One reason for this has to do with inevitable
measurement error, but there is also a sense in the literature that some of
the key in uences in people’s lives either cannot adequately be captured
by these models (Mayhew and Rosewell 1981: 243) or must be put down to
sheer luck (an argument rst proposed by Jencks, 1972). Bielby (1981), for
example, argues that potentially crucial in uences on status attainment,
such as the emotional quality of family life, are neglected in this tradition
of research and cannot easily be measured, and many other critics are scep-
tical about the use of IQ tests to measure ability.
Against such criticisms, however, it is clear that various aspects of per-
sonality can be adequately measured and incorporated into path models
(see, for example, Sewell et al. 1969 and Kelley 1990: 322). It is also the case
that advances in computing power and in statistical modelling now enable
us to develop highly complex path models incorporating many interacting
variables. Just as the development of loglinear modelling enormously
improved the sophistication of work in the class structuration tradition
from the 1970s onwards, so too the development of linear structural equa-
tions modelling has opened up new possibilities for status attainment

Routes of success
research today, for it is now possible to unravel chains of multiple causation
which could never have been analysed using older regression-based tech-
A second reason why the status attainment tradition has had relatively
little impact on British sociological orthodoxy is that it was accused early on
of being ‘functionalist’ and ideologically conservative. Key ndings – that
Britain is a relatively ‘open’ society, for example, or that ability tends to
count for more than class background in in uencing where people end up
in life – have been disregarded on the grounds that they are the product of
an inherently biased methodology. The charge pertains mainly to the use
of occupational prestige scales as the dependent variable in path models.
Horan (1978) is only one of a number of critics who have argued that such
scales erroneously imply a social consensus over the evaluation of different
occupational positions (an assumption Horan traced back to Talcott
Parsons and the functionalist theory of strati cation).
There are, however, occupational scales (such as the ‘Cambridge scale’)
which measure social patterns of association, rather than occupational pres-
tige, and which do not therefore depend on any assumption of value con-
sensus over the worth of different positions. These scales look very similar
to those based on prestige rankings (the Cambridge scale and the
Hope–Goldthorpe occupational prestige scale, for example, correlate at
0.88 – Stewart et al. 1980: 76), and this high degree of external reliability
seems to suggest that the problem of ‘ideological contamination’ of occu-
pational prestige scales has probably been exaggerated.
Furthermore, there are strong grounds for arguing for the use of occu-
pational measures involving a continuous scale and against categorical
schema (such as the Registrar-General class schema or the Goldthorpe
schema) which invariably run up against the problem of identifying class
‘boundaries’ where in reality there are none (see, for example, Kelley 1990
and Prandy and Bottero 1995). Given the continuing debate in British soci-
ology about the relevance of class analysis, there is surely a strong case for
using occupational scale measures as much as categorical schema in empiri-
cal research. Indeed, as Hope (1984: 16) suggests, there is no reason why
we should not use both.
The third problem which critics identify in the status attainment tradition
is that it is ‘individualistic’ and that it ignores ‘social structure’.4 Lewis Coser
(1975), for example, suggested that the status attainment approach fails to
address the question of how the structures of power and privilege (which
constitute the class positions between which people move) come to exist
and to be reproduced in the rst place. The existence of the class structure
is taken as given, and the research focuses only on individual movement
within it – on the trees and never on the wood.
This is clearly true, in the sense that status attainment research does not
try to address the ‘big’ question, which lies at the heart of the class struc-
turation approach, of whether and how class relations are reproduced over
time. But it is not true to suggest that the status attainment tradition takes

Rod Bond and Peter Saunders
no account of ‘structural’ variables (most path models aim precisely to
identify the relative importance of ‘individual’ qualities such as ability or
motivation, and ‘structural’ conditions, such as schooling or material depri-
vation in the home), nor does it follow from Coser’s criticism that the ques-
tions which are addressed in this kind of research are in some way trivial or
unimportant. Indeed, as Coser himself recognized, ‘There is surely a need
for both types of studies’ (1975: 695).
There are, then, no good a priori grounds for privileging the structura-
tionist approach to social mobility over the status attainment approach.
Indeed, we have arguably reached a point where the hitherto dominant
tradition in Britain of class structurationist work has taken us as far as we
can usefully go without pausing to re ect on why certain individuals experi-
ence social mobility when others do not. Goldthorpe and others have
clearly documented the extent of social mobility in Britain, but they have
not addressed the question of why some people take advantage of the ‘struc-
tural’ opportunities available to them while others do not. This is the ques-
tion we now seek to answer.
The source of our data is the National Child Development Study which tar-
geted all seventeen thousand children born in Great Britain in one week in
March 1958, 98 per cent of whom were included in the initial sample
(NCDS0). These children were then followed up at ages 7 (NCDS1, 1965),
11 (NCDS2, 1969), 16 (NCDS3, 1974), 23 (NCDS4, 1981) and, most
recently, at age 33 (NCDS5, 1991). In addition, details of public examin-
ation results for all cohort members were obtained from their schools in
1978. By 1991 the size of the panel had shrunk to just under 11,400, and
this has led to some under-representation of those from lower social classes,
from poorer housing conditions, with lower aspirations and with lower
scores on tests of cognitive ability (Shepherd 1993).5 We have further
reduced our sample size by focusing only on males in full-time employment
at age 33 (in later papers we intend to consider separately the mobility pat-
terns of women and of those in part-time or no employment), and a further
13 per cent of cases have been dropped for lack of adequate data over a
range of variables. This leaves us with a nal sample size for this analysis of
4,298. Few of these cases have complete data on all variables, and mean sub-
stitution has been used to replace missing values.6
Our aim is to develop a model predicting the occupational grade
(measured on the Hope–Goldthorpe scale) achieved by this sample of men
at age 33.7 We have organized the predictor variables into four broad cat-
egories – social class origins, parental support, individual ability and moti-
vation in early life, and formal quali cations achieved from the age of 16
onwards. In each case, these variables are measured at different points in
time, corresponding to the various sweeps of NCDS (e.g. father’s class is

Routes of success
measured at ages 7, 11 and 16; quali cations are measured at 16 and at 33;
and so on), so that we build up a picture of how different in uences
produce different effects over time.
Social Class Origins
Our measures of the social background of our sample include whichever
was the higher of the maternal and paternal grandfathers’ social class, plus
the father’s class (recorded at the child’s birth, and again at ages 7, 11 and
16), the mother’s class (recorded when the child was 16), mother’s and
father’s schooling (measured by whether or not they left school at the
minimum leaving age), whether or not the child attended a private school
(at age 7 and again at 16),8 and the housing conditions (measured by an
index of overcrowding) in which the child was living at sweeps 1, 2 and 3.9
Between them, these twelve measures should adequately capture the kinds
of differences which the SAD thesis claims are crucial in explaining the dif-
ferential educational and occupational success rates of children from differ-
ent types of backgrounds.
The social class of grandparents and parents is measured on whichever
version of the Registrar-General’s schema pertained at the time when the
information was collected (e.g. father’s class at sweep 1, in 1965, is based
on the 1961 schema, and at sweep 3, in 1974, on the 1971 schema). In all
cases we have coded to just three categories representing the pro-
fessional/managerial/administrative ‘middle-class’ (classes I and II), an
‘intermediate class’ of skilled employees (classes IIIN and IIIM), and a core
‘working-class’ of semi and unskilled manual workers (classes IV and V).
For the purposes of our analysis, we treat these three classes as an ordinal
Because we have four measures of father’s class spread over 16 years, we
have analysed these measures at each point in time as fallible indicators of
latent constructs and have estimated their reliability. Treating them as fal-
lible (rather than perfect) indicators means xing their reliabilities at some
value less than 1.0. Following the method outlined by Werts, Joreskog and
Linn (1971; see also Heise 1969 and Wiley and Wiley 1970), the reliability
of the rst and last wave is not identi ed, but those of intermediate waves
are, (this assumes that errors are independent and that a lag-1 autoregres-
sive model represents development over time). Given four panel sweeps,
the reliabilities of the second and third sweeps were separately identi ed
and this model was tted (a second model was then tted where these two
reliabilities were assumed to be equal, and it was found that the t was not
signi cantly worse, suggesting that the assumption of equal reliability was
reasonable). The reliability estimated was 0.64 and this was assumed to be
the reliability for all four social class measures.11
Figure I presents the results of these analyses and depicts the pattern of
intragenerational social mobility for fathers. Not surprisingly, there is a fair
degree of stability in father’s social class, although from the respondent’s

Rod Bond and Peter Saunders
Age 7
Age 11
Age 16
Father’s RG
Father’s RG
Father’s RG
Father’s RG
class Birth
class Age 7
class Age 11
class Age 16
x 2
D x 2
D df
Identity at age 7 and 11

Equal errors
Equal reliability = 0.64
Lag–1, reliability = 1.0
Father’s class: Intragenerational mobility: standardized solution
birth to when he is aged 16 the stability coef cient is 0.62, indicating only
38 per cent of the variance in father’s class when the respondent is aged 16
is predictable from his class at the respondent’s birth. When the grandpar-
ents’ class and the panel members’ own occupational status at age 33
(measured using the Hope–Goldthorpe scale) are added, the extent of
intergenerational mobility between three generations can be assessed.12 For
grandparents’ class, the path to father’s class at the respondent’s birth was
0.45 but it was also necessary to include a path (coef cient 0.12) to father’s
class when the respondent was aged 7. The coef cient from father’s class
when the respondent was aged 16 to the respondent’s occupational grade
at age 33 was somewhat smaller at 0.36, but the similarity is striking given
that different measures of class are used. Clearly there has been consider-
able social mobility between grandparents and fathers (where 80 per cent
of the variance in father’s class at birth is not explained by his grandfather’s
class), and between fathers and panel members (where 87 per cent of vari-
ance is not explained by the father’s class).