Notes on: Sullivan, A., Parsons, S., Wiggins, R., Heath, A. & Green, F. (2014) 'Social origins, school type and higher education destinations'.  Oxford Review of Education 40 (6): 739-63.  Doi: 10.1080/03054985.2014.979015

Dave Harris

This is based on the 1970 British Cohort Study (BCS70) to analyze trajectories forwards.  Social origins remain as an influence.  Attending a private school is predictive of gaining a university degree specially from an elite university.  Grammar school attendance conveys no advantage.

The Oxford studies examined inequality in the tripartite system, and also examined Boudon and Bourdieu. The work remains influential.  This paper addresses the same essential issues using different data.  These birth cohort studies have produced a lot of work lately [summarised 740].  There is a view that the declining grammar schooling has led to a decline in social mobility, although early work showed that overall, any advantages were balanced by the disadvantages of secondary moderns.  However, recent analysis has shown that 'selective school systems increase earnings inequality in adult life'[Burgess, Dixon and Macmillan 2014, Working Paper number 14-19, Department of Quantitative Social Science --??].  School differences have also been examined using various cohort studies. For the 1958 cohort, 'private and grammar school pupils achieved significantly higher exam results at age 16 than their peers at comprehensives, while those at secondary moderns fared significantly worse'.  Private schools seem to be linked to an earnings advantage in later life.  However, this cohort was unusual in that a number of changes took place in school regimes, including raising the school leaving age, and changing from a tripartite to comprehensive schools. 

Using the cohort for 1970 tests for an era when the comprehensive system was well established [by 1986, they claim].  The cohort also had changing opportunities in HE, from an age participation index in 1978 of 14% for men and 11% for women, to 20% and 20% respectively in 1990.  Student loans were introduced in that year, but tuition fees were much later [1998].  In 1992, the divide between polytechnics and universities was abolished, although this was too late for this cohort who were applying in 1988.

Status differentials for British universities are still important.  This might be because of 'maximally maintained inequality' [741, citing Lucas], which says that inequalities are maintained through status distinctions despite expansion.  Elite universities link to elite jobs, and private schools still seem to produce better access to them, although there is a view that this arises simply by better exam results, irrational discrimination by elite institutions, poor advice from state schools, or self exclusion.

For Bourdieu, cultural resources are fundamental [Reproduction…].  However, cultural capital should not be defined exclusively in terms of elite cultural activities, but should include knowledge and skills.  This links with early studies on the presence of books in the home and reading behaviour, and reading seems particularly important.  Bourdieu has been seen as vague.  Boudon talks about primary and secondary effects, cultural inequalities first, then different costs and benefits affecting different educational routes, especially at key points, and these are primarily the result of economic inequalities.  All individuals attempt to avoid downward social mobility, but the importance and consequences vary according to parental occupation.  Secondary effects may be exponential.  Ethnic minority status can work in the opposite direction, producing increased participation by minorities at particular levels of attainment [citing Waters, Heath, Tran, and Bolivar 2014, in Alba and Holdaway (Eds).  The children of immigrants at school: A comparative look at integration in the United States and Western Europe.  New York: New York University Press].

Secondary effects have been defended by some recent studies, but there is a methodological problem in gaining effective data about initial attainment, since measures are only 'weakly differentiated'.  There is a problem also with people who leave school with no qualifications (742), and this affects the 1970 cohort.  There also some additional difficulties in that primary stratification is assumed to be a given, while secondary effects reflect choices.  However measured ability itself varies over time and can be affected by actions and investments.  In the last decade, there has also been an increased focus on the early years, both in policy and in research [including some on the 1970 cohort].  Some of this shows that cognitive inequalities also emerge early and reflect advantage.  Cognitive development itself needs to be seen as [only one?] factor in 'individual learning trajectories'.  Current policy focusing on early years might over estimate the affects, however.  Also, primary affects are seen as culturally determined, and secondary effects as economically determined, but both are likely to interact in affecting educational decision makers, especially if we take cultural factors to include knowledge of the education system, or beliefs about mobility.  This interaction is a separate dimension from the ones separating primary and secondary.  Additionally, there is an assumption of clear branching points in secondary careers, and this compares with 'the relatively untidy accumulation of qualifications during the life course' (743), and the neglect of processes before those branching points.  A final complication is that expectations can affect current efforts and therefore attainment.

This paper uses a life course approach, allowing for cognitive trajectory by using cognitive scores at a range of ages, and taking qualifications only at the end of compulsory schooling.  Social background influences both test scores and educational attainment.  Cognitive scores can be assumed to measure intellectual development, but qualifications can also 'reflect such factors as an understanding of the "rules of the game" of doing well in particular high stakes examinations', as well as opportunities to be entered.  These different factors can be unpacked and their weight assessed.  However, motivation, compliance 'as well as potential stereotype - threat (anxiety due to potential to confirm a negative stereotype about one social group)', also affect cognitive scores.  It is also difficult if multiple choice tests are used, since these 'do not capture the full range of academic competencies', and show gender differences.  There can be no attempt to estimate 'innate intelligence' (744) from these tests of attainment, which measure only particular performances in particular conditions.

The first question was to estimate chances of gaining an elite undergraduate degree and ordinary one by the age of 42.  Access to elite universities is a new issue for longitudinal studies.  Several questions emerged - what are the effects of early years and cognitive scores, is type of secondary school relevant, what are the effects of socio economic advantage explained by secondary schools, can the influence of schooling be captured by cognition up to 16, and then exam performance, and is it confined to that stage, are there effects of social origins on degree in elite degree attainment  other factors have been controlled, and if so are secondary effects like this is due to economic or educational and cultural resources.

The 1970 cohort study has 17,000 people in England, Scotland and Wales born in a single week.  It collects data on health, physical, educational and social development, and economic circumstances.  In a series of surveys '(or "waves")' at ages 5, 10, 16, 26, 30, 34, 38 and 42.  016, 16 instruments were deployed, but unfortunately there was a teachers' strike or at the time which affected completion, leading to 'substantial instrument non response', for those tests of cognition administered at school (744).  There are therefore gaps in the trajectory, although this [ESRC] project attempts to repair the data by using 'previously unused cognitive variables' (745).

First of all, missing data are identified, then 'multiple imputation' is used to suggest values, using a technique called 'Schafer's algorithm', which assumes randomness in missing values.  This is strengthened by including auxiliary variables, 'to protect against departures from multivariate normality' [these auxiliary variable have  a known distributional quality, and are used to match the missing data, as are demographic variables in conventional surveys?].  The 'analytical sample' of 7767 focuses on people resident in England and Wales in 1986, with a full set of birth characteristics, and who had provided information about degree and school type.  Scots members had different qualifications.  Further attempts to address missing date or involved replication, 20 times, using 'Rubin's Rule' which apparently helps you gauge the efficiency of estimation.

The dependent variable is degree attainment, in three categories - no degree, degree, elite degree, based on a question asked when respondents were 42.  Elite status was judged on the basis for the institution attended --membership of the Russell Group.  This is acknowledged to be arbitrary to some extent.  They considered taking only Oxbridge, but this led to insufficient numbers.  Polytechnic degrees were rated as ordinary.

'Multinomial logistic regression'calculates the log odds of falling into particular categories 'for various linear combinations of our explanatory or predictor variables'.  Because this is a longitudinal survey, a series of models can be produced to assess attainment at ages 5, 10, and 16 sequentially.  Luckily, the cohort surveys are 'particularly rich in measures of cognition', a series of cognitive tests.  Results from these tests are then turn into 'a single main component score' using 'varimax Principal Components Analysis', and reliability is assessed using Cronbach's Alpha.

Six models were produced 'which captured distinct stages of the life course in cumulative manner'.  Model 1 starts with characteristics of birth as well as indicators of economic cultural and parental educational background.  Cultural resources 'reflect the importance of books and reading'[748 -- they went round and counted number of books per household, as in the classic studies?].  Model 2 thirds cognitive scores at age five.  Model 3 adds cognitive scores of age 10.  Model 4 include secondary school type.  Model 5 includes cognitive scores and scores for attainment in public examinations at age 16 [these are scores based on usual calculation of points, it seems].  Model 6 includes post compulsory academic qualifications.

In model 1, birth characteristics included sex, birthweight, position in birth order, age of mother at first birth - all these are apparently predictors of educational chances.  Ethnicity was not a category, because so few in the sample were from ethnic minorities, and the same goes for single parents status at birth.  Educational and cultural resources included parents highest qualification, mothers or fathers, frequency of readings of the child, according to mothers'estimates, newspapers in the home, divided into tabloid, broadsheet, both, or had none, which assumes that 'the prose style of tabloids were simpler and geared towards the lower reading age and smaller vocabulary' [I am not at all sure this is reliable].  Newspaper reading was a 'strong cultural identifier' in the 1980s, and is assumed to be 'a stable characteristic' for the preceding years [it was only gathered at age 16].  Economic resources included occupational social class, on the only available RG scale.  The highest occupation was chosen based on either mother or father, current almost recent job.  Non working mothers were grouped with the same category as working fathers assuming that this was the only information available - the authors tell us that 'this is a small group'(749).  They also examined home ownership, and overcrowding, measured by 'the ratio of people in the household per room' assuming that more than one person per room is overcrowded.

In model two, aged five, cognition was added, based on results from five tests - copying designs, English picture vocabulary, human figure drawing, complete a profile, and Schonell graded reading.  In model three, aged 10 there were a further eight cognitive tests -Edinburgh Reading Test, pictorial language comprehension test, friendly maths test, spelling, British ability scale, two verbal sub scales (word definitions and word similarities, and two non verbal sub scales, (digit recall and matrices).  Model four included school type results from a questionnaire addressed to head teachers, a schools census, and a retrospective question asked in 2012.  Missing data was dealt with by using one or other of these surveys.  Model five, in 1986, included nine cognitive tests, although the results of only two (spelling and vocabulary) were deposited in the archives at the time, 'due to lack of funds'.  Arithmetic tests were added later, and in this study, the data from the six remaining ones have been added, for the first time.  Examination results of age 16 produced a total points score (O level grade A got seven points, down to grade E that got three, CSE grade one is seen as equivalent to O level grade C, receiving five points and so on, down to grade six CSC which got one.  Possession of either maths or English O level grades A to C, or CSE grade one received an extra binary weighting. Model six added A level qualifications, taken at age 18, although they included data up to age 20 to allow for retakes.  Categories range from three or more A to C, one or  two A-C, lower grades, none.  The authors note that these qualifications 'have been subject to substantial grade inflation since the abolition of norm - referenced marking in the 1990s, and that 'it was still possible [then] to get a place at a Russell Group university with C and B grades at A level' (750).

The first level of analysis offers descriptive information for the variables, by school type.  'A clear hierarchy of schools emerges', based on both socio economic background, cognition and qualifications [of students].  Private schools are at the top, then grammars, then comprehensives, and finally secondary moderns.  As examples, '52% of privately educated cohort members have at least one graduate parent, compared with 31% of grammar pupils, 14% of those who attended comprehensives, and 8% at secondary moderns'. 

16% of the sample got ordinary degrees, and 7% elite ones, and again school type was important - 31% of privately educated pupils got an elite degree, 13% of grammar, 5% of comprehensive, 2% of secondary modern [that last 2% must have been an extremely interesting group!].  All the cognitive and educational measures were correlated to produce a matrix.  Correlations vary between  0.28 [cognition aged five and A levels], and  0.68 [cognition at 10 and at 16].  Another table (753 below) shows 'multinomial logistic regressions predicting the odds of achieving an elite degree or an ordinary degree compared to no degree by age 42' (751)

Tab 3


Tab 3/2


In model one, there was no significant difference between male and female members, birth weight was not significant, older mothers did seem to be able to convey an advantage, but birth order did not.  Parental education 'is highly significant', however, and having a graduate parent gave you three times better odds of gaining a degree compared to those with parents with no qualifications (5.5 times better odds of gaining an elite degree).  Having tabloid newspapers compared with having no newspapers at all lowers the relative odds [!!] , but having broadsheet newspapers raises them, the more so for elite degrees.  Frequency of reading by parents is 'highly significant', again with a premium when it comes to elite degrees.  However, only the top social classes were significantly link to elite degree chances, and  'social class was not a significant predictor of getting a non elite degree'.  Home ownership raises the relative odds, overcrowding reduces it.

In model two, cognition at age five 'is powerfully linked to the relative odds of gaining an ordinary degree', more so with elite degrees [an odd way of calculating odds here 'a single standard deviation increase in the score translates into 1.6 times the odds of a degree and 1.9 times the odds an elite degree over not gaining a degree'].  Early cognition scorers seem to mediate partially the influence of birth characteristics and home background.  'Social class becomes non significant in this model', although the specific variables measuring cultural, educational resources and economic resources 'remain substantial and significant overall'.  The effect of having a graduate parent is 'reduced slightly'

Model three shows that cognition at age 10 is more 'powerfully predictive of higher education chances than cognition at five'(755).  Cognition of both ages now reduces the advantage produced by reading to insignificance.  Model four shows that private schooling, with 5% of the sample attending, produces 1.7 times the odds of gaining an ordinary degree, and three times or more the odds for an elite degree compared to comprehensive pupils 'with similar backgrounds and cognitive attainment'.  Grammar schools had no effect, and nor did any disadvantage of attending a secondary modern school.  Differences in home resources are only modestly moderated by school type including private schools, however.  Interactions between school type, cognition at 10, apparent social class, and parents education were designed to see if school types interacted with different characteristics - for example, did grammar school attendance help specifically working class pupils who attended them.  There were no statistically significant interactions overall.

Model five includes cognition tests at 16 and exam performance.  Cognition, A level scores, maths and English O level grades 'are all independently linked to degree chances' in achieving. O level maths improved the odds of an elite degree 'strongly'.  Cognition at aged 10 'remains powerfully significant', but this could be due to measurement error.  The advantages of private school attendance 'is only modestly attenuated': school type and age 16 attainment do not interact.  Social origins show secondary affects in this model.  Only home ownership confers an advantage.  Broadsheet newspapers become insignificant, although tabloid newspapers still reduce the odds of degree and elite degree.  Parental education remains 'highly significant, though reduced' (756) -having a graduate parent multiplies the odds of gaining a degree by 1.7, and an elite degree by 2.5.  This cultural secondary effect is greater than the effect of home ownership for elite degrees, probably due to greater knowledge of the system and of its status differentials.

Model six introduces exam performance, helping to explain any differences because of different rates of staying on and performance in FE.  Patterns of results remained 'broadly similar', however, and those with the same A levels results varied according to parental qualification.  Private school members also had increased odds - 1.4 for a degree, 2.5 for an elite degree - compared to comprehensive students with the same A levels.  This is 'clearly non meritocratic', because the differences persist across a range of conditions for cognition and exam results.

Overall, despite some advantages, limitations include not accounting sufficiently for diversity within school sectors, not looking at the interaction between gender and different school types, including the role of single sex schools: these are to be addressed in future papers. 

It seems that social inequalities in access to HE are mediated by early cognitive scores, 'but far from fully'.  Scores at age 10  leaves differentials supplied by parental education and home ownership.  However, 'early cognitive differentials are very important and have far reaching effects' (757), and independently predict degree chances even allowing for subsequent attainment, but again this is not a full accounting.  The emphasis on early years is important, but 'there is a risk that this can be exaggerated'.

School type is linked to the chances of gaining a degree and elite degree, cognition characteristics and resources having been controlled.  Private schooling conveys 'a powerful advantage', especially in gaining an elite degree, although 'this advantage is not necessarily causal', because there are likely to be 'unobserved differences between parents who send their child to a private school…  motivation and ambition…  differences in political outlook linked to attitudes towards "getting on"'.  Grammar schooling was not linked to any significant advantage, 'surprisingly', although they might have made a difference to attainment within compulsory schooling, at O level, for example.  However, 'this did not follow through to university chances', producing a '"leaky pipe"' between grammar school and university entrance.  The answer might lie in characteristics of schools in detail, including the degree of academic selectivity - some grammar schools were 'no longer highly selective'.  Overall, they seemed to provide no special benefit for working class pupils HE chances.

Social background differentials were also mediated by different types of school, 'slightly' by secondary school type, despite the other advantages associated with private schooling.  No significant interactions were found to moderate school differentials with pupils' social class, parents' education and prior cognitive attainment.

Comparing cognition and qualifications, shows that private school advantage is 'only very slightly reduced' by subsequent attainment at 16, implying that private schooling provides 'post- 16 factors'.  These might include the level of aspiration of parents and schools, and links between universities and private schools, especially for the elite universities and 'a small number of elite private schools' (758).  Again this is to be investigated.  There is no analysis for meritocratic views, that the privately educated represent some concentrated pool of talent, since it looks as if state educated pupils outperform privately educated ones once at university.

Secondary effects on degree chances can be examined by controlling for cognition throughout childhood and attainment at 16.  There is 'a significant robust association with home ownership, and especially with parental education': the latter gives advantages especially in the case of gaining an elite degree.  Social class itself [occupation and economic resources] seem to have 'no significant secondary affect', challenging Boudon's view.  Parental educational status 'is the largest source of secondary effects'

Overall, increasing access to HE while ignoring status differentials 'will tend to lead to an underestimation of social inequalities in access', more so for the current generation than it was for the 1970 generation, 'given the great expansion...and increased diversity of universities and courses since 1992'.  A 2013 study shows that pupils from lower class backgrounds applied to Russell Group institutions less frequently than those with similar qualifications from higher class backgrounds.  The same goes for comparing state and private schools, and the former 'are less likely to apply then to be awarded a place if they do apply'

Overall, 'being bright is not necessarily enough' []very very bright is, is my guess, but with diminishing returns - -like the Douglas study found years ago]. Social advantage in terms of origins and private schooling raises the chances 'of gaining a degree, especially an elite degree, beyond cognitive and examination attainment'.  Such 'non meritocratic processes have important repercussions', including 'the domination of Britain's ruling class by graduates of private schools and elite universities'.


back to more social mobility studies in 'files'