Impact of Social Origins on Educational Attainment: A Multidimensional Approach, Study notes of Statistics

Download Impact of Social Origins on Educational Attainment: A Multidimensional Approach and more Study notes Statistics in PDF only on Docsity! 1 Social inequalities in educational attainment. The changing impact of parents’ social class, social status, education, and family: England 1986 and 2010 Nicola Pensiero UCL Institute of Education, UK Ingrid Schoon UCL Institute of Education, UK Abstract There is controversy regarding trends over time in the association between social origins and educational outcomes in the UK. An explanation may lie in different methods of analysis. This article provides new evidence about trends in inequality between the 1980s and 2010s and informs the debate about the conceptualisation and operationalisation of social origins. It expands the multidimensional conceptualisation of social origins proposed by Bukodi and Goldthorpe (2013) by adding a separate indicator of family income to those of class, status, and education of parents. Results from two UK age cohorts born in 1970 (BCS70) and 1989/90 (Next Steps) show that social class, social status, education and income all have independent effects on educational attainment and can show different patterns of stability or variability over time. Moreover, the study highlights the importance of transitions to upper secondary education for a more comprehensive 2 understanding of inequalities in educational progression and attainment. 1 Introduction A series of longitudinal studies has confirmed social origins as important determinants of educational attainment. However, there is still controversy about the trend across British cohorts in educational inequalities, that is in the association between social origins and educational attainment (Blanden, Gregg & Macmillan, 2013; Goldthorpe, 2013; Bukodi and Goldthorpe, 2016). The UK seems to have followed an equalising trend in educational attainment in the first half of the 20th century (Breen, Luijkx, Müller, & Pollak, 2009), while throughout the second half, when radical changes in educational policy occurred, it is not clear whether educational inequality was falling, rising or on a flat line (Goldthorpe, 2013, 2016). Plausible explanations of the diverging evidence concern differences in methodology and conceptualisation of constructs. Some researchers model education outcomes as the highest level of qualification attained (for example Breen et al., 2009), whilst others consider a sequence of transitions from lower to higher levels (Mare, 1981; 1980). The definition of individuals’ educational attainment can also differ in respect to absolute versus relative value assigned to education (the latter treating education as a ‘positional’ good (Bukodi & Goldthorpe, 2016). This article argues that the uncertain direction of trends has another cause that has received relatively little attention: the incomplete 5 distinguish occupations in terms of social relations in labour market and in the production process. Social class thus defined, it is argued, is a valid indicator of income levels, security and prospects (Chan & Goldthorpe, 2007; Goldthorpe & McKnight, 2006) and therefore it well covers parental economic resources, making the use of an additional indicator of economic resources unnecessary. Furthermore, social class is at the same level of abstraction of socio-cultural resources rather than serving as a proxy for them and is, similarly, a relational concept. Thus, Bukodi and Goldthorpe (2013) suggest decomposing social origin instead of decomposing social class. They would complement social class with social status as an indicator of socio-cultural resources. Social status is understood as an indicator of a structure of relations of perceived social superiority, social equality, and s o c i a l inferiority, as expressed in selective intimate relationships and in distinctive lifestyles. While the class structure is grounded in relationships within labour markets and production units (Erikson & Goldthorpe, 1992; Erikson et al., 1979), the social status order refers to relations of perceived social standing. It distinguishes between those who, by virtue of their higher position and ascribed attributes, behave as superiors and those who have a less advantaged position and consequently behave with deference. Although the expression of social status is less overt and more implicit nowadays, it is still recognisable in social networks (Chan & Goldthorpe, 2007; 2010). Another indicator considered is parental education, understood as an indicator of “educational resources”, providing a supportive home learning environment and knowledge about how to navigate the educational system (Bukodi & Goldthorpe, 2013; Erikson & Jonsson, 1996;). Although they did also find some very small independent effect of family income in one cohort, Bukodi and Goldthorpe (2013) conclude 6 that social status in combination with social class and parental education should comprehensively account for social origin effects on educational attainment. In this article we follow Bukodi and Goldthorpe (2013) and Bukodi et al. (2014) in decomposing social origins into social, occupational and educational components, but add family income as a separate indicator. A substantial portion of permanent income, in fact, is unrelated to social class, which therefore might not be a sufficient proxy of the variation in economic status between families (Blanden et al., 2013; Grusky and Weeden, 2001; McIntosh and Munk, 2009). Furthermore, the relevance of social class for children’s educational attainment is not limited to the opportunities that income entails. More generally, social classes specify the s o c i a l r e l a t i o n s w i t h i n t h e labour market and can determine standards of living and life chances. Apart from the opportunities for income, social class also determines the quality of work conditions (job security, career opportunities, working hours and stress) which in turn have a specific relevance for children’s educational attainment by influencing the quality and quantity of the relationship between family members including family disruptions (Furstenberg & Kiernan, 2001; Menning, 2002, Perry-Jenkins, Goldberg, Pierce & Sayer, 2007; White & Keith, 1990). 3 Changes in the educational system In considering trends in educational inequality it is important to take account of the changing socio-historical context. Both of our cohorts faced a compulsory school leaving age of 16 (introduced in 1973), but this 7 study covers a period of further educational expansion and major changes in the education system from the late 1980s onwards. In 1988 there was a switch from GCE O-levels system (General Certificate of Education, Ordinary levels) to GCSEs (General Certificate of Secondary Education), see Table 1. Under the former regime, more academically oriented students took O(Ordinary)-levels at age 16 and A(Advanced)-levels at age 18. A-levels are the requirement for entering higher education. Less “academic” pupils could take the Certificate of Secondary Education (CSE) at 16 (which we treat as the lowest academic qualification) or vocationally oriented programs. The 1970 cohort was one of the last to be educated under the GCE O- level system. The 1988 reform combined O- level and CSE exams into General Certificate of Secondary Education (GCSEs), which were u s u a l l y taken at age 16 Students no longer had to decide whether to take the less academic CSE or the more academic O-level exams. This, in turn, is thought to increase the participation of those in the middle of the skill distribution and in particular of those at the borderline between academically oriented and less academically oriented students. Moreover, the GCE O-level system was based exclusively on exam performance whilst the GCSEs also take into account the coursework. Table 1 here GCSEs turned out to be more accessible than the O-levels and the result of the reform was that a higher proportion of students – 93% of members of the Next Steps (1990) cohort against 76% of the 1970 cohort – left the school with at least some academic qualifications. Focusing on the post-compulsory phase, 38% of the more recent cohort attained an A-level qualification, against 16% of the previous (Table 2). Level 2 10 the reforms also made more information about the effectiveness of schools available to parents and the public, in the form of publicly available test score information, known as’ League Tables’ (Hansen & Vignoles, 2005; Machin & Vignoles, 2006). Increased competition among schools and decentralisation of school finance can potentially enhance attainment, but can raise inequality as well because advantaged families are better able to take advantage of the diverse opportunities created by a more market-oriented system (Blanden, Gregg & Machin, 2005; Galindo- Rueda & Vignoles, 2005; Gibbons & Machin, 2008). While the 1970 cohort was educated in the period prior to the market-oriented reforms, the 1990 cohort experienced a system that was already transformed by those reforms. Whether the reforms are reflected in a change in social class differentials in attainment is explored in the next section. 4 Previous research Previous evidence suggests that the dependence of educational attainment on household income has increased over time in the UK at the tertiary level, while it has gone down for secondary qualifications after the introduction of GCSEs in 1988. Blanden and Gregg (2004) found that the relationship between family income and final educational outcome has been strengthening across cohorts born in 1958 and 1970. By contrast, Gregg and Macmillan (2010) showed that the gradient of educational attainment at age 16 by social origins (income or class) has lessened between generations born in the 1970s and those born in the 1980s and early 1990s. They relate the improvement in equality of educational opportunity in educational attainment at age 16 to the 1988 reform introducing GCSE qualifications. 11 Blanden, et al., (2005) confirmed an initial increase in inequality in post- 16 participation by family income, followed by a decline after the introduction of GCSEs in 1988, and an increase at the tertiary level. The rapid expansion of higher education, they argue, had benefited children from wealthier families, disproportionately. The argument is supported by Galindo-Rueda and Vignoles (2005). They examined the relative importance of family background and ability and found that the importance of ability in accounting for educational attainment has declined over time, whilst that of parental class and parental education has increased. They attribute this partly to the fact that less able children from advantaged backgrounds have benefited most from the largest increase in educational participation. Boliver (2011) has shown that educational expansion, in and of itself, has not caused educational inequalities to decline in the UK. Instead, she found that social class inequalities in British higher education (HE) have been maintained both quantitatively, in terms of persistence of social class differentials in HE enrolment, and qualitatively, in terms of differential access to higher status courses. Similarly, Schoon (2010) confirmed that the association between academic attainment and a composite index of family social background comprising parental education and social class has remained stable over time, while the association between academic attainment a n d general cognitive ability decreased for the 1970 cohort compared to the 1958 and 1946 cohorts. Social background (whether as class or parental education) also showed persisting associations with transitions at 16 to A-levels and at age 18 to university in cohorts born from 1958 to 1991 (Jackson, 2013). Most of these studies examined a limited variety of family background factors. Bukodi and Goldthorpe (2013) have dealt with the 12 omission of relevant factors from a conceptual perspective, by decomposing social origins into parental class, parental status and parental education. Using evidence from the 1946, 1958 and 1970 British birth cohorts, they found that these three components of social origins have independent and distinctive effects on children’s educational attainment. In detail they found a stable effect of parental class, a weakening effect of parental status and a stronger effect of parental education. From this follows that if any of these factors was chosen as the sole indicator of social origin, it would cause an overestimation of the effect of that factor and an underestimation of the total effect of social origins. 5 Research questions This article contributes to the debate over trends in educational inequality by addressing the following questions. Do parents’ class, education, social status and family income show an independent effect on children’s educational attainment? If so, which of the different socio- economic family resources are implicated in producing educational inequalities. Does one set of resources become more important compared to another? Do the different indicators show similar or different trends? For the younger cohort, we include information on highest qualifications attained by 2010, as collected at age 19/20. By then most cohorts members will already have attained level 3 qualifications (entry to university qualifications), but not all of them will have decided whether to enter university. For this reason we focus on level 3 qualifications as the final educational outcome, which enables us to assess inequalities before making the step to university. 15 information on the Socio-Economic Group of both parents (SEG) at respondents’ age 10. Following the recoding procedure described in Goldthorpe and Jackson (2007) we recoded the SEG to the Goldthorpe class schema. In the Next Steps cohort we coded parental class using the National Statistics Socio-economic Classification (NS-SEC), which represent the Goldthorpe class schema for Britain (Goldthorpe, 2007) (see also Office of National Statistics – The National Statistics Socio- economic classification (NS-SEC)). In cases where both parents are employed we select the higher of the parents’ class in line with the dominance approach (Erikson, 1984). Our indicator of parental social status is based on the scale proposed by Chan and Goldthorpe (2004), which is derived from the occupational structure of close friendship relations. Cohort members’ parents are coded to the 31 categories of the scale on the basis of the allocation to Standard Occupational Classification 1990 (SOC90) occupational unit- groups. Where both parents can be allocated to the scale, we adopted the dominance approach. In the first wave of the Next Steps, there was insufficient detail on both parents’ occupational unit group, so we used data from the second sweep to construct the social status indicator. Parental education is defined as the highest academic qualification of either parent (dominance approach). It has been shown that the commonly used qualifications variable, which treats vocational and academic qualifications (NVQ) as equivalents has less predictive power of children’s educational outcomes than a variable giving prominence to academic qualifications (Sullivan, Ketende & Joshi, 2013). Accordingly we classified parental education on the basis of academic-qualifications in the same way as the cohort members, adding a level (4) for degree level qualification or higher. 16 Information on family income is banded in both in the BCS and Next Steps, therefore income cannot be directly operationalised as an interval variable (or percentiles). We constructed an indicator of four groups that is the finest-grained possible given the limits imposed by those bands. The resulting variable distinguishes between the bottom 7% of families, a second group comprising the next 30%, a third group of 34% , and finally the top 29% of families. We did not attempt to construct a continuous estimate of income because the covariates that would be used to impute values within intervals might introduce multicollinearity. In order to deal with the potential issue of multicollinearity arising from the use of different indicators of social origins we used two main diagnostic procedures: regressing each of the independent variables on the others (and a dummy variable indicating the cohort) and calculating the (pseudo)- R-squared value; and secondly, latent class analysis of the different indicators, assuming that they are manifestations of a single latent factor (Hagenaars & McCutcheon, 2002; Muthén, 2001). The findings disconfirm that multi-collinearity is an issue that might bias the estimates of our models. The (pseudo)-R-squared value for the multinomial logistic regression estimating social class is 0.2, the adjusted R-squared value for the OLS regression estimating social status is 0.39, and the pseudo-R-squared values of the ordinal logistic regression estimating parental education and family income are respectively 0.16 and 0.12. The magnitude of the (pseudo)-R-squared values does not reach the threshold of 0.8 one would expect in the case of large communality. The highest R-squared value found in the case of social status (0.39) indicates a low level of multicollinearity. If multicollinearity were an issue and the use of a common factor were the best fitting strategy, then the results from the latent class analysis would show the number of classes to be “limited”, 17 most cases would be found in classes representing consistent combination of indicators, for example a class comprising cases with high scores on all indicators, a class with middling scores on all and one with low scores on all indicators. Inconsistent classes in which indicators behave differently (high scores on one indicator and low scores on other indicators) should not emerge or would only contain a residual proportion of cases in such a hypothesis. The results show that the solution with 8 classes including inconsistent classes fits better the patterns of relationships between the indicators than the consistent 4 class solution. The sample-size adjusted Bayesian Information Criterion (BIC), in fact, equals 328693.437 in the first case and 343358.265 in the second. 7 Results Table 2 shows that a considerable number of cohort members did not achieve level 2 qualifications by age 20, 42% of the 1990 cohort compared to 36% in the 1970 cohort, pointing to persisting low levels of achievement. However, it has to be taken into account that while BCS70 data is based on self-report, information about qualifications in Next Steps is taken from the NPD. We also see that there had been an increase in level 3 qualifications for the later born cohort and a decrease in children not attaining any academic qualification. Table 2 around here Do social status, parental class and education and family income show an independent effect on children’s educational attainment? We adopt a sequence of logit models that reflect the ladder structure implied by the English education system (Mare, 1981; 1980) to estimate the likelihood of attaining 1) at least level 1 qualifications versus 20 Table 4 shows various patterns of change and stability of inequality in educational attainment across the social origins indicators. Table 4 here The social class attainment gap had a prevailing pattern of stability, there is no clear indication of reducing inequality, although it widened somewhat for specific groups at certain transitions. The differentials between class VII (routine) and class III (intermediate) in attaining at least level 1 qualifications have widened across cohorts, there is no support for a change in the gap between class VII and other classes at that transition. At transition 2, the gap between class VII and class IV has widened. At transition 3, only the attainment gap between class VII and class VI (semi-routine) has widened across cohorts. The social status attainment gap appears to have slightly increased across the three transitions, reaching statistical significance only at transition 2. With respect to parental education, the gap in the first transition has become wider when comparing children of parents with levels 1 and 2 qualifications with those of unqualified parents. The pattern is reversed at the next transition (to ‘O-level’), where the differentials between no parental qualification and parental qualifications at levels 1 or 2 are significantly narrower. Similarly, at the transition to A-level, among those qualified to level 2, the trend over time is equalising for all parental education categories. Household income: the attainment gap between the least affluent income group and the (two) more affluent income groups enlarged at transition 1. At transition 2, the progression gap in academic 21 qualifications has widened in a significant way only between the two extreme income groups. At transition 3 there is no evidence of a significant change inequality in attainment over time. In summary, the most salient patterns of changes in inequality are that at transition 1, from no to any qualifications, there has been a widening or persisting attainment gap between children from families with different socio-economic resources, while at transitions 2 and 3 there has been a reduction regarding the role of parental education. This means that the different socio-economic groups have taken advantage of the expansion of the access to and attainment of academic qualifications at different paces. The later born cohort experienced a generalised increase in the rates of completion of the first transition: the proportion of children without any qualification at age 20 declined from 23% to 7% roughly (Table 2). Yet, children from advantaged parents have experienced a more rapid decline, indicating that the expansion of educational attainment at the first transition has benefited the advantaged groups more than the disadvantaged. At transition 2 (attainment of at least ‘O-level-type’ academic qualifications at age 20), there does not appear to have been expansion and, at face value, the proportion of children completing this transition declined from 64% to 58% (Table 2). This decline may be overstated or even artefactual, given the possible biases noted above of overstatement of GCSEs reported BCS (Shepherd, 2001) and a possible small downward bias to records of Level 2+ qualifications in Next Steps. Even considering this caveat, many of the Next Steps cohort still found the second transition beyond their reach. Assuming that the overstatement of GCSEs reported in BCS is not systematically related to cohort members social origin’s indicators, we argue that the results regarding the variation of inequality of opportunity in educational attainment are not affected by 22 the likely upward bias in level 2 qualifications in BCS.At the third transition another major expansion has occurred, which more than doubled the proportion of children who have attained an A- level academic qualification, which increased from 17% to 38%. At this point, the expansion was accompanied by a reduction of inequality associated with parental education and by stable inequality in relation to social class, social status and family income. 8 Conclusions This article addresses a vexed question about change or stability in social inequalities in educational attainment. We compared two British age cohorts born in 1970 and 1989/90. We argue that the controversy regarding trends in social inequalities arises partly because parental social class or income should not be the sole indicators of social origins. A multidimensional conceptualisation of social origins should embrace, social class, social status, parental education and family income. Each of these factors shows independent associations with offspring’s educational attainment, suggesting independent mechanisms by which growing up in different families leads to diverging educational outcomes. When social class is used as the sole indicator of social origins the extent of social inequality in educational attainment is underestimated and the extent of social class inequality is overestimated. The findings also suggest that omitting family income from a more differentiated and comprehensive conceptualisation of social origins would still lead to incomplete conclusions. in that family income has an independent effect on educational attainment. 25 strengthened impact of family income at the first two transitions is the marked rise in income inequality (Johnson & Webb, 1993) and child poverty in the UK during the 1980s (Department of Social Security, 1998:1999; Gregg, Harkness & Machin, 1999). The stable impact of social class indicates that the advantages that the concept of class captures have changed little during the time period considered. Relating this finding to the results from Bukodi and Goldthorpe (2013) regarding the stability of class-related inequality for older cohorts, a long-term trend emerges indicating that the social class inequalities in educational attainment have not changed since the 1950s up to 2000s. The increasing gap across cohorts in attainment at the second transition associated with social status might reflect the increasing return to information associated with social status. It might also indicate that social networks have become more important in facilitating access and progression in the educational system. Among our four a s p ec t s o f social origins, only parental education had a generalised expansion and equalisation, due to the educational expansion at the secondary and higher level that occurred since the late 1960s. The parents of the 1990 cohort benefited from the rapid increase in participation in secondary and higher education that was determined by this expansion. Other dimensions of social stratification have not changed in this equalising and expansive way. This is probably the reason why inequalities in relation to parents’ education have reduced over time at transition 3, while there hasn’t been a reduction in inequality in relation to other factors. Another aspect to be considered is that the students who have passed level 2 qualifications are a selected group with relatively high skills. At transition 3 part of the effect of parents’ education on skills is captured by previous attainments; once p a s t the hurdle of obtaining at least level 2 26 qualifications, the expansion of education among the parental generation can finally lead to an improvement in equality of opportunity. By contrast, at transition 1, the expansion of education participation has mostly b e n e f i t t e d advantaged children from well-educated parents because of the persistent inequality in the long tail of low achievement. In summary, the results suggest that social class, social status, education and income all have independent effects on educational attainment. Furthermore, their effects can vary in different ways, i.e. they show different patterns of stability or variability over time and for different qualification levels. When they vary, they can trend in either direction. What is needed for a better understanding of these independent effects is to hypothesise and test the specific social processes or mechanisms that underlie the observed associations. Acknowledgements: This article is based on research in the LLAKES Research Centre funded by the ESRC: grant reference ES/J019135/1. We thank Tak Wing Chan for providing his code converting SOC90 to the social status scores. We thank Heather Joshi for her comments. References Blanden, J., Gregg, P. & Machin, S. (2005). Educational inequality and intergenerational mobility. In S. Machin and A. Vignoles (Eds.). What’s the Good of Education? The Economics of Education in the UK, chapter 6. Princeton, NJ: Princeton University Press. 27 Blanden, J., & Gregg, P. (2004). Family income and educational attainment: a review of approaches and evidence for Britain. Oxford Review of Economic Policy, 20(2), 245–263. Blanden, J., Gregg, P., & Macmillan, L. (2013). Intergenerational persistence in income and social class: the effect of within group inequality. Journal of the Royal Statistical Society: Series A (Statistics in Society), 176(2), 541–563. Blanden, J., & Macmillan, L. (2016). Educational Inequality, Educational Expansion and Intergenerational Mobility. Journal of Social Policy, 45(4), 589–614 Boliver, V. (2011). Expansion, differentiation, and the persistence of social class inequalities in British higher education. Higher Education, 61(3), 229–242. Bourdieu, P. (1984). Distinction: A Social Critique of the Judgement of Taste. Cambridge, MA: Harvard University Press. Breen, R., Luijkx, R., Müller, W., & Pollak, R. (2009). Nonpersistent inequality in educational attainment: Evidence from eight European countries. American Journal of Sociology, 114(5), 1475–1521. Brooks, G., Pugh, A. K., & Schagen, I. (1996). Reading performance at nine. Slough: NFER. Bukodi, E., Eibl, F., Buchholz, S., Marzadro, S., Minello, A., Wahler, S., . . . Schizzerotto, A. (2018). Linking the macro to the micro: a multidimensional approach to educational inequalities in four European countries. European Societies, 20(1), 26-64. Bukodi, E., & Goldthorpe, J. H. (2013). Decomposing ‘social origins’: The effects of parents’ class, status, and education on the educational attainment of their children. European Sociological Review, 29(5), 1024–1039. 30 Politicians and the Limits of Educational Policy, Journal of Social Policy, 42: 431-50. Goldthorpe, J. H. (2016). Social class mobility in modern Britain: Changing structure, constant process. Journal of the British Academy, 4, 89–111. Goldthorpe, J. H., & Jackson, M. (2007). Intergenerational class mobility in contemporary Britain: political concerns and empirical findings 1. The British Journal of Sociology, 58(4), 525–546. Goldthorpe, J, H. & McKnight, A. (2006) The economic basis of social class. In S. Morgan, D. B. Grusky, & G. S. Fields (Eds.). Mobility and Inequality: Frontiers of Research from Sociology and Economics. Stanford: Stanford University Press. Green, A. (2002). The many faces of lifelong learning: recent education policy trends in Europe. Journal of Education Policy, 17(6):611–626. Gregg, P., Harkness, S., & Machin, S. (1999). Poor kids: trends in child poverty in Britain, 1968–96. Fiscal Studies, 20(2), 163–187. Gregg, P., & Macmillan, L. (2010). Family income, education and cognitive ability in the next generation: exploring income gradients in education and test scores for current cohorts of youth. Longitudinal and Life Course Studies, 1(3), 259- 280. Grusky, D. B., & Weeden, K. A. (2001). Decomposition without death: A research agenda for a new class analysis. Acta Sociologica, 44(3), 203–218. Hagenaars, J. A., & McCutcheon, A. L. (Eds.). (2002). Applied latent class analysis. Cambridge: Cambridge University Press. Hansen, K. & Vignoles, A. (2005). The United Kingdom education system in a comparative context. In What’s the Good of Education? The Economics of Education in the UK, chapter 2. Princeton: Princeton University Press. 31 Heath, A., Sullivan, A., Boliver, V., & Zimdars, A. (2013). Education under New Labour, 1997–2010. Oxford Review of Economic Policy, 29(1), 227–247. Jackson, M. (2013). Determined to succeed? Performance versus choice in educational attainment. Stanford: Stanford University Press. Jæger, M. M. (2007). Educational mobility across three generations: The changing impact of parental social class, economic, cultural and social capital. European Societies, 9(4), 527–550. Johnson, P., & Webb, S. (1993). Explaining the growth in UK income inequality: 1979–1988. The Economic Journal, 103(417), 429–435. Lupton, R., Heath, N., & Salter, E. (2009). Education: New Labour's top priority In J. Hills, T. Sefton, & K. Stewart (Eds.). A More Equal Society? Assessing Policy Towards Poverty, Inequality and Social Exclusion 1997–2008 (Vol. CASE Studies on poverty, place and policy. pp. 71–90). Bristol, UK: Policy Press. Machin, S. & Vignoles, A. (2006) Education policy in the UK. Technical report, Centre for the Economics of Education, LSE. Machin, S., & Vignoles, A. (Eds.). (2005). What's the Good of Education? The Economics of Education in the UK. Princeton: Princeton University Press Mare, R. D. (1980). Social background and school continuation decisions. Journal of the American Statistical Association, 75(370), 295–305. Mare, R. D. (1981). Change and stability in educational stratification. American Sociological Review, 46(1), 72–87. Marshall, P. (2013). The Tail: How England’s schools fail one child in five – and what can be done. London: Profile Books. McIntosh, J., & Munk, M. D. (2009). Social class, family background, and intergenerational mobility. European Economic Review, 53(1), 107–117. 32 Menning, C. L. (2002). Absent parents are more than money: The joint effect of activities and financial support on youths' educational attainment. Journal of Family Issues, 23(5), 648–671. Muthen, B. (2001). Latent variable mixture modeling. In G. A. Marcoulides & R. E. Schumacker (Eds.). New developments and techniques in structural equation modeling, pages 1–33. Mahwah: Lawrence Erlbaum Associates. Office for National Statistics. The National Statistics Socio-economic classification (NS-SEC). Website name: Ons.gov.uk. Available at: https://www.ons.gov.uk/methodology/classificationsandstandards/otherclassifi cations/thenationalstatisticssocioeconomicclassificationnssecrebasedonsoc201 0. Accessed 19/11/2018: Perry-Jenkins, M., Goldberg, A. E., Pierce, C. P., & Sayer, A. G. (2007). Shift work, role overload, and the transition to parenthood. Journal of Marriage and Family, 69(1), 123–138. Raffe, D., Brannen, K., Fairgrieve, J., & Martin, C. (2001). Participation, Inclusiveness, Academic Drift and Parity of Esteem: a comparison of post- compulsory education and training in England, Wales, Scotland and Northern Ireland. Oxford Review of Education, 27(2), 173–203. Schoon, I. (2010). Childhood cognitive ability and adult academic attainment: evidence from three British cohort studies. Longitudinal and Life Course Studies, 1(3):241–258. Shavit, Y., Arum, R. & Gamoran, A. (2007). Stratification in higher education: A comparative study. Stanford: Stanford University Press. Shepherd, P., (2001). NCDS/BCS70 1999-2000 Follow-ups Guide to the Combined Dataset. Joint Centre for Longitudinal Research. Available at: https://cls.ucl.ac.uk/wp-content/uploads/2018/05/guide.pdf. Accessed 02/12/2018. Sullivan, A., Ketende, S., & Joshi, H. (2013). Social class and inequalities in early cognitive scores. Sociology, 47(6), 1187–1206. 1 1 Table 2: Descriptive statistics:: row percentages unless otherwise specified Academic qualifications 1970 cohort 1989/90 cohort (weighted values) None Level 1 Level 2 Level 3 Total None Level 1 Level 2 Level 3 Sex Male 24.97 13.54 44.94 16.55 100.00 8.70 37.64 20.38 33.28 100.00 Female 20.69 13.19 48.52 17.60 100.00 5.35 31.99 20.84 41.82 100.00 Total 22.77 13.36 46.78 17.09 100.00 7.05 34.85 20.60 37.50 100.00 N 8480 12264 Social class Routine occupations 34.18 18.72 39.29 7.80 100.00 19.21 53.22 14.61 12.96 100.00 Semi-routine occupations 31.84 17.09 45.31 5.76 100.00 12.63 48.52 18.66 20.19 100.00 Lower supervisory occupations 29.22 16.39 47.51 6.89 100.00 8.39 52.94 20.86 17.81 100.00 Small employers 25.29 15.69 45.69 13.33 100.00 5.41 40.30 22.09 32.19 100.00 intermediate occupations 23.77 15.24 50.66 10.33 100.00 4.04 34.60 25.41 35.95 100.00 Lower managerial and professionals occupations 15.41 9.95 49.23 25.41 100.00 2.93 26.14 22.50 48.43 100.00 Higher managerial and professional occupations 9.29 5.52 44.53 40.66 100.00 2.27 14.29 18.47 64.97 100.00 Total 22.86 12.56 48.30 16.28 100.00 6.26 34.34 20.88 38.52 100.00 N 7638 11214 Social status (mean) -0.33 -0.38 -0.10 0.50 -0.10 -0.52 -0.15 0.32 0.74 0.26 N 7682 11686 Parental education No qualifications 30.55 19.18 43.83 6.45 100.00 18.54 49.38 14.67 17.41 100.00 Level 1 27.75 15.76 48.43 8.06 100.00 11.07 53.82 17.43 17.68 100.00 Level 2 20.82 12.52 49.69 16.98 100.00 5.53 40.94 24.55 28.98 100.00 Level 3 15.53 8.82 54.61 21.05 100.00 3.39 25.98 23.78 46.85 100.00 Level 4 8.70 3.98 43.18 44.14 100.00 2.19 11.58 17.79 68.45 100.00 Total 22.48 13.33 46.79 17.40 100.00 7.01 34.62 20.62 37.76 100.00 N 7771 12029 Family income (percentile) I group (7%) 29.27 19.27 42.93 8.54 100.00 15.23 42.72 15.87 26.18 100.00 II group (30%) 30.00 15.39 45.30 9.30 100.00 13.15 45.30 18.57 22.98 100.00 III group (34%) 22.07 14.36 50.29 13.28 100.00 5.37 39.78 21.63 33.22 100.00 IV group (29%) 18.14 10.63 47.22 24.00 100.00 2.11 19.64 22.04 56.21 100.00 Total 22.50 13.29 47.10 17.11 100.00 7.12 34.52 20.53 37.82 100.00 7283 9468 Table 3 Transitions through three academic levels by age 20/21 by cohort, parental class, social status and education, and family income: Main effects, binary logistic models, odds ratios Level 1 and higher Level 2 and higher Level 3 vs No qualifications vs Level 1 vs Level 2 Cohort (Ref.: 1970) 1989/90 cohort 3.93*** 0.24*** 4.37*** (0.265) (0.013) (0.230) Female 1.45*** 1.42*** 1.20*** (0.079) (0.063) (0.058) Parental Class (Ref.: Routine occupations (VII)) Semi-routine occupations (VI) 1.19* 1.11 0.90 (0.119) (0.103) (0.125) Lower supervisory occupations (V) 1.32** 0.98 0.86 (0.156) (0.099) (0.131) Small employers and own account workers (IV) 1.68*** 1.35*** 1.28* (0.219) (0.144) (0.189) Intermediate occupations (III) 1.80*** 1.52*** 1.06 (0.171) (0.145) (0.140) Lower managerial and professional occupations (II) 2.17*** 1.50*** 1.30* (0.238) (0.149) (0.174) Higher managerial and professional occupations (I) 2.76*** 2.00*** 1.71*** (0.427) (0.248) (0.248) Family social status 1.07** 1.16*** 1.13*** (0.030) (0.031) (0.031) Parental education (Ref.: No qualifications) Level 1 1.09 1.04 0.97 (0.082) (0.075) (0.096) Level 2 1.47*** 1.63*** 1.36*** Lower managerial and professional occupations (II)*1989/90 cohort 1.19 0.97 1.01 (0.337) (0.202) (0.275) Higher managerial and professional occupations (I)*1989/90 cohort 0.87 1.18 1.31 (0.336) (0.315) (0.385) Family social status 1.00 1.09** 1.12*** (0.030) (0.042) (0.042) Family social status*1989/90 cohort 1.15* 1.11** 1.04 (0.091) (0.060) (0.062) Parental education (Ref.: No qualifications) Level 1 1.01 1.32*** 1.08 (0.081) (0.132) (0.157) Level 2 1.28*** 1.86*** 1.95*** (0.106) (0.196) (0.241) Level 3 1.90*** 2.68*** 1.97*** (0.241) (0.420) (0.282) Level 4 and higher 2.43*** 5.17*** 3.70*** (0.327) (0.919) (0.483) Level 1*1989/90 cohort 1.43* 0.65*** 0.64** (0.266) (0.096) (0.140) Level 2*1989/90 cohort 1.59*** 0.78* 0.41*** (0.278) (0.111) (0.077) Level 3*1989/90 cohort 1.36 0.86 0.57*** (0.343) (0.163) (0.119) Level 4 and higher*1989/90 cohort 1.24 0.95 0.46*** (0.379) (0.207) (0.096) Family income (Ref.: I group (7%) II group (30%) 0.81 1.18 0.98 (0.114) (0.195) (0.227) III group (34%) 1.02 1.17 0.94 (0.144) (0.193) (0.214) IV group (29%) 1.08 1.23 1.25 (0.164) (0.219) (0.288) II group (30%)*1989/90 cohort 1.71** 0.99 0.95 (0.418) (0.205) (0.279) III group (34%)*1989/90 cohort 2.06*** 1.09 1.09 (0.531) (0.225) (0.315) IV group (29%)*1989/90 cohort 3.01*** 1.68** 0.92 (0.903) (0.370) (0.268) Constant 1.66*** 1.91*** 0.16*** (0.248) (0.346) (0.039) Observations 15,466 13,524 9,797 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1