Download Impact of Social Origins on Educational Attainment: A Multidimensional Approach and more Study notes Statistics in PDF only on Docsity!
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
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
sociologists, social origins have been seen through a one-dimensional lens, f o c u s i n g primarily on parental social class (Breen et al., 2009; Goldthorpe, 2016; Shavit, Arum, & Gamoran, 2007). The notion of social class identifies the worker’s position within the relationships of production in firms and labour markets (Erikson & Goldthorpe, 1992; Erikson, Goldthorpe, & Portocarero, 1979). Parental education h a s been also included to complement the definition of social origins, especially when the outcome of interest is the child’s education. Economists h a v e mostly used family income instead of parental social class, as for example in the literature on educational inequalities in Britain (Blanden, Gregg & Machin, 2005 ; Blanden & Macmillan, 2016; Gregg & Macmillan, 2010). Others suggested that the predominantly one- dimensional treatment of social origins is inadequate and might partly explain the divergent findings on trends in educational inequalities (Bukodi, Erikson, & Goldthorpe, 2014; Bukodi & Goldthorpe, 2013 ; Goldthorpe, 2013; Jæger, 2007). Jæger ( 2007 ), following Bourdieu (19 84 ), p r o p o s e s that in analysing the effect of social origins on educational outcomes, social class should be accompanied by other factors aiming at capturing the resources that social class might proxy – parental economic, cultural and social capital. In this way he argues the effect of social class can be decomposed into more specific effects.While Bukodi and Goldthorpe ( 2013 ) endorsed the multi-dimensional treatment of social origins, they criticised Jæger’s approach from a Weberian perspective. In particular they contend that commonly used social class schemata such as such as the Erikson-Goldthorpe-Portocarero (EGP) or the Comparative Analysis of Social Mobility in Industrial Nations (CASMIN) (Erikson & Goldthorpe, 1992; Erikson et al., 1979) are not designed to proxy other kinds of resources, but rather are intended to
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 (201 3 ) 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, 201 3; Erikson & Jonsson, 1996 ;). Although they did also find some very small independent effect of family income in one cohort, Bukodi and Goldthorpe (201 3 ) conclude
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
academic qualifications were reached by 58% of the more recent cohort. The previous cohort’s attainment of level 2 appears higher (64%), but this could reflect an overstatement of GCSEs qualifications as discussed later. Table 2 Around here In the 1980s, participation i n post-compulsory education in the UK was low by international standards. In an attempt to raise it, two other major policies were introduced. The first was designed to enhance the labour market value of vocational qualifications. The second was the introduction of an Education Maintenance Allowance (EMA), which paid individuals from disadvantaged backgrounds a small means-tested allowance if they stayed on in full-time education beyond the age of 16. Evaluations of the EMA suggest that the subsidy increased participation not only in full- time education beyond the compulsory school leaving age but also in full-time education subsequently. Started in 1999 on a pilot basis, EMA was rolled out throughout the UK in 2004 and would have been available to the Next Steps cohort. Research suggests that it is one the factors that have enhanced the post- compulsory participation (Dearden, Emmerson, Frayne, & Meghir, 2005). The most recent reforms of the vocational education system, the development of the General National Vocational Qualifications ( GNVQs), introduced in 1992 and National Vocational Qualifications (NVQs) from 1998, have not, overall, been successful in terms of enhancing the labour market value of vocational qualifications (Dearden, McIntosh, Myck. & Vignoles, 2002 ; Machin & Vignoles, 2006). The system of vocational training and qualifications in the UK
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.
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 (201 3 ) have dealt with the
6 Data and operationalisations
6.1 Data We use data from two cohort studies, the 1970 British Cohort Study (BCS70) and Next Steps, formerly the Longitudinal Study of Young People in England (LSYPE). The BCS70 has collected rich information from a sample of around 17 , 000 individuals, all of whom were born in one week in 1970 (Elliott & Shepherd, 2006). Subsequent surveys took place when the cohort members were aged 5, 10, 16, 26, 30, 34, 38 and 42 years. Our study sample comprises around 8, 500 study members who l i v e d i n E n gl a n d a t a g e 1 0 a n d participated in both the 10 - year survey (for the social origins indicators) and 30 - year survey (when education history was collected through self-reports). Next Steps is a cohort study of pupils in England born between September 1989 and August 1990 and their parents (or carers). Data were collected annually between 2004 and 2010 (waves 1–7), with data currently available up to wave 8 , collected in 2015 at age 2 5/26. A sample of around 15,800 members participated in wave 1. Next Steps uses a complex survey design to over-sample deprived areas, thus requiring the use of sample weights in order to restore population representativeness. The data on educational qualifications are taken from administrative records, the National Pupil Data (NPD) which were linked to the survey members by the Department of Education. This study sample comprises 12,264 individuals who participated in the 13/14 years-old survey and had non-missing values on educational attainment by age 20/21 (from the NPD).
6.2 Variables In operationalising cohort members’ highest educational attainment, we focus on academic qualifications (ie excluding vocational qualifications). The coding of educational qualifications reflects the step structure of the UK education system. Our sequence of qualifications has a baseline of no academic qualifications (level 0); Level (1) is attainments immediately above this GCSE grades D-G; and CSE grades 2- 5 ; Level ( 2) is /O-levels / CSE grade1/ and, for the second cohort, GCSE grades A*–C. Level (3) for both cohorts is A-levels (see table 1). Data on completed qualifications in the Next Steps are available up age 20/21, using NPD. In order to generate a comparable indicator for the BCS70, we used the history data of qualifications reported by the BCS70 members at age 30. From the retrospective self-reported information regarding qualifications, we were able to derive the A-level qualifications obtained at age 20. The definition of the qualifications implies that if cohort members have not gained A-levels by age 20 they are assigned a level 2 qualifications (if they have one) regardless of qualifications attained later on. It should be noted that in the BCS 30 - year-old survey, level 2 qualifications are likely to be biased upward. More than 1000 cohort members reported having obtained one or more GCSEs before the introduction of GCSEs examinations (1988) (Shepherd 2001 p 42), possibly due to confusion of CSE and GSCE qualifications (the former are more likely to have been level 1 than GCSE). On the other hand, in Next Steps, there is a possible small downward bias to records of Level 2+ qualifications due to under-reporting of students attending independent schools (personal communication, Dr Morag Henderson).The variables available in the datasets on parental social class allow us to code class origins in the Goldthorpe schema, seven-category version. The BCS70 contains
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”,
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 BCS 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
parents with level 1 qualifications do not have a significant relative advantage over parents with no qualifications. Family income shows some independent association with both earlier transitions, over and above the other social origins’ indicators. There is a significant difference between children from families at the bottom income group and children from families at the third and at top income group at transitions 1 and 2, At the third transition, the likelihood of success does not seem to be associated with family income. To assess the relative importance of the predictors, we looked at the proportion of correctly predicted cases in the full model and then remove one predictor at a time to assess how much predictive power is lost each time. The full model for progressing to A-levels correctly predicts 68% of cases; parental education is the most important variable (predicted cases drop to 67.2%), followed by income (predicted cases drop to 67.8%), social status (predicted cases drop to 67.9%) and social class (predicted cases do not drop). In summary, Table 3 suggests that when considered together, parental social class, social status, education and family income each exerts an independent effect on educational attainment. Consequently social origins indicators should be regarded as distinct aspects of social origins, with the implication that if one or more of them were missing, the total effect of social origins would be underestimated. We now turn to the question of whether the effects of parental class and education, social status and family income on children’s educational outcomes changed across the two birth cohorts, addressed by adding an interaction term between cohort and each of the social origins indicators into the logistic models.
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
