Expansion and social selection in education in England and Scotland

Abstract

This paper examines trends in social class inequalities in young people’s educational attainment and HE entry between the mid‐1980s and the end of the 1990s in England and Scotland. Using time‐series data derived from the Scottish School Leavers Surveys and the England (and Wales) Youth Cohort Study, changes in both absolute and relative social class differences within and across the two countries were analysed through the use of a series of ordered logits. The results show that Scotland has higher educational attainment rates but also higher social class inequalities than England. Moreover, while in England social class inequalities at upper‐secondary and tertiary level have declined over time, in Scotland no evidence of such trend has been found. The conclusions highlight that possible explanations for these patterns reside in the different features of the two education systems and in the remarkable educational success of the Scottish middle class.

Acknowledgements

This paper is a product of the research project on Education and Youth Transitions in England, Scotland and Wales, 1984–2002, supported by the UK Economic and Social Research Council (R000239852). This paper has benefited from comments received from Richard Breen, Lindsay Paterson and David Raffe.

Notes

1. We had information about Welsh young people in the data but we had to omit them because the sample was too small to be analysed.

2. We have investigated whether men and women have differentiated patterns of social class inequalities in educational attainment. Our findings show that in both countries educational expansion has mostly benefited women but that social class differences in educational attainment do not vary by gender. For this reason the results presented in this paper refer to the whole sample.

3. YCS1 data (the year 1984) also do not cover independent schools. Moreover, the coding of social class in these data was different from, and cruder than, that used in the later surveys. For these reasons the data of this time point in England have been omitted from the analyses presented in this paper.

4. We would have liked to have dealt better with the problem of missing information in the social class variable. However, the best available predictor of missing information on this variable would have been our dependent variable, respondents’ educational attainment. So we could not specify a reliable model for the mechanism causing the missing data which would have not been confounded with our outcome. Respondents with missing information on social class appear in the category of the ‘unclassified’ as the analysis will show, this category is very likely to be composed of young people from lower social backgrounds.

5. There are two potential problems in relation to the information on respondents’ highest educational attainment. The first relates to the Scottish surveys data about vocational qualifications: this information is limited and was inconsistently collected. This may lead to an underestimation of young Scots who achieved post‐16 vocational qualifications. The second potential problem relates to the English surveys data. At 18–19 many more English young people were still studying for a non‐advanced qualification (5–6% of them for an A‐level qualification) than their Scottish counterparts (see Iannelli, 2007, for further details). If they subsequently entered HE we may underestimate the number of people in HE in England.

6. There can also be a problem that the data might be heteroscedastic. In our models we assume that the variance of the dependent variable, i.e., educational attainment, is the same (homoscedastic) across all levels of the independent variable, i.e., social class of origin. We tested whether this assumption was true in our English and Scottish data in the following way. We first estimated the residual variance for each social class of origin within country by calculating the simple variance of the residuals for each class (across years) within country. We then used F‐tests to compare these variances (across classes) within country. The results showed that the residual variance was significantly different for only one class: in England, the variance was much higher for young people who originated in the professional/managerial class than for any other class versus each of the other three classes. An investigation of the residuals showed that this was due to one outlier in the residuals of the professional/managerial class. Taking out this outlier, the only significant difference remaining was in the comparison between those originating in the professional/managerial classes and those from the ‘unclassified social class’. The results of these investigations led us to conclude that the assumption that the residual variance is the same across social classes is not seriously violated.

7. The modelling presented in Table 3 assumes that the effect of country and social class is the same across the different educational attainment levels (the thresholds); this may not be the case. We have tested for violation of the proportional odds assumption using the Brant test in STATA (Long & Freese, 2006) and we have found in our data that the assumption of proportional odds is violated. However, despite the violation of the proportional odds assumption for certain variables in the model, the main conclusions of the ordered logit model presented in Table 3 are confirmed (see Table A.3 in Appendix 2).