Origins versus Education 153 8 Origins versus Education: Are there 'Glass Floors' and 'Glass Ceilings'? In the two preceding chapters we have examined in some detail the associations that exist between individuals' social origins and their educational attainment and between their educational attainment and their class histories and eventual class destinations - that is, the associations on the OE and the ED sides of the OED triangle. The results we have presented show that the OE association has not weakened nor the ED association strengthened in the ways that would be expected if the expansion and reform of the British educational system over the last half-century or more had resulted in movement towards an education-based meritocracy. In the present chapter we turn to the third side of the triangle, the OD side. This relates to the 'direct' effect of origins on destinations or, that is, to that part of the association between origins and destinations that is not mediated via education. If an education-based meritocracy were emerging, then, with the weakening of the OE association and the strengthening of the ED association, the direct OD association should also weaken (see Figure 5.1). In other words, the overall association between origins and destinations should become increasingly mediated through education, while the direct effect falls away. Conversely, the persistence of a direct effect must imply that individuals' origins continue to qualify the effects of their education on their destinations; and, insofar as this is so, the degree to which any reduction in social inequalities in educational attainment is translated into a weaker overall OD association - that is, into greater social fluidity - will be restricted. 'Glass floors' will prevent children of more advantaged origins but with relatively poor educational attainment from being downwardly mobile, while 'glass ceilings' will limit the upward mobility of children from less advantaged origins who have performed well educationally. In Chapter 5 we noted previous research that has indicated that in Britain the direct effect of social origins - or DESO in the usual sociological abbreviation - has remained remarkably stable over time, and we reported results consistent with this conclusion from our own analyses of data from the 1946,1958 and 1970 birth cohorts.1 Then in Chapter 7 we have also shown how, across these cohorts, individuals' social origins have a persisting effect on the class trajectories that they follow, independently of the level of their educational qualifications at labour market entry. However, we now wish to take the question of DESO further in three different respects. First, we examine how DESO is affected if we consider social origins not simply in terms of class but in the more comprehensive way that we introduced in Chapter 6 in regard to social inequalities in educational attainment: that is, if we take into account, in the way there described, parental status and education in addition to parental class. If this is not done, it could be that DESO will be underestimated. Second, though, we also move to a larger view of educational attainment and take into account the further qualifications that individuals may obtain after labour market entry - that is, in the course of their working lives. If this is not done, it could be that DESO will be otwestimated. And third, having arrived at our best estimates of DESO, we seek to gain some better idea of how DESO comes about. To refer to the direct effect of social origins on destinations, in the sense of the effect that is not mediated via education, is in fact to refer to an unexplained effect One of the papers previously cited, Vandecasteele (2016), appears in a collection (Bernardi and Ballerino, 2016) in which the extent of, and changes in, the direct OD association arc examined in fourteen modern societies. It is found that whether origins and destinations are considered in terms of income, socioeconomic status or class, a significant direct association is in all cases present. This association is weakening in only two cases, is strengthening in two, and stable over time in the rest. A recent British study (Sullivan et al., 2017), drawing on the 1970 birth cohort dataset, claims to show that there is no residual effect of social origins on destinations once education is taken into account. However, several limitations of this study have to be noted. First, it is concerned only with access to NS-SEC Class 1 and not with the association between class origins and destinations more generally; second, class origins are not likewise treated on the basis of NS-SEC but on that of the now outmoded Registrar General's Social Classes, which very likely underestimates their effect; and, third, the definition of 'education' is extended to include not only qualifications obtained but also type of school and university attended. But the effect on class destination of attending a 'Tatler public school' or a Russell Group university over and above the level of qualification obtained could as well be seen as an effect of social origins as of educational attainment per se. 154 Social Mobility and Education in Britah Origins versus Education 155 of social origins. If education is not the mediating factor, then what other factors are involved? These more detailed concerns do, however, mean that, because of data considerations, we have to limit our analyses to the 1970 birth cohort. It is for this cohort that we have the fullest information on the nature and timing of qualifications gained after labour market entry; and it is only for members of this cohort that we have information about whether they have ever received any kind of direct help from their parents in searching for and obtaining jobs - which is of course one possible way through which DESO could operate. In turn, because the numbers in our analyses are thus reduced, we work with only a threefold version of NS-SEC, in which Classes 1 and 2 and Classes 6 and 7 are collapsed, as well as Classes 3,4 and 5, and we focus on the chances of individuals being found, at age 38, either in Classes 1 and 2, the managerial and professional salariat, or in Classes 6 and 7, the wage-earning working class. Further, as regards educational attainment at labour market entry and again at age 38, we form, on the basis of a detailed classification of cohort members' academic and vocational qualifications, three relative qualification levels, rather than the four we have previously used, with qualifications being relativised according to age at labour market entry.2 We label these levels, which cover roughly equal numbers of individuals, as low (comprising no, or only sub-secondary, qualifications) intermediate and high (comprising higher secondary and tertiary qualifications). And finally, since preliminary analyses revealed no significant gender effects, we treat men and women together, although in the analyses on which we report gender and also part-time working are always included as control variables. Table 8.1 derives from analyses in which we fit three successively more inclusive statistical models to data for the 1970 cohort, and shows significant positive and negative effects on the chances of being found at age 38 either in the salariat or in the working class for 2 The logic of this relativising of qualification level by age at labour market entry is the same as that of relativising across birth cohorts, i.e. that what mattets as regards labour market returns is not just how much education individuals have but how much relative to those with whom they are in most direct competition. Five age groups are used covering the entire age range from 16 to 38. Further details can be found in Gugushvili, Bukodi and Goldthorpe (2017: Online Appendix A). Table 8.1 Effects of social origins and of change in relative level of qualification on class position at age 38, by relative level of qualification at labour market entry Relative level of qualification at labour market entry Low Intermediate High Class at age 38 Class at age 38 Class at age 38 1 and 2 6 and 7 1 and 2 6 and 7 1 and 2 6 and 7 Model 1 Class of origin Classes 6 and 7 Classes 3, 4 and 5 Classes 1 and 2 — ++ reference ++ -- ns + reference + — ns + reference + ns Model 2 Class of origin Classes 6 and 7 Classes 3, 4 and 5 Classes 1 and 2 ns + reference + ns ns + reference ns ns ns + reference + ns Parental status Parental education ns -- ++ -- + ns ++ -- ns ns ++ ns Model 3 Class of origin Classes 6 and 7 Classes 3, 4 and 5 Classes 1 and 2 ns + reference + ns ns + reference ns ns ns + reference + ns Parental status Parental education ns -- ++ -- + ns ++ -- ns ns ++ ns Change in relative level of qualification Worsened Did not change Improved reference ++ -- — + reference ++ -- -- + reference Note (a) For symbols, see Table 6.1 Source: Gugushvili, Bukodi and Goldthorpe (2017) 156 Social Mobility and Education in Brit, Origin versus Education 157 individuals who entered the labour market at the three relative qualification levels that we distinguish.s With Model 1 the only explanatory variable included is that of parental class. It can be seen that this has rather systematically significant effects. Compared with individuals of intermediate class origin (the reference group), individuals of salariat origin have a higher probability of being themselves found in the salariat, and individuals of working-class origin a higher probability of being themselves found in the working class regardless of their level of qualification at labour market entry. And for individuals with low- or intermediate-level qualifications, being of salariat origin also has a negative effect on their risk of ending up in the working class. In other words, there is here clear evidence of DESO being in general operation and in creating glass ceilings and, especially, glass floors. With Model 2 we introduce parental status and education in addition to parental class. It can be seen that the effects of parental class now in some instances weaken or indeed become non-significant, although they still play a part in maintaining intergenerational immobility. Parental status is of rather limited importance - perhaps the main point of interest to emerge is that it is now higher parental status, rather than more advantaged parental class, that appears to create a glass floor protecting individuals with low entry qualifications from ending up in the working class. But parental education proves to have rather consistent effects. Higher parental education has a positive effect on the chances of individuals at all qualification levels accessing the salariat, and a negative effect on the risks of those with low and intermediate levels of qualification being found in the working class. These results do then indicate that DESO will tend to be underestimated if only parental class is considered. As we argued in Chapter 6, although parental class, status and education are correlated, the correlations are not perfect, and cumulative effects may therefore be expected in some degree to arise. We take this matter further in Figure 8.1. In the left-hand panel of the figure we show the probabilities, under Model 1 of Table 8.1, of men of salariat, intermediate and working-class origins being found in the salariat and in the working Full details of the models - known as linear probability models - and of the construction of the variables included are given in Gugushvili, Bukodi and Coldthorpe (2017). class in relation to each of the three qualification levels that we distinguish at labour market entry. Then in the right-hand panel we show, under an analogous model, the corresponding probabilities for men allocated to the three parental groups defined by combinations of class, status and education that we introduced in Chapter 6. It is evident that the effects of social origins are greater in the latter case than in the former - and results for women are on much the same pattern.4 In other words, as regards the first issue that we raised, we can directly confirm that DESO will be underestimated insofar as social origins are not treated in a sufficiently comprehensive way. At the same time, though, what also emerges from Figure 8.1 is that the differences resulting from DESO are more marked, the lower the level of qualification - and this shows up especially strongly in the right-hand panel. Thus, it can be seen that for individuals with low-level qualifications the difference between those with parents in the least and most advantaged groups is up to almost 30 percentage points in their chances of being found in the salariat and to over 30 percentage points in their risks of being found in the working class. These are disparities of a rather remarkable order, testifying to the actual extent to which the experience of men and women in the 1970 birth cohort diverges from what would be expected if an education-based meritocracy had been brought into being. They also reinforce previous indications that DESO contributes more to the creation of glass floors than glass ceilings. This is an outcome much in line with the importance that we have attached to the particular concern of parents and their children - stemming, we suggest, from the psychology of loss aversion - to avoid intergenerational downward mobility. Where the children of more advantaged parents do not perform well educationally before entering the labour market, other factors would appear to operate that still give them a good chance of maintaining their parents' position or, at all events, of not falling too far below it. Moving on now to the second issue we wish to address - that of the effect on DESO of taking account of qualifications gained after labour market entry - we introduce in Model 3 in Table 8.1 a further variable that we have constructed specifically to allow for change in individuals' relative qualification level in the course of their working lives. We compare each cohort member's relative qualification level at labour See further Gugushvili, Bukodi and Goldthorpe (2017: Online Appendix B). B « a g 00 g - "2 *a T3 a, 5 " 2 £ .SP 60 -S. .9 c