CLASS ORIGINS, EDUCATION AND OCCUPATIONAL ATTAINMENT IN BRITAIN: Secular trends or cohort-specific effects?

One evident advantage of the ‘status attainment’ approach to social mobility research (e.g., Blau and Duncan 1967; Featherman and Hauser 1978) was that it allowed for inter- and intragenerational mobility to be treated in an integrated, if rather schematic, way. The status of ‘first occupation’ (i.e., on entry into the labour market) served as a key intervening variable, following on that of ‘years of education’, in causal path models aiming to link individuals’ social origins to the status of their current or last occupations. More recently, however, studies of intergenerational mobility, as carried out within a class structural context, and of intragenerational mobility, as carried out on the basis of detailed occupational histories, have tended to move apart from each other – focussing on different substantive issues and using different kinds of data-set and analytical technique (compare, for example, the papers collected in Breen 2004; Blossfeld et al.2006; Blossfeld and Hofmeister 2006). This divergence is unfortunate. In large part, we would suggest, it results from certain, often implicit, assumptions that unduly limit the attention that is given (1) in analyses of intergenerational class mobility to the part played by occupational mobility over the course of working life and (2) in analyses of occupational histories to the influence of social, and especially class, origins.

As regards (1), the underlying assumption is that a secular tendency exists, as a feature of ‘modernisation’, for education to become the ever-more dominant factor in whether individuals remain in or move away from their class of origin. It is within the educational system that individuals primarily acquire their human capital; and it is then their human capital, as indexed by their educational qualifications, that primarily determines not only the occupational level at which they enter the labour market but, further, their chances of subsequent mobility within it.¹ While it may be recognised that the effects of education on occupational level tend to weaken as individuals age, occupational mobility in course of working life is still in effect treated as in some large degree epiphenomenal: that is, as simply reflecting prior educational attainment rather than – as was more typical in pre-modern times – the acquisition of human capital in employment itself, and in such a way that might compensate for a lack of educational attainment or opportunity.

As regards (2), the underlying assumption is that a secular tendency exists, again as a feature of modernisation, for the influence of individuals’ social origins on their own work histories to be increasingly channelled via their education. Individuals’ class origins are recognised as a major influence on their educational attainment and as having thus an important ‘indirect’ effect on their life-chances within the labour market. But the importance of other, ‘direct’ effects of class origins is taken to be in decline, as employing organisations follow increasingly ‘meritocratic’ personnel selection policies in which educational qualifications are given predominant weight. If, then, in studies of work-life occupational mobility due attention is paid to individuals’ educational attainment, class origin effects will in this way be largely captured, and analyses can concentrate on the dynamics of occupational change per se without further reference to its intergenerational context.

The assumptions to which we refer could in some degree be defended as providing useful simplifications: no analysis of social phenomena can include all potentially relevant variables. None the less, there is growing evidence, deriving largely in fact from the research in which these assumptions are embedded, that they are now in need of some re-examination.

On the one hand, good grounds have emerged for querying whether the role of education in intergenerational mobility processes is in fact of steadily increasing importance. It would appear that in many advanced societies over recent decades the association between individuals’ educational qualifications and their class destinations has, if anything, tended to weaken (cf. Breen and Luijkx 2004). Moreover, while studies of work-life occupational mobility sensitive to the effects of ‘globalisation’ rather than of modernisation have emphasised the disadvantages experienced by individuals with only low levels of qualification (cf. Mills and Blossfeld 2006), they have at the same time pointed to a growing looseness and unpredictability in the transition from education to employment (see e.g., Blossfeld et al.2005, 2008). In particular, the range of occupations and forms of employment contract initially taken up by individuals with higher-level qualifications have become far more heterogeneous (Bukodi et al.2008). And it is by no means clear that ‘over-qualification’ occurring on entry into employment is then more or less automatically corrected through upward occupational mobility in later working life (see e.g., Büchel et al.2003; Büchel and Mertens 2004).

On the other hand, while some studies have lent support to the view that social origins essentially impact on individuals’ work-life occupational attainment via their education (e.g., Warren 2001; Warren et al.2002), others have shown that the effects of parental social class in particular still persist when education is controlled. In other words, it would appear that the intergenerational transmission of class inequalities continues to shape individuals’ occupational life-chances, not only through the creation of advantage or disadvantage in regard to educational attainment but in a range of other ways: for example, through the development of personality or subcultural attributes or of social networks that can also produce significant returns in working life (for Britain, see Breen and Goldthorpe 1999, 2001; Jackson et al.2005; Jackson 2006; Goldthorpe and Jackson 2008).

In sum, it would seem important in future research to cease to rely on supposed secular tendencies and to recognise, rather than discount, two possibilities. First, processes of work-life occupational mobility need not be shaped simply by the human capital that individuals first bring with them to the labour market via their education; these processes may themselves have some independent role in determining rates and patterns of intergenerational class mobility. Second, the influence of class origins on individuals’ work-life mobility can extend beyond its effects via their educational attainment – and in ways that may be difficult to reconcile with ideas of either human capital or ‘meritocracy’.

Taking this approach, arguments that imply secular tendencies in the role of education or of class origins in social mobility can then be explicitly set against arguments that would rather emphasise the possibility of ‘cohort-specific’ effects: i.e., effects following from the particular temporal relationship that certain birth cohorts have with historical events or conjunctures (Ryder 1965). As regards employment, a question of particular interest is that of whether, in cases where individuals’ early working lives coincide with adverse labour market conditions, a damaging kind of path dependency is thereby set up (‘hysteresis’) rather than some recovery occurring once labour market conditions improve (‘resilience’). The typical concern of economists in this regard has been with the possible ‘scarring’ effects of early unemployment on individuals’ future employment and earnings prospects (cf. Arulampalam et al.2001). But of greater relevance to our own wider concerns with the early experience of recession, whether involving unemployment or not, is the approach taken by economists such as Moscarini and Vella (2008). These authors suggest that in depressed labour markets a ‘noisier’ sorting of workers across jobs tends in general to occur, so that individual comparative advan tage becomes less relevant to occupational choice – and with implications for the level of returns that qualifications bring. If such wide-ranging and lasting cohort-specific effects do impact on work-life mobility processes, then even in the presence of forces making for secular tendencies in these processes, the overall outcome may still prove to be one of merely ‘trendless fluctuation’.

In this paper, we aim to make a start in pursuing the research programme indicated above by examining aspects of the occupational attainment in early-to-mid working life of men in three British birth cohorts. Two limitations of the paper will be obvious. The first is the exclusion of women. Significant gender differences exist as regards many of the issues that concern us, and these are the subject of a separate paper (Bukodi 2009). The second limitation is that, while we bring class origins into our analyses, we do not attempt the further step of linking work-life occupational mobility – as measured here in terms of occupational status and earnings (see further below) – to intergenerational class mobility. This is a matter that we will be better placed to consider when we have information available for men in each cohort at a later stage in their work histories and can thus establish class destinations more securely.² We do, however, believe that the linkage is an important one to make. Even if one's ultimate interest is in class mobility – because, say, of its known, wide-ranging consequences – it is occupational advancement in terms of status and pay that in the course of individuals’ working lives is likely to be subjectively salient and most immediately pursued.

The three British birth cohorts with which we are concerned are those covered by the Medical Research Council National Survey of Health and Development (NSHD), the National Child Development Study (NCDS) and the British Cohort Study (BCS). These studies aim to follow through their life-course children born in Britain in one week in 1946, 1958, and 1970, respectively (for further details, see Ferri et al.2003: Appendix 1). The NSHD has so far undertaken 18 data collections up to age 53; the NCDS, 7 up to age 46; and the BCS, 6 up to age 34.

In each case, the data-sets of these studies include recalled information, recorded in months, on respondent's previous jobs, on absences from employment, on the timing of job changes and on occupation in each job (for further details, see Bukodi and Neuberger 2009). For our present purposes, we consider these work histories for men from the point at which they left full-time education and first entered the labour market up to age 34 – i.e., the latest age for which we have information for respondents in all three cohorts. The data-sets also include detailed information on respondents’ social backgrounds and on their educational histories and attainment that we are able to exploit. In particular, it should be noted that in our analyses we treat education as time-variant. That is to say, if men in our cohorts attained a higher level of educational qualification at any point after their entry into the labour market, this is taken into account from that point onwards.

As with all longitudinal studies, the problem of missing data arises. All three studies have suffered from a considerable attrition of respondents from one sweep to another – although a number of individuals subsequently ‘return’ – and for each sweep there is also some amount of item non-response. However, various analyses of attrition and non-response have been undertaken and the results are encouraging in suggesting that no major biases are being created (Despotidu and Shepherd 1998; Nathan 1999; Hawkes and Plewis 2006; Wadsworth et al. 2006).

Occupational data in each study have been re-coded to the official British SOC90 classification (OPCS 1990). In proceeding from these data to establish a basis for treating occupational mobility of a ‘vertical’ kind, we follow a strategy that is set out at length elsewhere (Bukodi et al. 2011). The essential point is that, rather than relying on a single occupational scale of a ‘synthetic’ (or ‘composite’) kind, such as a scale of the socioeconomic status of occupations or of their ‘general desirability’, we work with two ‘analytical’ scales, each of which aims to order occupations within a specific and well-defined hierarchy. These are (i) the occupational status scale developed by Chan and Goldthorpe (2004), based on patterns of close friendship and intended to capture status in something close to the classic Weberian sense and (ii) an occupational earnings scale developed by Bukodi which is in effect an update of that produced by Nickell (1982) and ranks occupations on the basis of average hourly earnings rates for full-time employees, using data from the New Earnings Survey 2002. The latter scale provides a score for each of the 77 minor occupational groups distinguished in SOC90 but the Chan–Goldthorpe scale gives scores for only 31 occupational categories that are either these minor occupational groups or collapses thereof. Thus, for purposes of comparability, we convert scores on both scales into percentile distributions.³

A positive correlation between scores on the two scales does of course exist but the correlation is not all that strong (see further Bukodi and Goldthorpe 2009). There is a tendency for occupations associated with the manufacturing, construction and transport sectors to yield high earnings relative to their status, while the reverse applies for occupations associated with administration, sales, and personal services.

In Figure 1 we plot by age the average occupational levels attained by men in each cohort on our occupational status and occupational earnings scales. It may be noted, first of all, that the general shapes of the curves shown are very similar both for the three cohorts and the two scales. Over the period covered, men have tended to move upwards occupationally during the course of their working lives, in terms of both status and earnings, and at broadly similar rates.

At the same time, though, some differences can be observed. In the case of the status scale, Figure 1, Panel A, points to secular change in that, as one moves from the earliest to the latest cohort, there is a rise in average occupational level on entry into the labour market and then in the level attained at almost all ages up to age 34. In fact, on the status scale the distribution of jobs ever held by men up to this age reveals a fairly steady ‘up-grading’ across the cohorts which can in turn be linked to long-term changes in the structure of employment: in particular, to the growth of non-manual occupations which have a generally higher ranking than manual occupations on the status scale (see further Chan and Goldthorpe 2004).

However, in the case of the earnings scale, Figure 1, Panel B, reveals no similar trend. There is no great difference across the cohorts in men's average occupational level on entry into the labour market, and then, from around age 22, the occupational attainment of men in the 1958 cohort falls below, and stays below, that of men in the 1946 as well as in the 1970 cohort – whose own trajectories are from this point almost identical. One relevant factor here is that on the earnings scale, in contrast with the status scale, there is no ‘upgrading’ of the jobs ever held by men across the three cohorts. If anything, some decline is apparent in the proportion of jobs in occupations with intermediate levels of earnings – for example, skilled manual jobs – consistently with the thesis of a ‘polarisation’ in employment shares in the UK between high- and low-pay work (Goos and Manning 2007).

In addition, though, the distinctive trajectory of the 1958 cohort may be associated with the fact that at the beginning of the 1980s Britain entered into a severe economic recession and a period of extensive ‘de-industrialisation’ and consequent re-structuring of the labour market. Male unemployment rates rose rapidly and remained at double-digit levels from 1981 through to 1988. Men in the 1958 cohort would then meet with these adverse conditions in the early years of their working lives or, in the case of those who had been longest in full-time education, at the very time of their entry into the labour market. Although these better educated men might still have had good chances of moving into relatively high status, non-manual employment – for example, associate professional or sales occupations – they would appear to have been less able than their counterparts in either the 1946 or 1970 cohorts to establish themselves in occupations that were also high paying ones – for example, managerial or technological occupations (see further Bukodi and Goldthorpe 2009).⁴

We now turn to a detailed analysis of the factors influencing the occupational level at which men in our three cohorts first entered the labour market. For this purpose, we use OLS regression models with the scores of men's first occupation on each of our two scales as the dependent variables.

In Table 1, Panel A, we show the results obtained with the status scale. It can be seen that, consistently with Figure 1, Panel A, occupational level at entry tends to rise across the cohorts. However, of main interest to us are the effects, and any changes in the effects, of educational qualifications and class origins. As shown, we treat qualifications on the basis of five ordered categories of highest qualification obtained, ranging from ‘less than O-level’ to ‘degree’, and class origins on the basis of a seven-class version of the Goldthorpe schema (Goldthorpe 1987, 1997).

From Model 1 in Panel A, it is evident that qualifications exert a strong influence on the status of the occupations in which men enter the labour market, and on an entirely expected pattern. And from Model 2, it is further evident from the education-by-cohort interaction terms that are introduced that what we might call the ‘status returns’ to qualifications on labour market entry tend to increase across the cohorts for all levels of qualification – except degrees. Degrees have the greatest status returns for men in the 1946 cohort – only 6 percent of whom had degrees as compared with 9 percent in the 1958 cohort and 17 percent in the 1970 cohort – while the difference in returns as between men in the two later cohorts is not significant.

Turning to class origins, it can be seen from Model 1 that these also have independent and quite strong effects, and again on an unsurprising pattern. From Model 2, it can then further be seen that while the advantage of originating in the professional and managerial salariat, as represented by Classes I and II, is most marked for men in the 1946 cohort, no significant difference occurs in this respect as between men in the 1958 and 1970 cohorts.⁵

These findings are set in relation to each other in Figure 2 which shows predicted status scores of first occupation by qualifications and salariat origins under a regression model that includes the same variables as in Model 1, but also terms for the interaction of qualifications and salariat origins, and that is fitted separately for each cohort. The stronger effect of qualifications relative to that of class origins, at least in the binary form here used, is apparent. But what also emerges is that the overall differentiating effect of qualifications is greater with the 1946 cohort than with the two later ones – at around 55 points on the status scale as against 40–45 points. And a further feature of interest is that while with both the 1946 and 1970 cohorts the effects of coming from a salariat background diminish as the level of qualification rises, this is not the case with the 1958 cohort. In this cohort, it is men with A-levels and sub-degree tertiary qualifications who appear to benefit most from advantaged social origins.

We may then say that while the results of our analyses based on the status scale are in some part consistent with the idea of education becoming of dominant importance in determining the level of first occupations, there are also, from this point of view, a number of anomalies. The returns to degrees are greatest with the earliest cohort, as in turn is the overall range of the effects of qualifications; the independent effects of class origins do not significantly weaken between the two later cohorts; and the relationship between the effects of qualifications and class origins is rather distinctive for the 1958 cohort. Is the picture at all clarified when we turn to analyses based on the earnings scale?

Panel B of Table 1 gives the results of these analyses. It can be seen that, in line with Panel B of Figure 1, there is no tendency for the level of first occupations to rise across the cohorts. Men in the 1958 cohort tend to enter occupations with lower levels of pay than do men in the two other cohorts.⁶

As regards qualifications, Model 1 shows that these exert a similarly large effect on level of first occupation when using the earnings scale as when using the status scale. However, Model 2 reveals a difference. There is no general tendency for the occupational earnings returns to qualifications (apart from degrees) to increase across the cohorts, as in the case of status. Rather, there is further evidence of a specific 1958 effect. For men in the 1958 cohort all levels of qualification, but especially higher levels, give lower earnings returns than for men in the other two cohorts.

Furthermore, turning to class origins, our main finding is that while under Model 1 their effects appear less strong – though remaining significant – than when using the status scale, Model 2 reveals that with the earnings scale it is men in the 1958, rather than the 1946, cohort who gain most in coming from an advantaged background within the professional and managerial salariat.

Figure 3, graphing predicted scores of first occupations on the earnings scale, is based on a model with the same explanatory variables as Figure 2. Two similarities with Figure 2 show up. As here measured, qualifications have clearly stronger effects than do class origins, with the overall range of the effects of qualifications being again wider for men in the earliest, 1946 cohort than for men in the two later cohorts; and the graph for the 1958 cohort is again rather specific – although in different ways than with the status scale. First, the relatively low returns to degrees and also to sub-degree tertiary qualifications are clearly brought out. Second, it can be seen that while for men in the 1946 and 1970 cohorts, salariat origin effects, given level of qualification, are very small – and especially as compared with our results using the status scale – for men in the 1958 cohort the difference made by salariat origins on level of first occupation is larger: i.e., at around 5–10 points at all levels of qualification above O-levels and equivalent.

Our results using the earnings scale would tend therefore to reinforce our scepticism concerning secular trends, and especially in the role of education in determining occupational level at entry into the labour market. Although educational qualifications are of large importance in this regard, there is no indication that their importance is steadily increasing, and there is no indication that the effects of class origins are in steady decline. Most notable in fact is the way in which results based on our earnings scale support those based on our status scale in suggesting that the experience of men in the 1958 cohort is in some degree distinctive. For these men, qualifications give clearly lower returns than for men in the earlier and later cohorts in terms of the earnings levels of their first occupations, while their class origins appear in this respect to matter more. An obvious conjecture is that for men entering the labour market in hard times, those from relatively advantaged class backgrounds will draw on the greater resources – economic, cultural, or social – that are available to them in order to compensate for the at least temporarily reduced economic value of their qualifications.

We now wish to extend our analyses so as to follow the men in our three cohorts from the occupational level of their first occupation to that they attain at some later stage in their working lives. The question thus arises of how this later stage might best be defined. As earlier noted, our data cover men's occupational histories from their first entry into employment up to age 34, the latest age for which we have information for men in all three cohorts. One solution would then be simply to focus on occupational level at age 34. However, this may not be the best solution. We know that in the early years of working life the frequency of job changing tends to be relatively high before falling off, usually somewhere between ages 30 and 40, and afterwards being less likely to involve significant changes of occupational level (cf. Figure 1). But it may also be supposed that men with differing characteristics will tend to reach this stage of what might be called ‘occupational maturity’ (cf. Goldthorpe 1987: 52–3) somewhat earlier or later in their working lives. Thus, age 34 is one at which men could be quite heterogeneous in this respect. Some will have already achieved occupational levels, whether high or low, that any further job changes are unlikely to affect, while others will still be in a period of their working lives in which occupational changes could occur leading to significant upward or downward mobility. We proceed therefore on the following lines.

Kaplan–Meier ‘survival’ estimates under our model (Bukodi and Goldthorpe 2009) show that up to around age 30 very few men in any cohort appear as having reached occupational maturity but that, after this age, the proportion increases rather sharply. However, men in the 1946 cohort would appear to reach maturity relatively slowly, over a third not having reached this stage by age 34 as against a fifth in the two later cohorts.⁸ This result does then strengthen our view that simply comparing men across cohorts at age 34 would not, in a sociological sense, be comparing like with like. And so too does the finding from further analyses (not here shown but available on request) that men we would regard as having reached maturity have significantly higher levels of occupational attainment at age 34, on both the status and the earnings scale, than do those we would regard as not having reached this stage.

In analysing the determinants of the occupational level attained by men in later working life, we therefore opt to do this not at age 34 but rather at the stage of occupational maturity as predicted under our model. It could be thought a disadvantage that in this way we discard a good deal of information – i.e., on those men we treat as not having achieved occupational maturity – and especially with the 1946 cohort. However, we have analysed the occupational attainment of these men, at age 34, under the same model as we apply in the following section, and the results reveal some significant differences that we comment on in notes to the text.

We here revert to OLS regressions models with the dependent variable being scores at occupational maturity on our two scales for those men we would predict to have reached this stage. We begin with results using the status scale, as shown in Table 2, Panel A.

An initial point to note here is that while the general level of first occupation as measured on our status scale increased across the cohorts, this is less clearly the case with level at maturity. Rather, a 1958 effect is again indicated in that men in this cohort have significantly lower levels of occupational status attainment overall.

Turning to what might be called the work-life variables, we can see that these are in fact of some enduring importance. As might be expected, status of first occupation has a significant positive effect on status at maturity but so too does the number of occupations that an individual has held between entry and maturity.⁹ And, moving from Model 1 to Model 2, we find no indication that these effects are declining across the cohorts.

In the case of qualifications, the main effects are, just as with status of first occupation, strong and on an entirely expected pattern.¹⁰ However, the interaction effects of qualifications with cohort, as shown under Model 2, are different. With first occupation, status returns to qualifications, it will be recalled, increased across the cohorts for all levels of qualification except degrees, which provided the greatest returns for men in the 1946 cohort. But at occupational maturity this trend is no longer apparent. The only significant changes are those indicating a 1958 effect. For men in this cohort secondary, and in particular A-level, qualifications give lower status returns than for men in the 1946 as well as in the 1970 cohort.¹¹

Finally, as regards class origins, it can be seen that these are also quite strong, and on the same unsurprising pattern as we found with status of first occupation. But Model 2 reveals that here again a difference arises with the cohort interaction effects. Men in the 1946 cohort who came from a salariat background benefited significantly more on entry into the labour market than did their counterparts in the two later cohorts; but at maturity this relative advantage has disappeared and now the positive effect of salariat background does not differ significantly by cohort.¹²

To bring out more clearly some of the implications of these findings, we present in Figure 4 predicted occupational status scores at maturity under a regression model that includes the explanatory variables of Model 1 plus terms for the interaction of qualifications (as at maturity) and first occupation and qualifications and salariat origins, and that is fitted separately for each cohort.

Again, the greater importance of qualifications as compared with class origins, as we here measure these variables, is apparent. But the overall differentiating effect of qualifications is no longer stronger, as it was at entry, for the 1946 cohort than for the two later ones. This effect is in fact much the same – at about 50 points on the status scale – for all three cohorts. And further it is no longer the case that only the 1958 cohort deviates from a pattern of salariat background effects declining as level of qualification increases. These effects are still, in all cohorts, least for men with degrees but otherwise show no regular pattern.

In sum, on the basis of our status scale, analyses of occupational level at maturity provide, if anything, less evidence of secular trends across our three cohorts than do our corresponding analyses of occupational level at entry to the labour market. We move on now to analyses based on our earnings scale as shown in Panel B of Table 2.

As regards cohort effects, the most notable result is that among men whom we would predict as having reached occupational maturity by age 34, it is those in the 1958 cohort who tend overall to have the lowest levels of occupational attainment – just as was the case at entry and also, as now appears, at maturity on the basis of our status scale. In other words, some lasting impact of the relatively disadvantaged start in the labour market that these men experienced is further indicated.

The main effects of work-life variables, as shown under Model 1, are similar to those we observe using the status scale: i.e., occupational level at maturity is found to increase significantly with level of first occupation but also with the number of occupations subsequently held. And, as can be seen under Model 2, there is, again as with the status scale, no indication of these effects weakening across cohorts. In fact, no significant changes of any kind are revealed.

Turning to qualifications, we see that the main effects are strong and on a similar pattern to those reported with the status scale, including the finding that graduates in the 1946 cohort no longer have higher returns than graduates in the two other cohorts, as was the case at entry.¹³ But what once more stands out is the pattern of results for men in the 1958 cohort. That is, the tendency for the returns to qualifications for these men in terms of occupational earnings to remain distinctively low, just as they were at entry and, as we have now also found, in the case of occupational status at maturity.

The main effects of class origins are similar to those revealed using the status scale, though appearing somewhat less strong; and, also as with status, Model 2 reveals that the effect of being of salariat background remains essentially the same across cohorts. The particular advantage that men with this background in the 1958 cohort gained at entry is still suggested but is no longer statistically significant.

Figure 5, showing predictions of occupational earnings scores at maturity, derives from the same regression model as Figure 4. The strong differentiating effect of qualifications is again apparent, although with the 1946 and 1958 cohorts this is clearly below the 50-point range found with the status scale, which is matched only in the case of the 1970 cohort. There is here a clear contrast with the situation at entry when it was for men in the 1946 cohort that the overall effect of qualifications was largest. Now the overall effect is largest for men in the 1970 cohort, and chiefly on account of the greater differences that result from having, on the one hand, tertiary rather than secondary qualifications and, on the other hand, some qualifications rather than virtually none. However, as regards the effects of salariat origin, we find a continuity with the situation at labour market entry. For men in the 1958 cohort, coming from a salariat background still retains at occupational maturity at least some of the advantage it had at entry so far as access to relatively high-paying occupations is concerned, while being of slight importance for men in the two other cohorts.

Once more, then, there are no consistent indications of the operation of social forces making for secular change in the relationships between class origins, education, and occupational attainment.¹⁴ The evidence would rather suggest that in Britain, over the historical period within which the early work histories of men in our three cohorts fall, it is effects specific to cohorts, deriving from the economic vicissitudes of the period, that have more importantly differentiated their labour market experience and mobility chances – a finding, it may be noted, highly consistent with that reported some decades ago by Thernstrom (1973) for men in early twentieth-century Boston.

We noted at the outset two, often largely implicit, assumptions in current research that tend in effect to create an undue divergence between studies of intergenerational class mobility, on the one hand, and of intragenerational occupational mobility, on the other: (1) that a secular tendency exists for education, as a source of human capital, to become increasingly dominant in determining individuals’ chances of intergenerational class mobility or immobility – and regardless of the part that may appear to be played by their work-life occupational trajectories and (2) that a secular tendency exists for the influence of individuals’ class origins on their chances of work-life occupational mobility or immobility to be increasingly channelled via their educational attainment, so that other effects of class origins become of little importance. Given the accumulating evidence that these assumptions may be mistaken, our aim has been to subject them to further examination in the light of the experience of men in three British birth cohorts extending over the second half of the twentieth century. The main conclusions that we would draw from the analyses we have reported are the following.

First, educational qualifications are obviously a very powerful determinant of men's occupational level at entry into the labour market and again at occupational maturity. This is the case whether we scale occupations according to their status or their average earnings, although qualifications appear to have generally stronger effects in regard to status than to earnings. However, we find no consistent evidence that the importance of qualifications is becoming greater across the three cohorts at either entry or maturity.

Second, the independent effects of class origins on men's occupational level, while less strong than those of qualifications, remain generally significant at both entry and maturity. There are, moreover, no clear indications that these effects are tending to weaken. What is of chief interest here is that class origin effects prove to be generally weaker when occupational level is measured in terms of earnings rather than of status. This supports the argument made elsewhere (Bukodi et al. 2011) that the occupational earnings hierarchy is more fluid or ‘open’ than the status hierarchy, and in turn underlines the need to treat earnings and status mobility separately rather than in terms of some ‘synthetic’ socioeconomic status scale.

Third, while men's qualifications and class origins play a major part in determining the occupational level at which they enter the labour market and this level has itself a positive effect on the level they reach at maturity, a further feature of their work-life experience still retains an independent importance: i.e., the more occupational changes that men make between entry and maturity, the higher their occupational level at this latter stage. And we find no evidence that the importance of work-life mobility in this regard is declining.

Fourth, although our analyses do not reveal changes across our three cohorts that would be consistent with the idea of education playing a steadily increasing role in the occupational attainment process or in mediating the influence of class origins in this process, they do clearly point to the possibility of cohort-specific effects. Repeatedly, we find that the experience of men in the 1958 cohort differs from that of men in both the earlier and later cohorts. The early work-life histories of men in the 1958 cohort coincide with a period of severe economic difficulties, labour market re-structuring and continuing high levels of unemployment. At entry to the labour market, adverse effects are then indicated, chiefly in regard to earnings. For men in this cohort earnings returns to all levels of qualification, and especially to higher levels, fall below those for men in either the 1946 or 1970 cohorts. At maturity, this disadvantage in regard to earnings largely persists and, moreover, at this stage secondary qualifications also bring lower status returns than for men in the other two cohorts. In other words, the effects of unfavourable labour market conditions in early working life would appear to be long-lasting.¹⁵ Finally, class origin effects also differ in the case of the 1958 cohort. At entry, men in this cohort appear to benefit distinctively from having a salariat background in that this helps to offset somewhat the relatively low earnings returns that they obtain from higher-level qualifications; and at maturity such an advantage is still apparent while being negligible for men in the two other cohorts.

In sum, the findings we have reported must further call into question assumptions regarding secular change that seem often present, if only implicitly, in studies of intergenerational class, or intragenerational occupational mobility. The experience of men in three British birth cohorts, as they move from their class origins, through their educational careers, into the labour market and then to a stage of occupational maturity, would not appear to be greatly illuminated by being placed within a narrative of ‘modernisation’. In many respects it is in fact the absence of change in the trajectories they follow, and in the apparent determinants of these trajectories, that is most notable. And where cross-cohort differences do show up, these would seem often better understood in the context not of some relatively benign transition from industrialism to post-industrialism but rather in that of what we can today readily recognise as the disruptive economic cycles endemic to capitalism.

Our research forms part of the ESRC Gender Network, Project 1 ‘Changing Occupational Careers of Men and Women’, reference: RES-225-25-2001. We are indebted to Diana Kuh for access to the MRC NSHD data-set. For helpful comments on earlier versions, we thank Carlo Barone, Shirley Dex, Robert Erikson, Steffen Hilmert, Heather Joshi, Yaojun Li, Colin Mills, Walter Müller, Antonio Schizzerotto, participants in a seminar at Nuffield College Oxford in January 2010, and an anonymous referee.

Erzsébet Bukodi is a Reader in Quantitative Sociology and Research Director of the National Child Development Study and the 1970 British Birth Cohort Study, Centre for Longitudinal Studies, Institute of Education, London. Her research interest includes intergenerational and intragenerational social mobility, transition from school to work and life-course research.

John Goldthorpe was an Official Fellow of Nuffield College, Oxford from 1969 to 2002 and is now an Emeritus Fellow. He is also a Fellow of the British Academy, a Member of the Academia Europaea and a Foreign Member of the Royal Swedish Academy of Sciences. His publications include The Affluent Worker (with David Lockwood and others, 3 vols, 1968–9); Social Mobility and Class Structure in Modern Britain (1980; 2nd edition, 1987); The Constant Flux: Class Mobility in Industrial Societies (with Robert Erikson, 1992); On Sociology (2000; 2nd, two-volume edition, 2007); and From Indifference to Enthusiasm: Patterns of Arts Attendance in England (with Catherine Bunting and others, 2008).