OCCUPATIONAL ATTAINMENT AND CAREER PROGRESSION IN SWEDEN

Research on intergenerational mobility is one of the best developed fields in sociology (e.g., Erikson and Goldthorpe 1992; Breen 2004; Breen and Jonsson 2005). This research has been done somewhat independently from studies on intragenerational mobility, or career mobility (however, see e.g., Mayer and Carroll 1987). Career progression is of interest both for understanding the functioning of labour markets, and also from a more general stratification viewpoint: the more occupational mobility there is, the better the chances for advancing from initially disadvantageous occupational positions (e.g., Esping-Andersen 1993). Career mobility also matters for our understanding of intergenerational mobility. Firstly, considerable career mobility hinders the possibility of drawing clear-cut conclusions on intergenerational mobility based on single observations of the parent and the child. Secondly, life course studies on class origin effects can give clues about the mechanisms of intergenerational class reproduction: which life stages are most important? Finally, we can learn about the stability of class origin effects over individual life courses (cf. Mayer and Carroll 1987).

Our objective was to describe occupational attainment and career progression patterns over time in Sweden. Sweden is generally placed among the more open societies in terms of intergenerational mobility (Breen and Jonsson 2005; Björklund and Jäntti 2009), but appears to have an average position in terms of intragenerational mobility (e.g., DiPrete 2002). However, we find gaps in knowledge of career progression patterns and in what these patterns tell about the Swedish stratification regime. Thus, we focused on three more specific questions of occupational mobility and career progression in Sweden.

Firstly, we wanted to present a general description of occupational attainment over the life course and across cohorts. How do careers develop from the first job to later career stages? Is there a stage of ‘occupational status maturity’ – a general stabilization of the occupational career? Secondly, we were interested in the role of educational attainment and class background on occupational attainment over one's career (cf. Sorensen 1975; Mayer 2009; 2010). Thirdly, we wanted to know more about cohort changes in career progression and in the effects of education and class background. Throughout the empirical analyses we conducted separate analyses for women and men, since women's careers are more frequently interrupted than men's by periods of parental leave, and earlier, by periods of housewifery. While we will not focus on gender differences per se, we can nevertheless indicate some gender differences in occupational attainment processes.

The theoretical foundations of career mobility research are well described in other contributions to this special issue. Therefore, we focus more on previous research, particularly on Sweden. Previous research has commonly used either path-analysis in which family background, educational attainment, and early (typically the first) occupations are used to predict later occupational attainment (Ganzeboom et al. 1991) or event-history approaches to analyze transitions between jobs or occupations (e.g., Blossfeld 1986; Carroll and Mayer 1986; Grunow and Mayer 2010). In this study, we approach our questions using growth curve analysis of occupational prestige scores over the life course (see also Barone et al. in this issue). This enables a broad overall representation of career progression over the career and an examination of the effects of independent variables at its different stages (cf. Mayer 2009).

Well-known features of the Swedish labour market include its early history of female labour market participation, a large public sector, and a compressed wage structure. Sweden also has an open, export-oriented economy, and its labour market is less regulated than often anticipated (see e.g., Holmlund and Storrie 2002; Korpi and Tåhlin 2006 for descriptions of the Swedish labour market). The steady growth and low levels of unemployment of the 1950s and 1960s were temporarily disrupted by the oil crisis. The major crisis did not arrive until the 1990s (Figure 1).

From the mid-1980s, the earlier steady growth of the public sector also came to an end, and even declined somewhat (Edin and Topel 1997). During the same period, wage differences have increased (le Grand et al. 2001). All of these developments can affect occupational opportunities and career progression (cf. Benner and Bundgaard Vad 2000).

Educational expansion in Sweden has meant that successive cohorts have – although not in a completely linear fashion – been increasingly likely to hold upper secondary as well as university degrees.¹ At the same time there has been occupational upgrading (Oesch 2006). However, it has been indicated that the trend of educational expansion exceeded the trend of occupational upgrading resulting in an oversupply of well educated people (Korpi and Tåhlin 2009).

The changes in the educational system are however not only quantitative, but Sweden has experienced two major educational reforms in the last decades. In the reforms of 1960s, tracking to vocational and academic streams was postponed to later ages, and barriers to higher education were lowered. These reforms reduced family of origin effects on educational and income attainment (DiPrete 2002; Mayer 2005; Meghir and Palme 2005, but see also Blossfeld and Shavit 1993). In the 1990s, the former division into 2-year vocational secondary and 3-year academic tracks was replaced by 3-year education in both tracks (for descriptions of the Swedish school system see Erikson and Jonsson 1996; Halldén 2008). However, this reform did not seem to remove barriers to higher education, as was intended (Hall 2009).

In terms of career mobility, contemporary Sweden occupies a middle position between less fluid countries such as Germany, where the connection between educational tracks and occupations are stronger, and more ‘liberal’ ones such as the United States (DiPrete 2002; Korpi and Mertens 2003; Mayer 2005). Such findings fit well-known institutional variation in the tightness of employment protection and the connection between educational tracks and occupations.

To present a more detailed picture of what is known about career mobility in Sweden, we go through previous research on these issues. We first discuss earlier findings on career progression and the question of occupational maturity in particular. We then look at research on the impact of educational attainment and class origin over the life cycle, and finally, discuss the possible changes over cohorts in career progression. We present our expectations at the end of each subsection.

Most studies on intergenerational mobility assume that – after some possible mobility – class positions stabilize in the 30s. Ceiling effects provide maybe the most obvious reason for such stabilization: since status scales of occupations have a limited number of positions, the chance of an increase probably decreases with every step up the career ladder (e.g., Sorensen 1975).

There are also more substantial reasons why people would be more mobile at the beginning of their careers. Firstly, initial matching problems in the start of careers tend to be resolved with years spent in the labour market. Secondly, work experience and on-the-job training function as additional forms of human capital (Mincer 1974), the marginal value of which is, however, likely to be concave. Thirdly, internal labour markets with their promotional ladders provide avenues for career progression, often particularly during early career stages (e.g., Sorensen 2000). Lastly, early careers overlap with the ‘rush years’ of the life course, in which increased financial needs due to family formation provide additional incentives for occupational progression, especially for men (Mayer and Carroll 1987).

Studies examining both occupational and income attainment over the life course generally tend to support the hypothesis of occupational maturity around the age of 30 or somewhat higher (Jonsson 2001; Böhlmark and Lindquist 2006; Breen and Jonsson 2007; Bihagen et al. 2010). Furthermore, the study by Böhlmark and Lindquist (2006) suggested that earnings of Swedish men correspond to their life time earnings around the 30s, although the pattern was somewhat less stable for women. Given these previous findings, we also expect occupational maturity to occur around age 30, although we will focus more closely when this happens and whether we see any patterns by cohort, gender, education, and class background.

A persistent finding in the literature on intergenerational mobility is that most effects of class origin on class destinations operate through educational attainment (Breen 2004; Breen and Jonsson 2005; Hout and DiPrete 2006). Fewer studies have analyzed how family background and education operate across careers and the life course. Some expectations can be laid out. Much research has followed the classic occupational attainment model (Blau and Duncan 1967), which emphasizes educational attainment and one's first job. According to this model, class origin mainly affects one's educational attainment, which then has a strong influence on the status of one's first job, which lays the foundations for subsequent occupational attainment (cf. Wolbers et al. in this issue). One's first job has been consistently found to affect later career opportunities (e.g., Mayer and Carroll 1987; Erola 2009; Bukodi and Dex 2010; see also other articles in this special issue). Also consistent with this thesis, the net effects of class and education are smaller for occupational status in later careers.

Class origins and educational attainment may nevertheless shape occupational trajectories throughout one's career. Early effects of class origin and educational attainment become weaker as workers gain valuable experience and further training. For example, Warren et al. (2002) found diminishing effects of education, especially for men, but more stable and persistent effects of ability over the life course. They also reported that social origin effects operate entirely through these two factors, therefore implying that the only effect that changes is that of education. Opposite predictions are also possible. If early careers are characterized by job mismatches, one could expect educational and class origin effects to ‘kick in’ mainly at more mature career stages. Furthermore, the notion of ‘counter-mobility’ suggests that people tend to diverge back to their origin classes (especially if they are relatively high); this suggests that the effect of class becomes relatively stronger at later career stages, while that of education diminishes (cf. Mayer and Carroll 1987).

We did not find any equally comprehensive analyses of the life-course patterns in occupational attainment for Sweden. However, using the same data as we do, but restricted to the period up to 1991, Jonsson and Erikson (1997) found some persistent effects of class origin on career mobility, which were consistent with the hypothesis that farmers and those from the upper professional classes in particular tend to return to their class of origin. Unlike Warren et al. (2002), Jonsson and Erikson did not control for educational attainment. In another study, Bihagen (2007) reported, using panel data, some direct effects (net of education) of class origin for women's downward mobility from the Service I class. Hence, there are some signs of remaining net effects for class origin later in careers in Sweden. We find it plausible that class background effects strengthen at later career stages, but we expect educational differentials to become weaker.

It is often suggested that globalization has a substantial influence on working life. Since the first job appears to be decisive for the successive career, such a societal shift would especially affect cohorts who entered the labour market quite recently. Results from the Globalife project do not, however, provide any clear evidence for increasing occupational mobility and career instability over time (Blossfeld et al. 2006, 2008; Blossfeld and Hofmeister 2006). Although there are some indications of increasing occupational mobility for Swedish men in the 1980s (Korpi and Tåhlin 2006), and for women for a longer time period (Korpi and Stern 2006; see also Jonsson 2001), such trends levelled out in later periods. Job stability is also indicated to have increased from 1970 to 2000 (Korpi and Stern 2006; Korpi and Tåhlin 2006). Finally, Halldén and Hällsten (2008) reported increasing difficulties at the beginning of the career with delayed school-to-work transitions and an increased risk of having a temporary job, but a declining risk of downward mobility over time.

Hence, these cast some doubt on the idea of one large scale change and it may be more plausible that cohorts are affected differently by recurrent business cycles. Given the possible long-term ‘scarring’ effects of unemployment (e.g., Gangl 2006), entry to the labour market during high unemployment can have long-term implications for the cohorts involved. There is also little research on the role of educational attainment and class origin for the career progression of different cohorts. We could expect a devaluation of educational attainment levels with the increasing proportion of well-educated workers over time, which may lower both gross and net effects of education across the life course. In line with inter-generational studies, at least concerning Sweden, class origin should also matter less for later cohorts (Erikson and Goldthorpe 1992; Jonsson 2004). Finally, we would expect different cohort patterns for men and for women.

We used retrospective occupational biography data for a random sample of Swedish inhabitants from the 1991 and 2000 waves of the Swedish Level of Living Survey (LNU) (see Jonsson and Mills 2001; Manzoni et al. 2010). Respondents reconstructed their occupational careers (occupation in and starting and ending times of each job) starting from the first job they held that lasted for 6 months or more. Although we were able to construct very detailed occupational careers, these data are naturally not without caveats, an obvious one being recollection bias. However, a recent study by Manzoni et al. (2010) reported that memory biases in the LNU – which especially affected reconstruction of complex careers and recall of events far back in time – do not generally affect substantive conclusions.

We had separate samples of 2,511 men and 2,426 women born between 1925 and 1974 and who had ever held a job for 6 months or more. Although homemaking among women used to be relatively common in the older cohorts, there were hardly any respondents who had never held a job for 6 months or more. Together with monthly information on occupational status, we had a total of 539,339 men-months and 436,824 women-months.

As a measure of occupational attainment we used the Standard International Occupational Prestige Scale (SIOPS), the Treiman scale of occupational prestige (Treiman 1977; Ganzeboom and Treiman 1996). This scale was developed to provide a cross-nationally comparative measure of the relative prestige or social standing – as assessed by the public – of a variety of occupations. Such assessments are close to being constant across time and between countries (Hout and DiPrete 2006), which makes it plausible to measure them with one scale (see also Magnusson 2009). This makes it suitable for our purposes as we look at change over time periods.²

For men in our sample, the grand mean of this variable (Table 1) is 40.3 (e.g., police officer, farmer, or locksmith), the standard deviation 11.6, the minimum 6 (e.g., hunters and trappers), and the maximum 78 (e.g., medical doctor or professor). For women, the mean is 38.7 (e.g., customer service clerk or prison guard), standard deviation 12.2, minimum 13 (e.g., garbage collector or street sweeper), and maximum 78.

Our independent variables are parents’ social class, educational attainment, work experience, cohort, and the number of children. We compare five 10-year cohorts, born in 1925–1934, 1935–1944, 1945–1954, 1955–1964, and 1965–1974. Although somewhat arbitrary, these cohorts cover the timing of major aspects of social change in Sweden. These are entered as dummy variables, with the oldest cohort as the reference category. The time-varying variable measuring educational attainment differentiates between five categories, namely compulsory education (folk-/grundskola), which acts as the reference category, vocational secondary education, academic secondary education (Gymnasium), lower tertiary (usually 2 years), and university education (3 years and more). Parents’ social class is measured using a five-class EGP scale, which differentiates between the higher (EGP I) and lower (EGP II) service classes, higher routine non-manual workers, supervisors, and higher-grade technicians (EGP IIIa and V), the self-employed and farmers (EGP IV), and the working class as the reference category (EGP IIIb, VI, and VII). The ‘dominance principle’ in which the highest class position of the parents determined one's class background was used (cf. Erikson 1984). As expected, there was a clear upgrading of both educational attainment and class background over the cohorts. In the case of education, this was clearest at the bottom, with the share of those with compulsory education falling, especially at the expense of vocational secondary and the Gymnasium. At the same time, the share of those hailing from the professional classes increased, particularly at the expense of those with farmer/entrepreneur backgrounds, while the share of working class backgrounds changed much less.

We measure time in the labour market as years since one started her or his occupational career, that is, held her or his first ‘real’ job. We specify this (0–25 years) using five 5-year linear splines. The parameter estimates tell how an additional year since one first entered the labour market affects occupational prestige within each spline, so for example, during the first 5 years, between the fifth and the tenth year, and so forth. This provides a flexible way of analyzing career progression without imposing stronger functional form, such as linearity or curvilinearity, on the relationship between work experience and occupational prestige.

We use growth curve modelling (e.g., Halaby 2003; Steele 2008), which is rarely used in sociological analysis of occupational attainment, but provides a fruitful approach for analysis of occupational attainment and career progression (cf. Barone et al. in this issue). Mayer (2009) characterized growth curve analysis as lying between ‘analytic’ approaches (which usually use event-history analysis) and ‘holistic’ ones (which use, for example, sequence analysis). Although not focusing on the specific job shifts that make up an occupational career, growth curve modelling shares with event-history analysis the incorporation of independent variables. And, by providing depictions of career trajectories, our approach has resemblance to analyses of careers through sequence analysis (e.g., Bison in this issue).

In addition to this baseline model, we estimate two models with interaction terms. The first adds interactions between cohort, class background, and education on the one hand, and the time-in-labour-market splines on the other. This model is used to examine how occupational attainment varies across the career by birth cohort, class background, and educational attainment, respectively. The second model includes interaction terms between birth cohort and class background, and birth cohort and educational attainment, to study cohort change in the effects of these two predictors of occupational success. We do this using both the categorical birth cohort measure, and to improve efficiency due to the limited sample size, a linear year of birth measure.

We estimate cluster-robust standard errors (cf. Wooldridge 2002: 262–3; Cameron and Trivedi 2009).

We begin by describing occupational attainment in different cohorts over two time indicators, the life course and historical time. These patterns (Figure 2) refer to respondents aged 15 to 50, and who were working during the month in question. The upper two panels show how average occupational prestige scores develop over the life course and historical time, respectively, among men, and the two panels below present the same information for women.

There are both commonalities and differences between the sexes and across cohorts. Firstly, in line with the occupational maturity hypothesis, career progression seems to slow down and flatten around age 30. However, there are some main differences across cohorts so that in younger cohorts, career progression continues until age 40 and beyond. Secondly, women have experienced steady ‘occupational upgrading’ where each cohort attains, on average, higher prestige occupations than the previous one. The youngest cohort, born in 1965–1974, even ends up having higher occupational attainment at the end of the follow-up (at around age 30–35) than their male peers. Although these results fit the story of women's increased labour market involvement, the scope of the change may be surprising. We do not find a similar monotonous pattern of occupational upgrading for men. For example, the cohort of men born between 1955 and 1964 appear to do somewhat worse than both the cohort born before and after them. We return to these results in the next sections.

We continue by analyzing career progression in Sweden with growth curve models, separately for men women (Table 2). The first model is the ‘empty’ model without covariates, which simply decomposes the total variance to variance between individuals (around the constant) and variance over individual careers (residual, or within-individual variance). A substantially higher share – approximately three-quarters – of overall variance in occupational prestige results from variance across individuals rather than across individual careers. In other words, intergenerational mobility does not overcome (more permanent) occupational attainment differences between individuals in Sweden.

In the next model we introduce our two time variables, cohort (as dummies) and time in the labour force (as splines). The estimates for the latter show a similar pattern to what was already found in Figure 2: an initially faster rate of career progression in the beginning of the career, which then slows down. Men and women gained on average 0.5 to 0.6 points in occupational prestige per year during their first 5 years of work. Men's occupational prestige continued to improve at an average annual rate of 0.4 points during the next 5 years, and generally on average between 0.1 and 0.2 points per year for 15 years thereafter. An average man in our sample would, therefore, have started from a job with a SIOPS score of 35 (e.g., blacksmith or telephone/telegraph installer) and gained 6.6 points, or 0.57 standard deviations, in occupational prestige during 25 years of work (ending up as, for example, a machinery or electronics mechanic).

Women's careers started, on average, from a lower level (32 points; e.g., shop assistant or hairdresser) and after the first 10 years, women's careers were flat. After 25 years, our average female worker would have gained 3.9 points, or 0.32 standard deviations, in occupational prestige, finding herself working for example as a clerk or a receptionist. These results again point to the relative stability of work careers and the relatively limited chances of overcoming initial inequalities in occupational attainment through career mobility, especially for women.

The cohort estimates show interesting patterns. Particularly, among men, the 1955–1964 cohort deviates clearly from a general trend of occupational upgrading and did not reach higher occupational prestige than the oldest cohort, born in 1925–1934. For women, there was a steadier increase in occupational positions (also seen in Figure 2). For both men and women, cohort and work experience explain only a small share of the variance between or across individual careers.

In the next model we introduce three new variables: class background, educational attainment, and the number of children. The returns to education are, overall, similar for both sexes. Women benefited somewhat more than men from lower tertiary education. Overall, tertiary education – and university education in particular – clearly brought benefits for both men and women. Class background also mattered for men's and women's occupational attainment, net of own educational attainment. There were no differences between those having working class and entrepreneurial or farming backgrounds, but those from higher social classes, and especially the service class (Classes I and II), attained higher occupational prestige. However, it is worth pointing out that occupational prestige differences by educational attainment are clearly larger than according to class background. Further analyses (not shown) revealed that controlling for education reduces the estimates of class background by 40 to 50 percent, while the estimates of educational attainment hardly changed after adjusting for class background. In other words, educational attainment has a stronger independent effect, and class background differences in occupational attainment operate largely through education. Children are not associated with men's occupational attainment, whereas each additional child is associated with a reduction of approximately one SIOPS point in women's occupational prestige.

These three variables together reduce the between-individual variation in occupational attainment by approximately 40 percent, both for women and men. The reduction in within-career variance is negligible for men, and below 10 percent for women. In other words, they account for a major share of differences between individuals (and their entry jobs), but little across individual careers. The estimates for birth cohort also change. For men, the occupational advantage of the younger cohorts (with the exception of that born in 1955–1964) disappeared. In other words, occupational upgrading in these cohorts was due to improved resources, particularly higher levels of education. The cohort of men born in 1955–1964 actually had more difficulties in attaining occupational prestige – even with similar qualifications – than older cohorts. Improved resources also seem to account for women's occupational upgrading. However, women in the two youngest cohorts seem to have been somewhat disadvantaged compared to those from two immediately preceding ones.

The above results suggested some cohort differences in occupational attainment, but they do not tell us where in the career these lags come about: for example, do men in the youngest cohorts begin their careers from less prestigious jobs, or do their careers progress at a slower pace?

To examine this, we here turn to interaction models, which include interaction terms for cohort and time in the labour market, class background and time in the labour market, and educational attainment and time in the labour market. The interaction rows between cohorts and time in the labour market are not jointly significant for men (χ²=23.10; df=19; P=0.23), but are jointly significant for women (χ²=35.27; df=19; p=0.01). There is thus clearer evidence of cohort change in career progression for wo men. However, in both cases and especially in the case of men, some of the estimates of specific interaction effects are significant. Here and in subsequent analyses, we do not present these interaction estimates – which are available from the authors on request – but comment on them in the text. Instead, we focus on the bigger picture of career progress across cohorts (and later by educational attainment and class of origin), which we examine by plotting predicted values of SIOPS scores for each cohort by work experience (keeping other variables at their reference values). The results for cohort change are shown in Figure 3.

Starting with men, we can find that the career paths of the three oldest cohorts were similar, with fast progress during the first 5 years, and somewhat slower progress during the 10 following ones (the third cohort deviates somewhat). The last two cohorts started at somewhat lower prestige jobs (although the differences are not statistically significant), and their careers progressed at slower paces than in the oldest cohort. More specifically, men in the 1955–1964 had slower progress during their first 5 years, with no catching up at later career stages. These developments meant that these men worked in jobs that were approximately 2 to 3 points lower in prestige after 15 years of working life, corresponding to our estimates in Table 2. These men started working in the late 1970s and early 1980s. Despite the internationally low unemployment rates around this period, the new challenges experienced by Sweden's traditionally strong industrial sector could have affected the labour market advancement prospects of the cohort entering the labour market during this period (Benner and Bundgaard Vad 2000).

Men in the 1965–1974 cohort, on the other hand, experienced a dip, with a later catch-up, between 10 and 15 years into their working lives. Closer analysis, with results not presented here, revealed that the specific cohort that was particularly hit consisted of those born between 1970 and 1974. They entered the labour market to lower status occupations during the 1990s recession, and never fully recovered from this experience.

Although the interactions between cohort and time in the labour market are jointly significant, cohort differences in women's career progression are generally smaller. The joint significance of the interaction terms comes, in contrast to men, from cohort differences in later careers: the oldest cohort of women experienced losses in occupational status after 20 years of work. This was, however, restricted to this oldest cohort, and later cohorts maintained their positions also at later stages of their career.

Controlling for educational attainment, the improved occupational attainment of women in the 1935–1944 and 1945–1954 cohorts is linked to the higher prestige of their first jobs (difference significant at the 5 percent level). The two youngest cohorts, on the other hand, started working in lower prestige jobs than these two middle cohorts, and held first jobs of same status as the oldest cohort. Closer analysis (not shown) identifies the 1970–1974 cohort to be particularly hit in terms of their first jobs, even though women in the youngest cohort generally appears to continued to make progress also after the first 5–10 years. Although the 1955–1964 cohort attained lower occupational status than the cohort preceding it, it appears to have been somewhat less affected than the same cohort of men. A possible explanation is that the public sector continued to expand despite problems in the industrial base of the Swedish economy (Benner and Bundgaard Vad 2000).

All in all, these results – and particularly the rougher early careers of men and women born between 1955 and 1974 (and between 1970 and 1974 in particular) – point to potentially lasting effects of early disadvantage in the labour market. These effects can operate both through effects on entry jobs or slower career progression at early stages.

The above results showed that both educational attainment and class background matter for occupational attainment. However, they again do not tell us about life course differences in their effects: what are the most important stages that produce these differences?

The interactions between education and the time-in-labour-market splines are jointly significant at the 5 percent level for men (χ²=43.39; df=20; P=0.00) and at the 10 percent level for women (χ²=29.25; df=20; P=0.07). At a closer inspection, the only specific interactions that are jointly significant for men are between university education and time in the labour market (joint significance test: χ²=23.90; df=5; P=0.00) and for women, the only jointly significant interactions are between vocational secondary schooling and the splines (χ²=10.15; df=5; P=0.07).

Figure 4 plots the predicted values of the interactions between educational attainment and time in the labour market, holding the other variables at their reference values. Educational attainment clearly mattered throughout the career, but differently at different stages, due to variation in career progression patterns by educational level. Consider the occupational attainment of men. Men with only compulsory schooling or vocational secondary schooling started working in relatively low-prestige occupations, and could expect to gain approximately 6 points in 25 years. The next two educational groups, men with academic secondary education and lower tertiary degrees started from somewhat higher prestige jobs, but made faster progress, ending up in jobs with 9 more SIOPS pioints after 25 years. Finally, university graduates started working in clearly higher prestige jobs (e.g., associate professionals), but made slower career progress, probably due to ceiling effects. Given academic men's later entry to the labour market, the gap between them and other men was the highest at the beginning of their careers, on average around the mid-to-late 20s. Those with academic secondary education or lower tertiary schooling then partly catch up at older ages, and increase the gap to those with less education.

Women's career progression by educational attainment shows similar patterns to men's. A clear difference is the higher starting point, but lower rate or progression, of women with lower tertiary degrees. Women with academic secondary degrees begin their careers from lower prestige jobs, but close the gap somewhat during their first 20 years. Women with only compulsory levels of education start off even lower, but almost close the gap to those with vocational secondary schooling during 20 years of work. Summing up, educational differences in occupational attainment in Sweden vary over the career. However, the question of how the effects of education vary is not easily answerable, but depends on which educational attainment levels are being compared.

The results in Table 2 showed that class background differentiates occupational attainment much less than educational attainment. But do the effects vary at different points of the career? Tests of joint significance between class background and the work experience splines do not suggest this either for men (χ²=22.17; df=20; P=0.33) or for women (χ²=23.87; df=20; P=0.25). Closer inspection again reveals some class background differences in career progression. For men, the main difference is between those from service class backgrounds and others (χ²=14.11; df=5; P=0.03). Figure 5 – which shows how class background affects men's career progression – points to generally similar trends across men with different backgrounds. However, the careers of men from the upper service class seem to have continued progressing longer than for other men, thus increasing the gap in occupational prestige, although the differences are generally rather small. This gives some, though marginal, support to the thesis of growing class differences and of men from higher social positions reaching their parents class at later career stages (Mayer and Carroll 1987; Jonsson and Erikson 1997). For women, some of the separate interactions are significant at the 10 percent level. As seen from Figure 5, the overall career patterns are very similar across class backgrounds.

Overall, class background differences are not only smaller than those based on educational attainment, but they also remain more stable across the occupational career. In other words, the direct effect of class on occupational attainment functions mainly through entry level jobs, whereas educational differences are produced both through differential access to entry level jobs, and through different subsequent career progression patterns.

Lastly, we examine cohort differences in the role played by educational attainment and class background in occupational attainment. We analysed these by testing for interaction effects both between educational attainment and class background, and the 10-year cohorts as used throughout the text, and between education and class background, and a linear birth year variable. The latter was done for efficiency reasons, given our limited sample size. However, both point to similar directions.

Beginning with the question of whether the effects of educational attainment have changed over cohorts, we find that the interactions between educational levels and cohorts are jointly significant for men (χ²=28.09; df=16; P=0.02), but not for women (χ²=17.25; df=16; P=0.37).³ Specifically, there was a significant change in the value of secondary vocational schooling, as compared to compulsory schooling. As can be seen from Figure 6, this was true both for men and for women, although the value of secondary vocational education began declining earlier for women than for men. Additionally, there was a significant interaction effect (at the 5 percent level) between the linear birth year variable and Gymnasium for women, as can also be detected from Figure 6. The value of this degree for women declined by 7 SIOPS points: while a Gymnasium-educated woman in the older cohort could expect to have a lower-level white collar job as a coding proofreading clerk, one in the youngest cohort was more likely to have a working class job as a baker or other craft worker.

Whether the effects of class background have changed depends on the time variable used: the interactions are not jointly significant when using the categorical cohort variable (Men: χ²=16.56, df=16, P=0.41; Women: χ²=13.46, df=16, P=0.64), but are significant when using the linear year of birth variable (Men: χ²=8.70, df=4, P=0.07; Women: χ²=13.31, df=4, P=0.01). Again, when examining the interactions more closely one can find changes in class background effects (in line with previous research: Jonsson 2004; Breen and Jonsson 2005), which depend, however, on the background class in question. In the oldest cohort of men, class backgrounds grouped roughly into three groups, with those from the upper service class at the top, those from lower service classes and higher non-manual classes coming next, and men with working class or entrepreneurial backgrounds having the lowest occupational status. Over time, the advantage of those from the upper service class diminished (both the linear, P<0.01, and non-linear interactions being significant) so that from the post-war cohort onwards the main cleavage was between those from the service and the higher routine non-manual class and those from the working classes and farming/entrepreneurial backgrounds.

In the oldest cohort, women from both service classes attained higher occupational status jobs than those whose parents were in the routine non-manual class, who then attained higher status occupations than those from working class or entrepreneur/farmer families. There were also some, although less clear, changes in class background effects also for women. Interactions with the linear year of birth variable pointed to a significant reduction in the relative advantage of women from the lower professional class, and the interactions and Figure 7 suggest that this happened from the 1955 to 1964 cohort onwards. In the youngest cohorts, women from the upper service class have a slight advantage in terms of occupational prestige, but class background inequality was otherwise rather small.

How do our empirical analyses answer the questions laid out at the beginning of this article, and what do they tell us about social stratification in Sweden? Regarding our first question of general occupational attainment across the life course, the common assumption of ‘occupational maturity’ around the 30s is supported, particularly among men. Thus, occupational standing by the age above 30 is a reasonable approximation of a person's class of destination and inter-generational studies seem to be on safe ground concerning this assumption.

Secondly, we were interested in how educational attainment and class origins matter differently across the career. As a general conclusion, we find similarly to many other countries that differences in occupational attainment by education and class background appear already at the beginning of careers. Although they do not stay completely stable over time – for example, those with university degrees (especially men) generally start their careers and remain in high prestige jobs while others (especially those with lower tertiary or academic secondary education) continue to gain in prestige (and catch up) longer across their careers – there are somewhat limited chances to correct for inequalities at early career stages. Nevertheless, effects of education do vary across the career. These patterns also depend on the educational groups being compared, making it more difficult to reach generalizing conclusions about educational effects across the life course (cf. Warren et al. 2002). Class background effects operated mainly through educational attainment and remained more stable across the life course, although there were some indications that men from the upper service class tend to approach their class origin at later career stages. Overall, the ‘standard’ model of occupational attainment seemed to fit, with some refinements, the Swedish case rather well.

Our third question concerned cohort change: are there general trends in occupational attainment and career progression across cohorts and have the predictors of occupational attainment remained stable? In terms of general occupational attainment, the clearest change is visible in the improvement of women's occupational attainment levels, which accounts for much of general occupational upgrading in Sweden. Women in the younger cohorts continue to make occupational gains later in their careers than women in the oldest cohort: as a result, the youngest women even enjoyed higher average occupational prestige in their early 30s than men. These changes are mainly driven by women's improved educational attainment levels. These results speak in favour of the idea that cohort replacement will solve some of the gender inequalities in occupational attainment seen in Sweden, although research with newer data is warranted.

We also found long-term effects of labour market entry during the economic recessions of the mid-1970s and early 1990s. While the latter one seemed to hit both men and women alike, there were some indications that the former affected men more severely, possibly by affecting male-dominated manufacturing industries (Benner and Bundgaard Vad 2000).

Regarding the cohort (in)stability of the predictors of occupational attainment, we found declining direct effects of class background, in line with previous research and our expectations about cohort differences (Erikson and Goldthorpe 1992; Jonsson 2004). Regarding the effects of educational attainment, the clearest trend concerns the steady decline of returns to secondary schooling. Partly due to this, and in contrast to our expectations, the gap between those with tertiary degrees and the rest has widened, although for men it seemed to have peaked for the 1945–1954 cohort.

We believe that our study has contributed to a more comprehensive picture of career progression and social stratification in Sweden, and shown the fruitfulness of our methodological approach by simultaneously describing the patterns of career progression and the continuities and changes in occupational attainment differences over the life course. However, much more work remains to be done in order to corroborate these findings and to further investigate causes behind them.

Previous versions of the paper have been presented at the Tilburg meeting of this special issue's working group and at the Swedish Institute for Social Research. We thank the participants, and in particular Colin Mills, Carlo Barone, Antonio Schizzerotto, John Goldthorpe, and an anonymous reviewer for valuable comments. The responsibility for any mistakes remains ours.

Juho Härkönen works at the Swedish Institute for Social Research (SOFI) and at Stockholm University Demography Unit (SUDA), both at Stockholm University. His main research interests relate to social stratification and family demography from a life course perspective. His recent research has been published or is forthcoming in Demography, European Journal of Population, European Sociological Review, and Population Studies.

Erik Bihagen is placed at the Swedish Institute for Social Research (SOFI) at Stockholm University, where he directs a large research project based on Swedish register data (SUNSTRAT). His research deals with social stratification and class analysis, particularly in regard to economic differences, careers, and the recruitment of social elites. Recent publications include Acta Sociologica (2005, 2007, 2009), Social Indicators Research (2007), Sociology (2008), and chapters in various books.