The present study, based on a Swedish cohort born in 1965 which was followed from age 16 to 42 years, aimed to identify typical life-course trajectories of education and LMA in 18–42 year old men and women and examine to what extent parental socioeconomic position and level of youth depressive symptoms could predict which type of trajectory the individuals would follow. The novelty of the study is mainly represented by the long follow-up period with low attrition rates, as well as the complexity of data in terms of the number of states underlying the sequences. To our knowledge, no other study has mapped the education and LMA trajectories from late adolescence into mid-life with such rich data. Exploring how depressive symptoms in youth are related to which trajectory people follow is also relatively unique within the field.
We identified four typical trajectories for men and five for women. In men, compared to those in the trajectory of long education and stable employment, those following the short education or the continuously unstable trajectories were more likely to have had parents with a low occupational class position or parents who were unemployed. Overall, the level of youth depressive symptoms showed non-significant differences regarding membership across the trajectories. Adjusting for depressive symptoms did not significantly alter the associations between parental socioeconomic position and the life-course trajectories of education and work.
The findings show support for an inter-generational social reproduction also in women, although associations were only found when parental socioeconomic status was operationalised as occupational class (as opposed to unemployment). Similar to men, depressive symptoms did not differ significantly across the trajectories, nor did the adjustment for depressive symptoms explain the associations between parental occupational class and the trajectories of education and labour market attachment. Hence, we fail to conclude that intergenerational social reproduction is linked to depressive symptoms in adolescence among women.
Trajectories
The following discussion will mainly focus on inter-generational social reproduction and mental health selection. However, we will first discuss the trajectories that were identified. The characteristics of education and LMA trajectories, in the current as well as other studies, depend highly on socio-historical circumstances [
21] as well as the period under investigation. While the current study treats the issue of time primarily in regard to age, it should be emphasised that age effects are very difficult to separate from cohort and period effects. For example, over the past decades macro-social and macro-economic conditions have undergone major changes, with implications for making life courses less conventional [
29]. The individuals examined in the present study, were teenagers and young adults in a time of a relatively stable labour market, just before the economic crisis in the early 1990s with sharp increase of youth unemployment [
60]. Compared to most other West European countries Sweden was less severely hit by mass unemployment in the 1970s and 1980s [
55].
Despite the specific characteristics of the socio-historical context in which the lives of the Northern Swedish Cohort members are imbedded, our results are similar to other studies. For example, according to findings on women from the British Cohort Study, the participants followed trajectories of: Part-time employment; Unemployment; High level education; and Detachment from the labour market [
8]. Similarly, Widmer et al. [
28] identified occupational trajectories characterised by: Full-time employment; Mixed occupations; Return to part-time employment after decline in full-time ditto, At home trajectories; and Part-time employment, in Swiss adults between age 20 and 45. A longitudinal Scottish study of young people aged 16 to 23 years (of a cohort born in 1973), found eight transitional trajectories of education and employment: Long higher education; Short higher education; Enhanced education; Direct job; Assisted employment; Unemployment; Domestic; and Other [
26]. Another study applying sequence analysis identified the following youth-to-adulthood-transitions: Employment; Higher education; Further education; Joblessness; and Employment after long further education or training [
12]. Across all these studies, including ours, stable employment dominates whereas a minority of individuals follow a trajectory which initially is unstable but then becomes more stable with regard to employment. It also seems to be common across studies to identify a group of disadvantaged individuals with regard to short education and positions outside the labour market.
Although the distinct features of the trajectories for men and women are in line with previous findings [
28], they justify a further discussion. Men in our cohort tended to have more stable and less diverse trajectories of education and LMA than women. These differences are likely to be due to differences in possibilities for labour market participation and domestic responsibilities, including caring for children. For example, despite the gender equality regime of the Scandinavian countries, the Swedish labour market is remarkably gender segregated with men dominating the private sector and manufacturing industry, and women the public sector [
58,
59]. Public sector jobs are also more likely to be part-time and with a significantly lower pay compared to the male dominated sectors. Furthermore, Sweden has a generous parental leave cash benefit system for care of children as well as work time reduction for parents [
65]. Even though the system is not targeting mothers only, women use the majority of the benefits, with implication of weakening of the labour market attachment [
65]. Women, especially those working in the public sector, are also more likely than men to be on long-term sick leave [
66]. The analyses rendered trajectories of instability (or disadvantage) in both men and women, but different patterns were shown regarding precariousness. For example, in men it seems like those hit by unemployment were concentrated in one typical trajectory whereas unemployment was more spread out across trajectories in women. The activity state ‘outside the labour market’ was also more dominant among women than men, which, as noted above, most likely due to parental leave and sick leave. These findings are similar to those of Sweeting et al. [
6] based on a Scottish cohort of participants born in 1972 where the main reason for non-employment among men was unemployment and among women caring for home or family. The proportion of full-time employment in women is lower than among men, which most likely is due to women working part time and taking more responsibility for unpaid labour in the home [
65,
67,
68]. In 1994/95, at 30 years of age, approximately 17% of women and 1.5% of men in this cohort reported that they were on parental leave (data not presented).
There is a previous study of the Northern Swedish Cohort (8) which assessed LMA from age 30 to 42 by an ordinal scale variable (permanent employment, temporary employment, unemployment, inactivity) and searched the trajectories by latent class growth analysis [
69]. In line with the pattern of the present study, the trajectory of high level LMA was predominant in both genders, but women tended to assume more commonly the trajectories of strengthening or medium level LMA. Thus, as regards to LMA only, these different methodological approaches seem to yield largely similar results.
Prediction of trajectories
We found support for the hypothesis of inter-generational social reproduction, which is also in line with previous research using sequence analysis to identify education and work trajectories [
8‐
10,
12]. In other words, if the parents of a teenager have low occupational status and/or are unemployed, the teenager is more likely to follow trajectories of short education and employment instability compared to peers whose parents are better off. It should be noted that the strength of the associations seemed to vary between men and women. However, the gender-specific results cannot be directly compared since the typical trajectories were not the same.
Another and perhaps even more interesting finding was the heterogeneity within the categories of men and women with regard to which indicator of parental socioeconomic position that was considered. Men who had experienced parental unemployment showed a two-fold risk of ending up in the all other trajectories than “Long education into stable employment”. The relative differences across the trajectories regarding low parental occupational class were relatively larger – almost four-fold – but only present for the two least advantaged trajectories. Also among women, both indicators of parental socioeconomic position seemed to be associated with an increased risk of ending up in the less advantaged education and work trajectories. However, while these results were clear and statistically significant for parental occupational class, the differences were not statistically significant for parental employment status. These disparate findings – particularly evident for women – could be a consequence of the lack of statistical power, but could also be interpreted based on the differential meaning of occupational class and employment status of the parents for their offspring. Regarding parental occupational class, it is likely to reflect the general level of resources in the household and may be seen as a relatively stable condition. Employment status, on the other hand, will for most reflect a more temporary situation which mainly affects the availability of economic resources. Since the educational system is free of charge for all Swedish citizens, the family’s economic resources may not be as important for the youth’s future education and work trajectories as compared to the social and educational capital attached to the occupational class of the parents.
Depressive symptoms in youth showed no, or very limited, links with subsequent education and work trajectories also when taking parental socioeconomic position into account. The unadjusted model for men showed that those who reported elevated levels of depressive symptoms in youth were more likely to follow the most disadvantaged typical trajectory compared to those with the longest education and stable employment. This relates to existing research showing that men with poorer health are at risk of downward mobility [
41] but contrasts the findings by Birkeland et al. [
11] who found support for mental-health related social selection for girls only. However, our results were unstable and the overall conclusion from the analyses is, in accordance with some other studies [
32,
42], that there is poor evidence of effects of youth depressive symptoms on adult typical trajectories of education and LMA. At the same time, and rather surprisingly, when adjusting for parental unemployment, depressive symptoms in adolescence showed
reduced likelihood of following the trajectory of medium education into stable employment compared to the reference category in men. Are young men who intend to continue their studies more likely to report depressive symptoms than their less ‘study inclined’ male peers? It is possible that this is a reflection of being harassed for not adhering to hegemonic norms of masculinity in which being ‘studious’ does not fit [
70]. It is also possible that the longer education in men who had unemployed parents is a result of poorer mental health, for example that they had difficulties completing their degree, lacked motivation, or were enrolled in courses without really wanting to. Further speculations may be that these young men wished to work but were discriminated against at the labour market because of poor mental health and that they therefore chose to study.
Our results regarding the lack of association between level of youth depressive symptoms and trajectory membership can be understood from several perspectives. First, it might reflect a buffer effect of the educational system in Sweden. Although many young people experience depressive symptoms, the most common pathway for young people in the early 1980s was to continue to upper secondary school/senior high school after nine years of compulsory schooling (80% of 16-year olds) [
71]. Hence, there was a low risk of school drop-out which has been found to increase the risk of labour market exclusion and long-term welfare benefit dependency [
72]. In addition, research suggests an ‘equalising effect’ of school regarding the social gradient in health [
73]. According to this hypothesis, socioeconomic differences in health are greater before and after compulsory education, suggesting that schooling and the school environment may decrease the socioeconomic gap in health during this phase of life. If young people have the chance to continue in school despite elevated levels of depressive symptoms, it is possible that the selection effect is reduced.
Second, the findings might also reflect the buffering effect of the Swedish welfare state, providing a foundation of social and economic protection, for example via free health care and active labour market policies. For example, the school health care was well developed in Sweden at this time and all upper secondary schools/senior high schools had school nurses and counsellors employed to which young people could turn if they needed support for mental health troubles. The low level of youth unemployment at the time, together with the previously mentioned national policies regarding labour market programs, might also have contributed to a sense of security and predictability among young people and less discrimination of individuals with mental health problems than would have been the case in a more competitive insecure situation.
Third, it also possible that the depressive symptoms experienced at age 16 years (for both boys and girls) were not severe, long-term, or frequent enough to show lasting effects on their educational and employment-related choices later on in life. The symptoms assessed here represent relatively minor problems as opposed to, for example, clinically diagnosed major depression, which has been found to have long-term effects on people’s life chances [
14,
15]. Perhaps the level of depressive symptoms in youth has more immediate consequences or is related to other outcomes in their future lives than education and LMA. For example, according to a study based on the same cohort, men with mental health problems in youth were less likely to become fathers [
74]. It is also possible that the health selection effects are more prominent later in life, as identified in previous research [
34,
40,
47,
49].
Nevertheless, despite the lack of significant associations between depressive symptoms and the typical education and LMA trajectories identified, we would like to point out that the estimates suggest a positive association, that is, the higher level of depressive symptoms in youth, the greater likelihood of following a less fortunate typical trajectory than the one dominated by long education and stable employment. In the case of men, the unadjusted analyses showed that those following the typical trajectory of disadvantage (the continuously unstable trajectory) were more likely to report elevated levels of depressive symptoms than their peers in the long education and stable employment trajectory. This indicates health selection, reflecting what life-course epidemiology would call a ‘chain of risk’ [
75]. However, the adjusted analyses suggest that this association depends on parental socioeconomic position.
To the extent it indicates no discrimination of individuals with mental health problems, it is positive that youth level of depressive symptoms shows a weak association with adult status regarding education and LMA. Our findings suggest that typical trajectory membership is influenced by other factors than depressiveness in youth. This suggests that future research should focus on the importance of indirect selection processes as concluded by Foreskov et al’s analyses of the British household panel [
76]. A next step is to analyse mental health consequences of typical trajectory membership while taking both inter-generational and intra-generational social reproduction into account.
Strengths and limitations
The main strengths of the current study are the long-term cohort material covering 26 years, the low attrition, and the large number of states to model in the sequence analyses. The low attrition implies that those who are most likely to drop out, for example individuals with poor health [
77,
78], are still represented. Nevertheless, some limitations are worth noticing. For example, there is a risk of recall bias because data on activity was reported retrospectively, especially for the nine and twelve year periods between age 21 and 30 as well as 30 and 42. For the calculation of the individual sequences, gaps were imputed based on the states before and after the gap. Although commonly used, this strategy could have overestimated the stability of the trajectories. All data was self-reported which implies a risk of positive and negative affectivity, i.e. that individuals in a negative mood state may recall and report ‘worse’ activities (for example unemployment or sick-leave) than individuals in a more positive mood states [
79,
80].
With regard to the measures used, the variable of seven categories that reflects education and labour market attachment is relatively crude and restricts the variation in individual trajectories. It is also a limitation that the response categories differed across waves, for example that the option ‘education’ was not available for the period between age 30 and 42. Moreover, information on parental income and level of education would have facilitated a more complex analysis of intergenerational social reproduction. Unfortunately, such data were not available. Finally, depressive symptoms is only one aspect of mental health and although the measure used in the current study shows good psychometric properties and corresponds well with symptoms included in the DSM 5 manual, there is always a risk that the findings are strongly dependent upon the measure used. It is possible that the results would have been different had we used other indicators of (mental) health. Thus, our zero finding about mental health selection should be generalised with caution.