Introduction

Early marginalisation from working life among young people has caused growing concerns in several industrialised countries [1]. Understanding the determinants leading young people to dependency of social benefits rather than starting an education or entering working life is therefore of great interest for both researchers and decision makers, especially in times of economical decline. It is generally accepted, that education improves job prospects in general and the likelihood of remaining employed in times of economic hardship [2]. Also, positive early work experiences and education are critical to development of young people into healthy adults [3]. The risk of being excluded from the educational system and the labour market has shown to have negative health consequences, and during a severe economic recession, entry into working life for younger people becomes harder to achieve [1]. OECD finds that the worsening conditions on the OECD labour market are more severe for younger workers than for older workers. On the other hand, it could be argued that the ageing of the population and the decline in the population of 15-29 year-olds in OECD countries favour employment among young adults [1]. Nevertheless, in order to ensure future demands of labour and decrease the individual and societal costs associated with poor labour market participation, it is important to understand the mechanisms behind not entering into education or work in the early years of life, and rather rely on social benefits. Previous studies have concentrated on adulthood predictors of receiving social benefits, such as health status, sickness absence, work environment and socioeconomic status (SES) [4–6]. Extending the focus to cover potential determinants from earlier life circumstances seems to be justified: A Finnish longitudinal cohort study found a 3.5-fold risk of disability pension when the respondents had experienced 5–6 negative events (divorce or separation of parents, long-term financial difficulties in the family, frequent fear of a family member, severe illness of a family member and alcohol-related problems of a family member) during childhood [7]. A British birth cohort study on life course approach to long-term sickness absence, found that a measure of IQ, at age 11, had an impact on long-term sickness absence measured at age 42 [8]. Additionally, experience of stressful life events is a well-established risk factor for health measures such as the development of depressive problems in adolescence and adulthood [9–15]. However, expanding focus to include determinants from early life may increase our understanding of the complex process and determinants non-participation in working life, and provide insight into the mechanisms that determine why working life for some is over before it has begun.

The purpose of this study was to investigate whether adverse circumstances in terms of specific negative life events (NLE) in childhood determined later labour market participation, and to identify which negative life events constituted the largest risk. The analyses took into account the simultaneous effects of socio-economic position, school grades, educational plans, vocational expectations and general health at age 14-15.

Methods

Results

A total of 17.1% were categorized with low labour market participation at follow-up, significantly more men than women (Table 1). Across gender, a total of 78.2% did not receive social benefits at any time during the 12-month follow-up period, whereas 10.8% received benefits up to 25% of the 12-month period, an additional 4.2 received benefits from 25% to 50% of the period, further 2.4% from 50% to 75% and finally 4.3% from 75% to 100%. Of the latter group, a total of 2.7% received social benefits for the full 12-month follow-up period. Looking at the combined bulk of social benefits during the 12-month follow-up period, the majority of benefits (80%) fell within unemployment related benefits, whereas as 20% were health related (sickness absence compensation). Analysing the use of social benefits from age of eligibility (18 years), showed that 9 individuals (0.3%) were on permanent disability pension, 18.3% had at some point received sickness absence compensation, whereas 11% had at some point received unemployment benefits (not shown). Unpublished analyses have confirmed that responders and non-responders do not differ on educational and vocational outcomes (Winding TN, personnel communication).

The MacArthur Scale score of self-assessed SES strata in childhood was only marginally associated with future LMP. The number of childhood adversities experienced before age 14-15 was significantly associated with future LMP, and especially parents’ divorce, parents’ alcohol/drug abuse, having been abused or having witnessed a violent event, was associated with low labour market participation. Having fixed educational plans at the age of 14-15 and firm vocational expectations were not associated with future LMP, whereas school grades were. Self-rated health at age 14-15 was not associated with later LMP (Table 2).

The number of NLE was significantly associated with future LMP among females in a dose–response manner: The unadjusted ORs for having experienced 1-2 or 3-5 NLE vs. 0 were 1.54 (95% CI 1.13-2.11) and 3.50 (2.09-5.85) respectively. This only altered slightly when taking into account the simultaneous effects of SES, school grades, educational plans, vocational expectations and self-rated health, yielding ORs of 1-2 or 3-5 NLE of 1.54 (95% CI 1.07-2.22) and 3.37 (95% CI 1.82-6.24) respectively in the most adjusted model (Table 3, model IV). The risk associated with childhood adversities was generally smaller for males, and became insignificant when entering school grades and educational plans and vocational expectations into the model (Table 3). For men, test for trend was significant for models 1 and 2 (p=0.01 and 0.05 respectively), but insignificant in models 3 and 4. For women, test for trend was significant in all models (not shown).

Among females, separate analysis of the individual NLE items yielded a significant effect of experiencing ones’ parents divorcing, having parents who abused alcohol or drugs, being abused by someone known to the respondent, and having witnessed a violent event. These effects persisted independently of SES score. The effect of parents’ alcohol/drug abuse became insignificant when entering the four school variables into the model, leaving parents’ divorce (OR 1.82, 95% CI 1.26-2.63), being abused (OR 5.10, 95% CI 2.60-9.97) and having witnessed a violent event (OR 2.47, 95% CI 1.54-3.94) as significant risk indicators in the most adjusted models (Table 4).

For males, the crude models yielded a significant effect of having parents who abused alcohol/drugs, having been abused, and having witnessed a violent event. The associations became insignificant when entering the four variables assessing school grades, educational plans and vocational expectation into the model, yielding no significant associations of childhood adversities on future LMP among males.

Attributable fractions (AF) were calculated for childhood adversities based on the most adjusted models. For women, AFs ranged from 2.4% for parents alcohol/drug abuse, to 16.1% for parents divorce. Being abused in childhood attributed 12.7% of the excess risk of future low labour market participation. For males, NLE explained less than was the case for women: AFs ranged from 1.0% (parents abused alcohol/drugs) to 4.9% for those who witnessed a violent event.

Discussion

The present study showed, that among a cohort of 14-15 year olds, 17.1% received social benefits for at least 4 weeks in a 12-month period seven years later at the age of 21-22 years. This was the case for significantly more males (19.9%) than females (14.4%). The risk of low labour market participation increased if one had experienced certain negative life events in childhood, especially for females. Females who in their childhood had experienced their parents’ divorce, had been abused, or had witnessed a very violent event, had increased risk of future low labour market participation, when adjusting for SES, school performance, educational plans, vocational expectations and general health at age 14-15. Especially those who had been abused by someone they knew had highly increased risk (OR 5.10, 95% CI 2.60-9.97). This findings supports the findings by Harkonmäki and colleagues, who found frequent fear of a family member to be the strongest predictor of disability pension in a similar battery of NLEs in childhood [7].

Attributable fractions of various NLE ranged from 2.4% (parents’ alcohol/drug abuse) to parents divorce (16.1%) for women. For men, risk estimates were generally lower and tested statistically insignificant in the most adjusted models. Attributable fractions ranged from 1.0% (parents’ alcohol/drug abuse) to 4.9% for witnessing a very violent event.

When interpreting the study findings, one should consider the potential error induced by the 7 years from baseline to follow-up, where we have no information about the study participants. Life events could occur that affect LMP, which were unmeasured in this study. An individual experiencing for example an NLE during follow-up, would appear in the group of unexposed in the performed analyses. This could potentially underestimate the effects of NLEs on LMP. Regarding the measurement of NLE, this covers events taking place in an age span of 0-14/15 years of age, and therefore covers both early childhood, middle childhood and adolescence. Asking about events in early childhood could potentially introduce recall bias. This could potentially underestimate the effects of NLEs on LMP. However, considering the severity of the NLEs assessed in this study, we consider recall bias to be a minor issue. Another concern is the subjective nature inherent in certain of the NLE items. Surely, whether or not ones parents have been divorced or have died is not open for interpretation, but, for example, the degree to which an event is “very violent” or not, is to a large extent a matter of personal norms. Unpublished analyses of the data showed, that females more often report to have witnessed a very violent event than males did, but whether or not this is due to females witnessing more violence or interpreting violent episodes differently than males do, is unknown. However, in relation to this study, we argue that it is the experience that constitutes the adversity, rather than, in the case of this example, how one assesses a specific level of violence.

On the positive side, the study features a relatively high participation rate of 83%. An additional strength is the 100% follow-up due to use of register data, which furthermore eliminates common method variance [24]. A further advantage compared to the study by Harkonmäki and colleagues [7] is this study’s assessment of adverse circumstances in childhood while the study participants are still children, which minimizes potential recall bias compared to asking adults about events in early childhood.

The present study does not shed light on the pathway from NLE to future LMP: It can be hypothesised that NLE affects health [9–15,25], which again can affect LMP. To which degree this is the case for the participants in this study is unknown, but the larger proportion of the social transfer payments defining “low LMP” were related to unemployment benefits rather than health related benefits. We adjusted for health in Model 4 only, as the role of health as a confounder is dubious. Previous studies have also shown a strong association between socio-economic status and not only health [1,3,26–28], but also labour market attachment [4–6]. The present study only found a borderline significant association between socio-economic position in childhood and LMP in young adulthood. According to Rahkonen et al., both past and present socioeconomic status are important determinants of adult health [27]. This discrepancy could be due to the assessment of SES in this study: Strong evidence of the construct validity of the MacArthur scales is found in both middle-aged and older men and women [28], and the validity in relation to young people has been tested with good results previously [29–31]. However, Rahkonen et al. argue that the socioeconomic status of destination, that is one’s own education, has a higher impact on health status across age groups [27]. The weak association between childhood SES and LMP in young adulthood could also be due to the relatively young age of the study participants at follow-up: The differences in labour market participation caused by socio-economic inequalities in childhood might not have come through fully in the early twenties, especially not in Denmark, where people generally are older before entering working life compared to many other countries.

Conclusion

The effects of negative life events in childhood should be taken into account when considering both the determinants of labour market participation, as well as when identifying high-risk populations. Recognition and alleviation of negative life events in childhood should be promoted in relation to policies and practices aiming at integrating youth, especially young females, into education and working life.

Acknowledgments

Claus D. Hansen, Department of Sociology and Social Work. Aalborg University, Aalborg, Denmark is hereby acknowledged for his efforts in establishing and developing the underlying research project VestLiv.