Societal engagement in adolescence and future connection to education and employment in early adulthood – A follow-up-study

Background: Attainment to education and employment is essential for young people to develop the skills needed to participate in society and maintain a stable connection to the labour market in adult life. This study examined associations between societal engagement, measured by leisure time activities (LTA) and part-time work (PTW), in mid (14/15 years) and late (17/18 years) adolescence and a positive connection to education and employment in early adulthood, accounting for potential gender and socioeconomic differences. Method: A cohort of young people born in 1989 was followed in a prospective study with questionnaires in 2004 (n=3,054) and 2007 (n=2,400) where information on LTA and PTW was collected. Information on connection to education and employment was collected from a register of social benets when they were 25-29 years old and divided into high and low societal connection. The associations were examined using logistic regression and stratied by gender and childhood socioeconomic groups. Results: PTW was, both in mid and late adolescence, positively associated with societal connection (OR: 1.7 [95% CI 1.3; 2.2] and 1.9 [1.4;2.6]). LTA in mid adolescence was associated with societal connection (OR:1.6 [1.2;2.1]). Among the males LTA and PTW showed strongest associations with societal connection in mid adolescence (ORs up to 2.2), whereas the associations for females seemed strongest in late adolescence (ORs up to 2.8). Conclusion: The study showed that adolescent societal engagement had notable positive associations with connection to later education and employment, with stronger impact of PTW compared to LTA. The study identied differences between genders and the timing of engagement. Associations were consistent across socioeconomic groups.

. Among the males LTA and PTW showed strongest associations with societal connection in mid adolescence (ORs up to 2.2), whereas the associations for females seemed strongest in late adolescence (ORs up to 2.8).
Conclusion: The study showed that adolescent societal engagement had notable positive associations with connection to later education and employment, with stronger impact of PTW compared to LTA. The study identi ed differences between genders and the timing of engagement. Associations were consistent across socioeconomic groups.

Background
Around 170,000 young Danes between 15-29 years of age do not attain an education or work (1,2), and the negative consequences of youth unemployment persist into adulthood (3). It is a signi cant challenge in a welfare society as Denmark to aid and support all children and adolescents to enter adulthood with an education and with the skills needed to acquire a stabile job situation in the labour market (4). In 2018, Denmark spent 28% of the Gross Domestic Product (GDP) on societal expenditures (5). Due to an aging population, there is an increasing need to maximise the labour market participation, and there are several labour areas that are currently unable to get su cient employees, thus losing revenue. If young adults do not participate in the labour market, the society is both deprived of labour and challenged with increased expenditures in terms of social and unemployment bene ts (4,6).
Young adults without education and employment are at risk of disconnection from society (2,7). Being disconnected has negative consequences for the personal health and increases the risk of poor somatic or mental health, delinquency or premature death (8)(9)(10). Disconnection is in uenced by family background, and the negative in uence often persists throughout life with lower education, employment and income as results (2,11). Half of the disconnected 18-24 year olds in Denmark had disconnected parents (9). Disconnection varies between genders with more men being disconnected than women and often to a greater extent as well (7,9). In order to reduce the personal and societal costs of unemployment, and improve the future demands for labour, it is imperative to investigate factors and mechanisms associated with disconnection.
Part-time work (PTW) is paid employment while the employee is enrolled in an educational program.
Stressors experienced in PTW have been found to buffer the effect of more severe stressors later in life (16). Danish adolescents who worked 20-30 hours per month had less school absence and better grades in Danish, Math and English when nishing 9 th grade (12) and PTW between the age of 13 and 16 is found associated with a 40% increased chance of connection to education and employment by the age of 25 for young Danish adults with a non-western background, compared to non-western young adults without PTW (17). However, PTW is also found associated with lower levels of educational completion, lower grades, increase risk of dropping out, or reduce enrolment into tertiary education (18)(19)(20)(21).
Leisure time activities (LTA) are activities chosen for enjoyment, relaxation, or other emotional reasons, typically after school hours. The speci c type of LTA and its societal context affect social, educational, physical and civic development and shape young people's future lives (22). Adolescent LTA, performed alone or in unstructured settings, is associated with the risk of disconnection in early adulthood, lack of educational quali cations, unemployment, smoking, drinking, and crime (22,23), whereas organized and adult supervised LTA tends to increase later life outcomes positively e.g. satisfaction with life and physical and mental health. Participation in activities such as scouts and church clubs has been associated with positive adult outcomes regarding economic status, family situation, health and wellbeing, and citizenship (23)(24)(25). Relations with peers and activity leaders improve emotions and psychological adjustment, abilities that can be carried on into later life societal connection (22,26).
In order to reduce the number of disconnected young adults, it is important to identify tangible actions which can improve connection to society. The positive effects of PTW and LTA on later life competencies may buffer the risk of disconnection. When adolescents participate in PTW and LTA, they inevitably engage into the society surrounding them. In return, they develop new skills and traits. However, research on societal engagement and future connection to education and employment has been scarcely reported in the literature.
The primary aim of this study was to examine whether skills and traits attained through societal engagement (PTW and LTA) in mid and late adolescence were associated with connection to employment or education in early adulthood. Secondarily, the study aimed to investigate to what degree the associations between societal engagement in mid or late adolescence and connection to education and employment in early adulthood varied across gender and socioeconomic groups.

Design and population
The study was a prospective follow up study. Data was collected as part of the VestLiv Cohort, an ongoing longitudinal Danish cohort study following a complete regional cohort of young people born in 1989 who lived in the western part of Denmark when the cohort was established in 2004. The source population comprised 3,681 young people. In the rst wave of questionnaires in 2004, 3,054 (83%) participated. All the potential responders in 2004 have been re-invited to participate at the later waves. To date, the VestLiv Cohort has included waves of questionnaires in 2004 (age 15), 2007 (age 18), 2010 (age 21) and 2017 (age 28) (27). In the second wave, 65% (n = 2,400) of the source population participated.
Besides information from the participants, information was also obtained from the parents in the rst wave. A more thorough description of recruitment and data collection in VestLiv has been presented elsewhere (28).
Participants who answered questions about PTW and/or LTA in the waves of 2004 and/or 2007 were included in the study. They formed two separate study populations representing mid and late adolescence. The number of responders varied in the two waves: 2004 (n = 3,054) and 2007 (n = 2,400). Responses to at least one question about PTW or LTA were obtained from all 3,054 participants in the rst wave and from 2,366 participants in the second wave. Figure 1 illustrates the inclusion in both study populations.
Furthermore, register data was obtained from Statistics Denmark. Using the personal identi cation number (CPR number), which is given to every citizen in Denmark (29), a linkage between the participant and the parents was created through a unique family ID number, and subsequently between the different registers and databases using their CPR numbers.

De nition of outcome
The outcome "Connection" was de ned as the amount of connection to education and employment from the age of 25 to the age of 29. Lack of connection was measured using weekly registrations in the DREAM registry (Danish Register for Evaluation of Marginalization), a longitudinal database which contains weekly registration of the extent and kind of social bene ts such as unemployment bene ts, sickness bene ts and disability bene ts that Danish citizens may receive (30). All sorts of bene ts were categorized as lack of connection to education and employment, except maternity leave, educational grants and supplementary unemployment bene ts, which were categorized as positive connection to education and employment. The participants were followed in DREAM from 1 st of January 2015 to 2 nd of December 2018 (approximately 4 years). Based on the total number of weeks of received social bene ts (disconnection) reported in the DREAM registry (31) over the 4 year period, participants were dichotomized into low or high connection to education and employment.Participants with ≤ 52 weeks of receiving social bene ts were assessed as having high connection, those with >52 weeks of receiving social bene ts were coded as having low connection. DREAM data was not available for the last 4 weeks of December 2018 at the time of analysis. This means that participants born in December 1989 (n = 234) potentially could lack up to 4 weeks of registered disconnection.

De nition of exposures
The exposure variables contained in the term "societal engagement" were based on questionnaire information from 2004 and 2007 about adolescents' PTW and participation in LTA. Societal engagement was both examined in mid adolescence (14/15 years) and late adolescence (17/18 years) to identify if the timing of societal engagement played a role for future connection to education and employment.
LTA is de ned as activities carried out in places like after school clubs, scout troops, sports clubs, bands or voluntary activities such as church groups or theatres. Participants were categorized as having high LTA engagement if they had marked daily or weekly activities in at least two of the four categories of activities: 1) after-school classes/scout/youth club, 2) choir or orchestra/music group, 3) sports club or 4) other clubs. Otherwise they were categorized as having low engagement.
PTW was dichotomized into engaged or not engaged according to if they had a job regardless of the amount of work hours per week. In addition, PTW was recoded into a categorical variable based on the number of work hours per week "0 hours/w" "1-12 hours/w" ">12 hours/w". The cut-off values were based on studies indicating that higher amounts of work hours whilst still at school may have a negative effect on educational completion and attainment (14,20) and The Danish Working Environment Authority, which allows adolescents in compulsory education to work 2 hours per school day and up to 12 hours per week when attending school (32,33).

Adjustment variables
Equivalated childhood household income was generated by averaging the mean of the yearly household income if data was available for at least 3 of the 4 years, when the participants were 7-10 years old, using data from the FAIK (family income) register (34). Equivalated household income is a weighted measure taking household size and composition into account. The weighing is based upon the "OECD-  The variable chronic diseases were constructed based on responses to the questionnaires. In 2004, the parents answered whether the participants had any chronic conditions. In the 2007 wave, the participants answered the same question themselves. All conditions were evaluated by the main author and dichotomized into two categories, no and yes. Chronic diseases included all conditions expected to impact the ability of societal engagement or connection to education and employment. The 'No chronic diseases' category included participants without any chronic diseases or diseases not expected to affect the outcome or exposure in the study.
Depressive symptoms were assessed with an abbreviated 4 item version of the Centre for Epidemiologic Studies Depression Scale for Children (CES-DC) (40). The CES-DC scale measures the level of current depressive symptoms. All questions were answered on a four-point Likert scale with the options: "not at all", "a little", "some" and "quite a lot". Each answer was scored between 0 and 3, with high values corresponding to having depressive symptoms. In this study, depressive symptoms were dichotomized into absent or present at a cut point of more than 3.
Self-rated health is a valid measure of current physical health status, and a strong predictor for future outcomes such as morbidity and mortality (41,42). Self-rated health was assessed by a single question from the 36-Item Short Form Health Survey (SF-36). The participants were asked to rate their health on a ve-point Likert scale. Self-rated health was dichotomized into two groups; "good" self-rated health if participants rated their health as excellent or very good or "poor" self-rated health if participants rated their health as good, less good or bad.

Statistical Methods
Characteristics of the study populations and connection to education and employment in early adulthood were displayed according to gender and differences were tested applying Pearson chi-squared test (Table  1). Univariate analyses were performed to describe associations between each covariate and connection to education and employment in early adulthood (Crude analyses, Table 2 and 3).
In both study populations, correlation analyses were carried out for all covariates using Pearson's rank correlation coe cients. The tests showed no sign of multicollinearity (results not shown). The strongest correlation was between depressive symptoms and self-rated health with r = 0.26 at both time points.
PTW and LTA were only slightly correlated (r = 0.094 (2004) and r = 0.064 (2007)). Furthermore, there was no interaction between PTW and LTA in any of the study populations.
Logistic regression models were used to estimate the associations between societal engagement in mid or late adolescence and connection to education and employment at the age of 25-29 years. The adjusted analyses were additionally strati ed by gender to examine a possible gender difference and to facilitate the comparison of results with other studies. Results are presented with crude and adjusted odds ratios (OR) and 95 % con dence intervals (95% CI) ( Table 2 and 3, Adj. models). The adjusted models were strati ed by parental educational level and subjective social status to investigate the strength of the associations across different socioeconomic groups and measures of socioeconomic position (SEP) (Table 4a and 4b).
Data analyses were performed using STATA statistical package (version 15.0; Stata, College Station, TX, USA).

Ethics
Use of the data was carried out under the same conditions and with the same purpose as originally collected. The study was approved by the Danish Data Protection Agency, according to Danish law for studies using questionnaire and register data (The Act on Processing of Personal Data -Act No. 429 of 31 May 2000). Written informed consent was not required in questionnaire or register-based projects. Table 1 shows that the 2004 study population consisted of an equal number of females and males, whereas the 2007 study population consisted of more females (54%). In both study populations, a larger proportion of females had low connection in early adulthood (13-15%) compared to males (11%). Compared to mid adolescence where females and males were almost equally engaged in PTW, in late adolescence, more females (66%) than males (45%) had PTW and worked more hours. Females and males were equally engaged in leisure activities throughout adolescence but with a notable decrease in total high engagement from 34% in mid adolescence to 13% in late adolescence. More females than males reported depressive symptoms, low self-rated health and low subjective social status in both study populations. Table 2 and 3 show that all exposure variables from mid and late adolescence were associated with connection to education and employment in early adulthood. For all three SEP measures the strength of the associations increased with higher SEP (Table 2 and 3). The crude associations between SEP variables and a positive connection were more profound for males than females (crude analyses, Table 2 and 3).

Results
Across the analyses of societal engagement in mid and late adolescence, we observed associations between PTW and positive connection to education and employment with almost twofold increases for both time points (Table 2 and 3). LTA in mid adolescence was also associated with a positive connection (OR:1.6 [1.2;2.1]) ( Table 2).
Among females, we found that PTW in both mid and late adolescence was associated with an around twofold increase in positive connection to education and employment. When applying PTW as a categorical variable based on work hours in three levels, the same associations were seen for the females in both periods. However, in mid adolescence, working 1-12 hours/w was the category most strongly associated with positive connection (OR: 1.7 [1.2; 2.4]), whereas in late adolescence, the strongest association was seen for working >12 hours/w (OR: 2.8 [1.5;5.5]).
Among males, PTW was associated with a positive connection in both mid and late adolescence with a 70% increased risk in both periods. When applying PTW as a categorical variable the strongest association among the males was between PTW >12 hours/w in mid adolescence and connection to later education and employment (OR: 2.21.0;4.9]).
When stratifying the analyses on childhood SEP, we found varying but positive associations between societal engagement and positive connection to education and employment for all strata in both mid adolescence (Table 4a) and late adolescence (Table 4b). Participants who in mid adolescence considered themselves to have a low social status appeared to bene t the most from societal engagement. In late adolescence, the estimates were strongest for participants from homes with low educational level. However, the number of participants with high engagement in LTA was small (n = 297) resulting in small strata and thus imprecise estimates.

Discussion
In this prospective study, we found that societal engagement in adolescence increased the connection to education and employment in early adulthood. The strongest and most consistent associations were seen between PTW and connection to future education and employment. The timing of societal engagement seemed to have different impacts for females and males. LTA and PTW in mid adolescence were associated with a positive connection for both genders, but with somewhat stronger estimates for males. For females, LTA and PTW in late adolescence appeared to increase a positive connection in early adulthood as compared to males.
The ndings in this study are in line with the ndings from Lesner et al. (12) who found adolescent PTW associated with positive outcomes. In their study, students who had 20-30 hours of PTW per month during the school year achieved better exam grades and were faster to enrol into secondary education (12). These outcomes were measured as more immediate effects of PTW compared to our study, and they did not investigate if it eventually prevented disconnection from education and employment. The results are likewise in agreement with the results of a study of young non-western Danish adults, that found PTW was associated with an increase of 40% in positive connection at the age of 25 (17).
In contrast to Lesner et al. (12), who showed a positive outcome of a moderate number of work hours, the present study observed the strongest associations for participants working more than 12 hours/w. A cut point of 15-20 hours work per week during the school year has been proposed in several studies (14,18,20) as the threshold where the negative consequences of PTW outweigh the positive. Mortimer (14) identi ed that those working steadily a few hours per week most of a school year compared to those only working sporadically and/or working >20 hours per week had the most bene cial outcomes such as improved time management skills (14). Due to the a priori decision about the categorization of work hours, it was not possible to examine if participants working a very high number of hours had lower societal connection later in life.
The strong association between working >12 hours/week, and a positive societal connection in early adulthood is interesting in the light of recent Danish educational policies. Since 2014, the public Danish primary and lower secondary school has gone through a large reform. Pupils in mid adolescence must have 35 lessons per week and now have the longest school days in Europe (43). This study uses data from 2004 and 2007, prior to the reform in 2014, so the positive results found here comes from a period in time where adolescents had more spare time. The longer school days can make it di cult for adolescents to have enough time for both obligatory homework, LTA and PTW, particularly because the Working Environment authorities only allow adolescents to work 2 hours per school day and 7-8 hours on non-school days (32). The results from this study are relevant for policy makers when discussing future educational policies.
In both study populations, more females than males were disconnected in early adulthood. This is in contrast to Benjaminsen et al. (9) who identi ed more disconnected males. This discrepancy is interesting as females in this study worked more in late adolescence and the association between PTW and positive connection was strongest for females. The different results of the two studies can partly be explained by the fact that in Denmark (44), more males in late adolescence work in apprenticeships or vocational education. Therefore, fewer males are likely to have PTW during upper secondary education compared to females.
Another possible explanation for the difference can be related to the period where connection is measured, since many young people establish families at the ages of 25-29. Although our data has taken maternity leave into account, it was not possible to identify pregnancy-related sickness leave in the DREAM database (31). Absenteeism due to pregnancy complications are registered as ordinary sickness, and it is possible that some women were not truly disconnected, but rather on sick leave because of complications due to pregnancy. As a sensitivity analysis, register information about all childbirths from the female participants were collected between age 25 and 28, to which we had data. All recorded sickness bene ts six months prior to childbirth were subtracted the total number of weeks of received social bene ts. A total of 10 females had been categorized with low connection to education and employment instead of high. These females were recoded, and all analyses were repeated. The results only showed minor changes to the estimates (results not shown).
The univariate analyses indicated a social gradient, with participants from higher SEP having a higher connection to education and employment. Studies within the same cohort have previously identi ed a social gradient in dropout from secondary high school (45) and in levels of perceived stress (46). However, the multivariate analyses strati ed by SEP measures showed that the associations between both LTA and PTW and positive connection to education and employment varied within strata depending on the type of engagement, timing and measure of SEP. Some of the strata in the analyses were small, making the associations ambiguous and caution is advised in relation to interpretation.
The present ndings indicate that societal engagement is associated with a positive connection across different socioeconomic groups. However, in line with previous research this study did not nd a compelling social gradient when exploring the association between societal engagement and connection to education and employment. (47). Thus, if LTA or PTW are used as promotional strategies to improve connection to education and employment, considerations should include the de nition of the target group and the timing of implementation.

Strengths and limitations
A strength of the study was the use of a cohort with a relatively high initial participation rate in combination with high-quality register data. The register data was almost complete for all participants. The design was prospective, which allowed an evaluation of temporal associations. Furthermore, the study extended over 14 years, which ensured that almost all of the participants had participated in society through education and employment. Positive connection to education and employment was measured over four years to account for sporadic periods of disconnection. The use of both questionnaire and register-based data in the study is a strength, which minimises the risk of common method variance (48).
The exposures in the study were constructed from questionnaire data. The decision to participate in the original cohort was taken without the knowledge of any subsequent outcome in the study and we assume selection on participation was non-differential, which should not bias the estimates. In this study, information about non-participants' societal engagement and their connection to education and employment in early adulthood was unavailable. Based on the available register information, an analysis of non-participation in the two study populations showed that in both waves, more females than males responded, and a higher proportion of the responders came from families with high income and educational levels. Nevertheless, a previous study in the same cohort has shown that neither nonparticipation nor loss to follow up in uenced the size or directions of a number of exposure-outcome associations signi cantly (28).
Self-reported data is prone to information bias if some participants intentionally or accidently respond incorrectly resulting in misclassi cation of variables. Questions on PTW and LTA were rather harmless and not personal such as questions on weight or unhealthy behaviours (49). Also, we do not expect any misclassi cation to be associated with the outcome, thus any bias would most likely lead to an attenuation of the association estimates. Most adolescents in Denmark are employed on hourly basis, which increases the chance that the reported work hours were true. Yet, the number of hours adolescents work is likely to uctuate during the school year, which the questionnaires were unable to identify. This is anticipated to cause non-differential information bias.
Overall, the limitations identi ed were not considered to cause serious bias in relation to the associations found in the study. However, caution about causal interpretation is important since other factors may have confounded the associations, such as ambition, self-con dence, indomitability or "believing", which have been found to impact both societal engagement and positive connection to education and employment (50,51).
Previous research by Glasscock et al. has shown that the socio-demographical characteristics of the participants of the Vestliv Cohort are comparable to young people in the rest of Denmark (46). Many participants in Vestliv have moved and they are now living in different parts of Denmark re ecting various possibilities for education and employment. Therefore, the ndings from this study are generalisable to young people in Denmark and other countries with similar environmental and social conditions, taking into consideration that Danish adolescents today have longer school days making the balance between school and leisure time different than for participants in this study.

Conclusion
The study showed a notable association between adolescent societal engagement and a positive connection to education and employment in early adulthood, with stronger effects of PTW compared to LTA. Furthermore, the study identi ed both a gender difference in the association between societal engagement on positive connection and in the timing of adolescent engagement. The associations between societal engagement and positive connection varied across socioeconomic groups and depended on the chosen indicator for SEP. Municipalities, social workers and public health workers need to be more attentive to the positive effects of adolescent societal engagement in order to reduce the risk of disconnection in the transition into adulthood.
From a public health perspective, it is important to understand the factors associated with positive connection to education and employment in order to develop effective preventive strategies to match the needs of different groups. Based on the ndings in this study, both establishment of easier access to leisure time activities and increase the amount of part-time job positions for adolescents appear to be relevant and feasible preventive strategies.

Declarations
Ethics approval and consent to participate: The study was approved by the Danish Data Protection Agency. According to Danish law (Act on Research Ethics Review of Health Research Projects), available at: www.nvk.dk/english/act-on-research, questionnaire and register-based studies require neither approval by ethical or scienti c committees nor written informed consent.

Consent for publication: Not applicable
Availability of data and materials: The datasets used and/or analysed during the current study are available from the authors on reasonable request.