In terms of the research questions posed in the introduction, the main results of the study can be summarized as follows: a) There is a consistent social gradient in perceived stress irrespective of gender or the SES measure used. In general, adolescents with lower SES report higher levels of perceived stress. However, the amount of variance in perceived stress accounted for by SES is very small. b) There is also a social gradient in the number of negative life events experienced during adolescence, and c) the ways Danish adolescents attempt to cope with stress appears to be socially patterned. d) Life events and, to a greater extent, coping are associated with perceived stress. e) Furthermore, both life events and coping partly mediates the SES differences in perceived stress. When both life events and coping are included in the model between 37 and 56% of the social gradient in perceived stress is mediated, depending on gender and the measure of SES employed. Sobel’s tests confirmed the existence of mediation for all the associations with the exception of coping style as a mediator of income differences in perceived stress for boys. f) Finally, the analyses presented here indicate that different measures of SES are related to perceived stress in slightly different ways, that is, via different mediational pathways. In general, parental education appears to have more influence on female adolescents’ stress, while household income appears to have more influence on male adolescents’ stress.
SES and perceived stress
It was mentioned in the introduction that research into SES differences in stress amongst adolescents has been sparse, but that others have found higher levels of perceived stress among adolescents from lower SES families. Goodman et al.  report a similar finding, although in their sample, the association was only present when parental education was the indicator of SES, and absent when household income was employed. Both measures of SES were related to perceived stress in our sample, though the amount of variance in perceived stress accounted for by SES was very small. It is likely that both the amount of variance accounted for and the question of which SES-markers will be related to stress will be affected by contextual factors in the area in which a study is undertaken. It is possible, for example, that the amount of variance that can be accounted for will be larger in countries with a less egalitarian welfare system than that found in Denmark. Another possible explanation of the rather small differences in perceived stress across SES could be the inadequacy of tapping into actual status hierarchies influencing adolescents’ psychological well-being using information on parental status. It may be that the equalization of SES-differences in adolescents across a range of different measures of health  is caused by inadequate measures of SES among adolescents because their socioeconomic status is partly caused by the resources provided to them by their parents but is partly influenced by their own prestige and status in peer groups where different mechanism might be working (e.g. popularity due to sports achievements, intelligence, physical attractiveness etc.). If this is the case only the use of more subjective measures of SES could remedy this. Others have also reported higher frequencies of negative life events amongst lower SES adolescence [e.g. [36, 37, 54]].
SES and coping
Research into associations between SES and coping strategies is meagre . Wills et al.  examined associations between SES and many different coping variables, amongst them ‘helpless coping’, which consisted of items from Carver et al’s  disengagement coping scale. Items from this scale constitute 50% of the items in our avoidance coping scale. ‘Helpless coping’ was found to be significantly related to parental education; higher parental education was related to lower levels of helplessness, which is in line with our results. More adaptive forms of coping were found to be positively related to parental education.
Both life events and coping were, as expected, related to perceived stress levels. There are at least two possible explanations of why coping was more strongly associated with perceived stress than were life events. The first explanation concerns temporal proximity. While the life events measured here occurred throughout childhood, our measure of coping concerns how adolescents generally attempt to manage stress in the present. Coping is therefore closer in time to present stress levels than are life events. Many mediating and moderating variables intercede between initial stressors, such as life events, occurring during the course of life, and present day stress levels. Coping styles are likely to be the result of expectations of control, which in turn are shaped by previous experiences of success and failure in dealing with the challenges of life and an awareness of available resources. Coping is “farther along on the psychosocial chain as a mediator” . Secondly, there is likely to be a greater degree of causal reciprocity between coping and perceived stress than between life events and perceived stress. That is, stress levels are not purely a result of coping, but also something that partly determine the choice of coping strategy. When stress levels are higher, as they are for lower SES adolescents, this will often reflect expectations of inadequate control, which is likely to promote avoidance rather than attempts to change the (uncontrollable) situation.
Life events and coping as mediators
The main aim of the present study was to examine the extent to which life events and coping mediated SES-differences in perceived stress. In the present study, both life events and coping mediated some of these differences. For both boys and girls coping appear to be a stronger mediator than life events, irrespective of which SES-indicator is employed. In general, coping mediates SES-differences in perceived stress to a larger degree among girls than boys. This is most apparent for the gender difference in the effect of household income mediated by coping where 37% of the income differences in SES are mediated among girls and the mediation effect is not statistical significant for boys. These differences cannot be explained by the data available. It is possible that gender differences in socialisation and maturational processes play a part. That is, girls of this age may be psychologically more mature on average than their male peers, which might affect the extent to which their coping reflects strategies and habits that are learned from their parents. This may therefore be a very age-specific gender difference. It is also possible that girls for some reason are more sensitive to childhood conditions connected with parental education, that parental education has a more formative influence on coping for girls, while boys’ coping strategies are more influenced by material conditions reflecting household income. Perhaps higher parental education seems to promote more active coping and less avoidance coping.
Unfortunately there is very little in the literature that can help unravel such speculations. We have found only one previous study that investigates possible mechanisms connecting SES with perceived stress among adolescents by looking for mediation effects. Using cross-sectional analyses, Finkelstein et al.  examined the extent to which optimism and adaptive coping influence the relation between parental education and perceived stress. SES was related to stress, optimism and ‘disengagement coping’, but not ‘engagement coping’. Higher optimism and engagement coping were also related to less stress, while higher disengagement coping was related to more stress. Conceptually these coping measures appear very similar to the distinction in the present study between active coping and avoidance coping. Engagement coping and active coping entail attempts to deal directly with problems that one acknowledges, and is considered more adaptive, while disengagement and avoidance coping involve cognitions and behaviours that aim at creating distance to problems which one tries not to acknowledge. However, unlike in the present study, adding the coping variables to the regression model did not attenuate the effect of SES on stress. While optimism was found to mediate the SES – perceived stress relationship, no evidence of a meditational role for coping was found. This may reflect the fact that relations between coping and stress and between SES and coping appear weaker in Finkelstein et al’s study than in ours. The age range in Finkelstein et al’s sample was larger (12–20 years) compared with our birth cohort. Since coping behaviours are likely to change quite dramatically from year to year during this crucial developmental period, employing a sample with such a large age-range may mask differences that are present at a more specific age. Furthermore, there is recent evidence that the developmental trajectories of psychological resources such as self-esteem and mastery, which are likely to affect coping, are influenced by social class. Falci  found that high SES adolescents experience steeper gains in self-esteem and mastery compared to low SES adolescents during the high school years.
It is also possible that different variables will mediate the SES-stress relationship depending on the social, cultural and economic context. This may especially be the case with coping. If coping behaviour is in part socially determined, as the results of the present study suggest, then its measurement is likely to be sensitive to the social context in which it occurs. This presents researchers of adolescent stress and coping with a challenge. One approach would be to create context specific coping measures, which might make it easier to uncover relationships if they exist. The drawback of such an approach would be to muddy comparisons between studies and contexts.
While Wills et al.  did not include a measure of perceived stress in their study; they did test whether or not life events and coping mediated their observed relation between parental education and adolescents’ substance use. Life events were found to mediate this relation but coping variables did not.
Chen et al.  used two composite measures of SES, one combining parental education and occupation into a prestige-based SES measure, and one combining income and savings into an assets-based SES measure. Their cross-sectional study employed a sample of 15 to 19 year old adolescents, and tested the role of psychological interpretations of threatening and ambiguous stimuli in the relationship between low SES and physiological stress responses. They also collected data on negative and positive life events during the previous 6 months. Prestige SES was related to adolescents’ interpretations of ambiguous videos such that lower SES adolescents were more likely to interpret these as threatening than higher SES adolescents. This was not found when the analysis was repeated using the assets based SES measures. Lower prestige based SES was also related to higher heart rate reactivity and diastolic blood pressure while talking about the ambiguous video. Again, this was not found when using assets based SES. Threat interpretations partially mediated relationships between SES and physiological reactivity. General life events (e.g. a lack of positive events) partially explained the relationship between low SES and threat interpretations. Chen et al.  note that “…. the experience of repeated and unpredictable negative life events in low-SES children may lead them to develop patterns of thinking in which they come to expect the worst even in ambiguous circumstances.” Using this reasoning in the interpretation of our own results, this may in turn lead to a greater use of avoidance coping amongst low SES adolescents. This illustrates how complex the relations between life events (and other sources of stress), coping and perceived stress are likely to be in reality. One may question the meaningfulness of trying to discern whether differential exposure or differential impact best explains SES-differences in perceived stress, since coping behaviours are presumably formed through socialisation and learning processes throughout childhood, and these processes will in turn be influenced by exposure. In this way vicious cycles may develop, so that greater exposure to negative and uncontrollable life events leads to expectations of less control and more threat, which in turn promotes more avoidance, which reduces the likelihood of successfully dealing with problems, even when such problems are objectively controllable. In other words, greater exposure may lead with time to greater impact. It is important that ‘dispositional’ measures of coping strategies, such as those used here, should not be thought of purely as properties of the individual, since they partly reflect socialisation to differential social conditions that promote or restrict these strategies. As Baum et al.  note, many psychological factors “are products of environmental exposures associated with lower or higher SES, such as learned helplessness, a sense of powerfulness, or an orientation towards mastery and efficacy…”
Using two measures of SES – advantages and disadvantages
Our results, in keeping with the studies discussed above, confirm the need to employ different indicators of SES in the same study. We found that both parental education and household income are related to adolescents’ perceived stress, while others have found this relation for parental education but not income . On the other hand, we found that the amount of the SES – perceived stress effect that could be explained by life events and coping was to some extent dependent upon the measure of SES used. In general, mediation was strongest when parental education was used. An exception is seen in model 2 for girls (Table 4), where the choice of SES measure does not have much influence on the extent to which life events mediate the SES – perceived stress relation. The choice of SES measure influences results most clearly in model 3 for boys (Table 3). Here, coping is seen to mediate the relation between parental education and stress, but not the relation between household income and stress. The present study suggests that employing different indicators of SES could provide a means of examining gender differences in meditational processes connecting SES with stress.
We found that a clear social gradient in life events was apparent irrespective of SES-measure. Brady et al.  also found greater life event exposure in lower SES adolescents. Interestingly, different indices of SES were predictive of different types of life events. For example, greater assets (a composite measure of material resources) predicted fewer negative life events that were considered to be independent of respondent behaviour, and fewer total life events, while fathers’ education predicted fewer negative life events that were not independent of respondents’ behaviour. The study also included a measure of positive life events, but these were not related to any SES index.
Strengths and limitations
Several limitations of our study need to be addressed. Firstly, given the cross-sectional nature of the study, the meditational analyses cannot conclude whether life events and coping operate in a causal fashion. Both could account for some of the variance in the SES-perceived stress relation. Given that major life events, such as parental divorce or death, were reported, it seems unlikely that present day stress levels would influence recall or actual event occurrence. With coping however, there is likely to be more reciprocity. As mentioned previously, a greater use of avoidance coping could conceivably be a consequence of higher stress levels as well as a cause. Our results do however indicate a need to further investigate the possibility of a mediating role for coping. A second potential limitation concerns the possibility of response bias. As might be expected, non-responders were more likely to have lower SES. However, since all SES-groups were well represented in the sample we consider any potential for response bias to be minimal. Ensuring sufficient numbers in each SES group was in part achieved via our strategy of gaining access to schools and asking pupils to fill out the questionnaires there, during class-time. Another possible limitation is that we only collected data on negative life events. Given Chen et al.  finding that lack of positive events, rather than a greater frequency of negative events, mediated the relation between SES and physiological stress reactivity, the present study could have been strengthened by measuring both types of events. A final possible limitation concerns the generalizability of our findings. As our sample we used a whole birth cohort from a well defined region in Denmark. The region is predominantly rural and relatively affluent. Denmark is one of the richest countries in the world. Because of a progressive tax system and a welfare state, that ensures free education and health services for everyone, material differences between rich and poor are relatively small. In less egalitarian societies one might expect SES differences in perceived stress to be larger. It was also mentioned above that different variables might mediate the SES-stress relationship depending on the social, cultural and economic context. Whether or not specific coping strategies, for example, are beneficial, is indeed likely to be very context dependent. The consequences of life events, such as parental divorce, in terms of changed life trajectories, chronic living conditions and future possibilities are also likely to be dependent upon the societal conditions in which they occur. Again, the Danish context of the present study seems more likely to weaken rather than intensify the ‘stressfulness’ of low SES, compared to countries with a less established welfare system.
The present study also has a number of strengths. Firstly we employ data from a large cohort study with a very high response rate. Secondly, since we use two independent sources of data, self-report questionnaire data combined with data from national registers, the relationships found between SES and perceived stress cannot be accounted for by common method variance. Thirdly, the use of national registers enabled a reliable and objective categorisation of participants into SES-groups. Fourth, because we use two different measures of SES, our study can contribute to the ongoing discourse regarding the nature of SES and its relation with stress. While income and education are related, our analyses suggest that they may exert an influence on stress via different mechanisms, and that these mechanisms to some extent vary according to gender.