Health inequalities between employed and unemployed in northern Sweden: a decomposition analysis of social determinants for mental health

Unemployment in a Swedish context

The present study unfolds in the wake of the financial crisis in 2008, in a welfare state actively trying to decrease health inequities among citizens [11]. The national unemployment rate is currently higher than before the crisis (7.9% of the working age population compared to 5–6%), men have higher unemployment than women, and youths (15–24 years) have the most difficulties finding work (22.9% in unemployment) [12, 13]. Sweden is characterised as a worker-friendly country with universalism, solidarity and de-commodification of social and economic equality between its citizens [14]. It is a welfare model constructed to reduce inequities between classes through benefits such as social insurance, parental leave benefits, state-funded pension, free higher education tuition, free health care, and day-care, child allowance and school lunches for all children. It is a costly burden of social services and transfers funded by public taxes, and thus with high demands on all citizens to contribute by working full-time and paying taxes [14]. For those in unemployment, i.e. involuntary lack of paid work, the Swedish state has a de-commodification system for income loss and self-worth by cash programmes to help unemployed to find ways back into work (by passive and active labour market matching) [9]. However, during the last few decades the Swedish welfare state has been undergoing major changes, in accordance with macroeconomic change in the world economy, leading to downsizing in income protection and other welfare benefits [11]. In fact, when the cost of social benefits increased during the economic crises and unemployment rates grew in the 1990s and in 2008, cuts were made in the state-funded unemployment benefits [9, 10], making those in unemployment a particularly economic vulnerable group. Recent reports have described an alarming development with increasing socioeconomic inequalities, child poverty and cuts in the state-funded social security systems [15–18]. Even though Sweden still has generous welfare policies and strong focus on equity in health [11], the system has not been able to reduce the health inequities between social classes in society [19].

Unemployment and social determinants of health

In this study unemployment is conceptualised as a social and material exclusion from the normative wage working society [20]. It is a stressful life event regardless of age and may affect health through embodied stress and emotional strain [21]. Previous research has shown a clear relationship between unemployment and concurrent health problems, both with mental health illness such as general poor mental health, anxiety and depression, and with physical health outcomes such as high blood pressure and cortisol levels, heart diseases and mortality [4, 5, 22]. However, studies on health inequalities have mainly focused on social determinants such as income, social status, gender and place of residence [23], overlooking work and unemployment as one of the most important social determinants of health inequality [20]. A direct consequence of unemployment can be poverty and economic deprivation, where income loss is likely to restrict life choices, the ability to plan ahead and the feeling of control [24]. Financial strain and low cash margin also tend to increase stress and impair mental health among unemployed [25]. Economic deprivation may also act as a moderating factor through socioeconomic circumstances of low education, working conditions of manual work and low income, leading to higher risk of unemployment and psychological distress. Beyond the income loss, unemployment is also a deprivation of psychosocial aspects of work, such as feeling of belonging, self-identity, social interaction and integration in society, which are important protective factors for mental health [26]. People in employment spend a lot of time at the workplace and it becomes an essential part of the social activities of everyday life. Hence, unemployed people tend to be less active and more socially isolated. It may partly be due to the lack of income, and thus the ability to be a consumer in a society organised around consumption of goods and services. It may also be related to loss of status and shame at not being a part of the working society [27]. However, access to social buffers can dampen the loss of work-related social isolation by increasing the individual’s ability to cope with the emotional strain of unemployment, such as lack of social interaction with colleagues. This might be the social network of family and friends, providing someone to talk to or practical support [28, 29]. Alternative role identities beyond the working identity, such as social participation, can also act as a social buffer for ill-health related to unemployment [27]. Participation in social movements and public debate may lead to better social, mental and physical health regardless of employment position [30], and thus challenge the normative idea of work as a critical source of individual growth, self-fulfilment, social recognition and status [27].

Another important aspect relates to how individual health status can be shaped by the macro- and meso-level systems of society, such as the organisation of work and the labour market, but also the availability of systems such as income protection and access to health care [31, 32]. Active labour market policies have been shown to moderate the impact of health hazards during unemployment [33–35], to facilitate the ways back to re-employment [36] and reduce the worries and fear regarding risk of unemployment among those in employment [37]. In fact, psychological and economic research has shown that job insecurities and fear of unemployment affect well-being and life satisfaction [37], implying that it is not just the actual implications of generous social policies that can reduce ill-health but also that subjective trust and confidence in institutionalised social benefits might have health-buffering implications. In contrast, those in highest need of health care and social security may also be those at highest risk of unemployment, with low resources for individual economic and social protection, leading to a decline in trust in the ability of welfare systems to care for the citizens. Thus, the social insurance may modify the health implications of unemployment, while preserving the inequalities in health in the population [38].

Unemployment and ill-health in different phases of life

Sociological and epidemiological life-course researchers recognise three main stages of life corresponding to the position on the labour market, social and psychological development and the social and material circumstances characterising each life phase [39, 40]. The first phase is young adulthood (late teens to mid-twenties). It is typically illustrated as the stage in life when young people are expected to enter adult society by normative age-specific events, such as leaving school, entering the workforce, leaving the parental home and engaging in long-term relationships. Even though none of these social roles transitions is required for the achievement of adult status, together they mark the adult stage of the life course development [41] and unemployment may delay some of these age-normative events. This is sometimes referred to as ‘delayed adulthood’ [41], with implications for the health status of young unemployed people. On the other hand, youths tend to have a relatively low financial burdens compared to people during other stages of life and they often share the experience of instability in labour market attachment with their peers. Thus, the stigma related to unemployment may be lower among youths. The second phase is adulthood (late twenties to 40), often illustrated as the ‘settling down’ phase which often involves an increased investment in work, family and friends [40]. It is the time in life between the intensive phase of becoming an adult and before the more critical health problems accruing in mid-life, thus a phase with typically good physical health [40] but nevertheless high lifestyle stress. This is a time in life when many become parents, with higher demands for financial stability and when the social norm of employment is high, i.e. the social expectation of wage work is closely related to the feeling of belonging, identity and taking part in society. Unemployment during this phase may affect the emotional strain in relationships with family and friends, illustrated in the higher risk of divorce among unemployed [42]. The third and last phase in the labour market is mid-life (40 to 65 years), sometimes referred to as the time of the ‘empty nest’; children move from the parental home and the financial burden is lower. Most people have a relatively stable position in the labour market during this period of life. However, becoming unemployed during this phase is troubling due to age discrimination in the labour market. Becoming re-employed has been shown to be difficult, due to ideas about this group having difficulties adapting to new technology and changing society and cultures [43, 44], and thus with implications for health status [45]. This is also a time in life when age-related health problems start to become more frequent and when unemployment may be an even higher contributor to health inequalities.

In sum, previous studies within the field are limited, typically focusing on the relationship between unemployment and health, rather than on which contributing factors can explain the differences in health [46]. There are also reasons to imagine that social and economic circumstances may differ across age groups [47], resulting in different factors influencing the health inequalities between employed and unemployed. The aim of the present study was to investigate the mental health inequalities between employed and unemployed across three age groups of youth, adulthood and mid-life in northern Sweden and to identify to what extent social determinants could explain the mental health gap between the two groups.

Sample and procedure

The current study was conducted on the Health on Equal Terms survey. This is a Swedish public health survey conducted every four years since 2004. The survey is administered by the Public Health Agency of Sweden in collaboration with the four northern county councils and Statistics Sweden [48]. The survey covers areas such as health and well-being, contact with health care, living habits, work and occupation, and social relationships. Register data from Statistics Sweden were also collected for all participants about areas such as education, income and civil status. In the four northern counties of Sweden (Västernorrland, Jämtland, Västerbotten and Norrbotten) the last data collection was conducted in 2014, comprising 708,641 inhabitants in the age 16 and 84. From this population, a randomised sample was selected, stratified by sex, age and municipality. The response rate was approximately 50% (n = 25,667). In the present study the effective sample was n = 10,407 individuals (of whom 7 .7% were unemployed), due to exclusion of people over age 65 (n = 8888), and people outside the labour market, such as students, on sick leave, on parental leave, in labour market measures (n = 4001). Individuals with missing values in some of the variables also became missing in the final analysis (n = 2371), most frequent in financial strain, cash margin and support. However, no systematic pattern was found in the missing values.

Measurements

Psychological distress, measured by the twelve-item version of the General Health Questionnaire (GHQ-12), was used as mental health outcome. It is a widely used self-administered screening instrument developed for non-psychotic mental illness, typically depressiveness and anxiety [49]. The participants were asked if they had experienced different symptoms or behaviour during the last few weeks, such as being able to concentrate and enjoying day-to-day activities, feeling unhappy, depressed, and capable of making decisions. Each item assesses the severity of the symptoms with a four-point scale as ‘less than usual’, ‘no more than usual’, ‘rather more than usual’ or ‘much more than usual’. The items were recoded into more or less severity and then summed up into an index with the range 0–12, and dichotomised by three or more symptoms, commonly used in surveys to indicate psychological distress [49].

The variable unemployment was self-reported, capturing the subjective experience of involuntary lack of paid work. Participants’ current labour market position was dichotomised into unemployed and employed (including self-employed), excluding those in labour market measures, students, those on parental leave and different forms of sick listing in order to make as homogenous comparison groups as possible. Different dichotomising of labour market position was tested, such as unemployed or in labour market measure and outside the labour market, with similar results.

Age was categorised into three groups: youth between age 16 and 25, adults between age 26 and 39 and middle-aged between age 40 and 65. Age groups were categorised in accordance with a sociological and epidemiological life course framework of different stages in life (see above).

Determinants of health inequities

Determinants of health inequalities were divided into four different domains: socioeconomic status, economic resources, social network and trust in welfare institutions. Because the health consequences of unemployment runs across all levels of the society, the variables education, civil status and occupational class measured socioeconomic status. Highest level of education was based on register data collected by Statistics Sweden 2013, and coded into high and low level of education. Low education included compulsory education and 2 years’ upper-secondary education without eligibility for higher education, and high education as 3 years of secondary education until PhD degree. Participants’ civil status was based on both register and questionnaire data, to capture the common living arrangement of non-married cohabitation in Sweden. Statistics Sweden’s register data of civil status registered in 2014 were used as married, unmarried, divorced and widow, while the self-reported questions were used for cohabiting living arrangement. The civil status variables were coded into married/cohabiting and unmarried/not cohabiting (including divorced and widow/widower). Occupational class was, unlike the self-reported unemployment, based on Statistics Sweden’s socioeconomic classification last registered employment [50] and coded into blue-collar worker (unskilled manual workers and skilled manual workers) and white-collar workers (assistant non-manual, intermediate non-manual, self-employed and other).

Economic resources were measured by three variables: income, cash margin and financial strain. Income was based on the individual disposable income of the register-based debt and assessed (such as earnings from employment and capital, housing and child benefits, student loan, repayments of student loan and alimony) and coded into five quintiles. Participants’ cash margin was captured by the question ‘Would you be able to get hold of 15 000 SEK in one week for an unforeseen situation?’ with the response option as yes or no. Financial strain was captured by the question ‘During the past 12 months, have you had difficulties managing your ongoing expenses, such as food, rent and bills?’ which was dichotomised into yes or no.

The social dimension was measured by the subjective feeling of access to social network, divided into practical, social support and trust in others and social participation. Practical support was measured by the ability to get help in case of practical problems or illness and dichotomised into ‘always or most of the time’ and ‘never or not for the most parts’. Social support was measured by the question ‘Do you have someone to share your thoughts and inner feelings with?’, while trust in others was measured by the question ‘Do you think, in general, that one can trust most people?’ both with the response options yes or no. Social participation in leisure time activities was measured by the matrix question ‘Have you participated in some of the following activities during the last 12 months?’ with fourteen different options of activities. Two different variables of alternative roles were created: socially active and participation in sports events. The first one was measured by participation in a public event, such as night club or dancing, larger family gatherings and/or private party. The second one asked whether or not they had participated in public sport events. Both variables were dichotomised into active or none active. The remaining alternative roles were also tested, such as participation in leisure time study circle, labour market unions or other types of non-governmental organisations, public debates as well as cultural events such as theatre, movies and art exhibitions, but they were not included in the final analysis since they did not contribute to the health gap.

Two variables were constructed to capture the general belief in the welfare institutions. The matrix question was worded as ‘Do you trust the following institutions/politicians in society?’, and asked to rate on a five-point scale from ‘very high’, ‘rather high’, ‘rather low’, ‘very low’ and ‘no opinion’ for twelve different institutions. The first constructed variable, confidence in unemployment benefits, consists of attitudes towards the state-administered unemployment office and the income benefit. The variables were added and dichotomised into ‘high or very high trust’ and ‘low or very low trust’. Those with no opinion were coded as missing. The second variable, confidence in health care, was elicited and coded in the same way as the variable above.

Analysis

Frequencies of the characteristics for unemployed and employed were calculated for each determinant of health. Pearson’s chi-square was used to assess statistical differences between unemployed and employed within each of the age groups.

To address the aim of the study, a Blinder-Oaxaca decomposition analysis for non-linear regression models was applied [51], decomposing the mental health gap between unemployed and employed into three age groups (youths aged 16 to 25, adults 26 to 39 and middle-aged 40 to 65). The analysis was conducted in two steps and performed stratified for each age group and for genders. First, the difference in mental health was estimated, between the disadvantaged group (the unemployed) and the non-disadvantaged group (employed), presented as log odds. Then a decomposition was conducted to assess the magnitude of the explained and unexplained components along with a detailed decomposition to display each predictor contribution of the explained and unexplained association with the health inequalities. The explained component reflects the contribution of each social determinant to the difference in health gap between the groups, and is interpreted as the difference of characteristics between unemployed and employed that could explain some part of the total health gap [52]. The unexplained component captures the part of the health gap that remains unexplained. It can be interpreted as a discrimination or an unequal treatment of the unemployed against the employed (with the assumption that all determinants of health are included in the analysis) or uncertainty due to residuals [53]. Estimations were reported as log odds along with the share of the relative contribution of each determinant of health (presented as percentage of the total explained contribution) to the health inequalities. Calculations of confidence interval were presented as the significance at 5% level [52]. This approach has previously been applied within fields such as gender income-inequalities, health inequalities and socioeconomic gaps in health [53–55].

All the analyses were performed in Stata 13. Estimations were obtained with the Oaxaca command specifying the logit and pooled options. The logit option is the non-linear decomposition for binary dependent variables, providing a way to estimate both aggregated and detailed decompositions [51] while the pooled option applied a pooled model (as an additional variable) of both groups’ coefficients as the reference coefficient [56]. Independent variables, such as trust in school, social service, parliament, local politicians and trade unions, without contribution to the health gap (coefficients< 0.00) were excluded from the analysis. Ordinary Blinder-Oaxaca analysis and Fairlie decomposition [57] were also carried out, giving similar results.

Health gap between employed and unemployed

Decomposing the health gap

The second step was to decompose the health gap, reported for each age group in Tables 2, 3, and 4. In the age group of 16 to 25 (Table 2), the explained proportion of the health inequality were 43 and 51% among men and women, respectively, attributed to the social determinants of health between unemployed and employed. Among men, the single most contributing factor was cash margin, which explained 43% of the health gap (p < 0.05). In addition, experience of trust in others and trust in health care (15–16%) showed to be contributing factors, but not significant, among young men. Among young women, none of the single predictors showed a statistical significant contribution. However, cash margin (15%), financial strain (12%) and practical support (11%) were the most relevant.

In adulthood (26 to 39 years old), the social determinants of health explained 92% of the health gap among men and 42% of the health gap among women (Table 3). Financial strain showed to be the single most contributing factors for both sexes (25 and 27% respectively, p < 0.05). Among adult men, the significant contributors to the health gap were income (21%) and trust in others (11%), while the corresponding factors among women were cash margin (16%) trust in others (10%), and social support (8%). A negative contributing factor among adult women was occupational class (− 27%, p < 0.05). This implies that among women in the age group of 26 to 39 years old, employed were more often white-collar workers than unemployed, and this, in combination with white-collar work being associated to poor mental health among unemployed adult women leading to increase, rather than decrease, in health inequality between the groups.

The social determinants of health in middle age (age 40 to 65) explained 60 and 65% of the health inequality between employment status among men and women, respectively (Table 4). Similar to previous age groups, the economic factors showed to be the main contributing factors; cash margin (19%) and financial strain (15%) among men, while financial strain (25%) among women (p < 0.05). Other key significant determinants of health were social support (15%), trust in others (13%) and practical support (9%) among men, and trust in others (19%), trust in health care (12%), participation in sport events (7%) among women. In middle age, education showed to be a factor increasing the health gap between employed and unemployed (− 7% among men, − 13% among women, p < 0.05). Among men, occupational class also showed to be negatively contributing to the health inequality (− 15%, p < 0.05).

Overall, socioeconomic background did not have a key role in explaining the health inequality. Having low education, not cohabiting and blue-collar occupational class were characteristics more common among those in unemployment, emphasising the uneven distribution of socioeconomic resources between employments.

Methodological considerations

The limitations of the study need to be addressed. First, with a response rate of 50%, the non-response could be an issue regarding representatively to a larger population. The sensitivity analysis performed by Statistics Sweden showed that the majority of dropouts were registered as not accessible while only a small part were mail returns, declining participation, questionnaire problems or unable to contact. Regarding labour market position, the dropout was only 1.4%, although we were unable to know whether unemployment was a reason for not participating in the survey. Internal dropout due to missing values in one or several explanatory variables resulted in a relatively large share of missing individuals in the analysis, particularly variables such as financial strain, cash margin and support. However, as mentioned in the methods section, no systematic pattern was found in the missing values. Moreover, the cross-sectional study design could not account for previous health status. The health selection may be troubling given the two-way causal direction of unemployment and health [65]. A further limitation is that the study design also limited our ability to account for the complexity of different types of unemployment and unemployment durations, and thus, the actual social and economic change. Earlier research shows that the effect of unemployment on health varies depending on timing, duration and characteristics of the unemployment situation, factors that we did not have the possibility to measure [4, 66]. Lastly, our model hypothesised that unemployed as compared to employed were at higher risk of social and economic exclusion, which in turn is damaging to health. However, unemployment, particularly during shorter periods of time, may also be a health-improving life event, due to poor work environment (such as stress and conflicts), less conflicts related to the work-home balance and fear of unemployment. This heterogeneity within the group of unemployed could not be taken into account due to the study design, which might result in smaller health gap estimations between the employed and unemployed. Different operationalisations of unemployment (including outside the labour market and labour market measure) and employment (including all excluded groups such as those in parental leave, students and sick-leave) were tested, with similar results supporting the validity of our results.