Introduction

Occupational health research has established solid evidence on the impact of work stress on health, mainly based on the application of theoretical models in epidemiological cohort studies. Work related stress has been found, for instance, to be associated with increased risks of cardiovascular disease [1–3], affective disorders [4–6], and a range of other health risks [7–9]. Importantly in many, but not all [10,11] studies, measures of work stress followed a social gradient, with higher levels of work stress among workers in more disadvantaged socioeconomic positions (SEP) [12–14]. This finding is relevant in the context of explaining social inequalities in health among working populations, because work-related stress may be an important intermediate factor linking low SEP with poor health. Most existing studies supporting this ‘mediation’ hypothesis, show that associations between SEP and health are generally weaker once work stress is considered in multivariate analyses [15–17].

Against this background it is important to explore socioeconomic differences in work stress in more detail, specifically in a cross-country perspective. This perspective offers the opportunity to compare associations between SEP and work stress in a variety of countries. In addition it is possible to study whether there is a link between the extent to which specific national policy regulations are implemented and the magnitude of socioeconomic inequalities in stressful work. This latter question is of particular interest in view of the explicit goal of national welfare policies to improve living and working conditions of their socially disadvantaged population groups. In this paper we set out to study this research question in the frame of a cross-national survey of older employees in 16 European countries. More specifically, we study differences in work stress between different educational groups. Work stress was operationalised in terms of two theoretical models, the demand-control model [18], and the effort-reward imbalance model [19]. While work stress in the first model is a result of a distinct job task profile, i.e. high demand in combination with low control (‘job strain’), its definition in the second model is based on the work contract: It results from high efforts spent at work not being reciprocated by appropriate rewards in terms of salary, job promotion or security, and esteem. Before describing the study design and the methods we explain the role of labour market policies and education in this context.

There is increasing evidence that the overall level of work stress in a country is related to its wider political context [20]. Existing welfare state typologies constitute an important conceptual background at this point, but recent empirical research has followed a more comprehensive approach that uses quantifiable indicators of labour market policies. For example, active labour market policies (ALMP, measured as % of GDP) were related to favourable working conditions, in particular policies that promote further qualification among adult workers and that invest in supported employment and rehabilitation services [21,22]. Furthermore, recent findings suggest that the strength of associations between work stress and employees’ mental health is generally smaller in countries with extended unemployment protections, in particular, the unemployment benefit levels [23]. These findings suggest that two types of labour market policies are of special interest, protective and integrative policies. Integrative policies aim to promote return to work in case of unemployment or support job maintenance for those in precarious employment (e.g. through ALMP or workplace training). Protective policies refer to social provision through compensation of income loss following unemployment or precarious employment (e.g. level of unemployment benefit). In terms of existing country-variations, European countries have generally experienced a substantial decrease of protective policies and an expansion of integration policies. Although the decline of protective policies has been very pronounced in case of Scandinavian countries (in particular Sweden) and continental European countries, these countries still have relatively extensive protective policies in particular compared to Anglo-Saxon countries. Similarly, albeit the expansion of integrative measures was observed in most European countries, Scandinavian countries (in particular Denmark) still have highest spending levels of ALMP followed by continental countries, and lowest levels in Anglo-Saxon and Eastern European countries [24].

Yet, an open question is whether these two types of labour market policies also have an impact on the magnitude of the socioeconomic differences of stressful work at a national level. For example, protective policies offering financial compensation in case of job loss (e.g. replacement rate) may reduce socioeconomic differences in work stress, because they offer financial security employees can rely on [25]. In contrast, as long as individuals' standard of living entirely depends on market performance, it is likely that employees are forced to accept any kind of job, even if this means that they have to work under stressful conditions. In a similar way, integrative policies supporting those who experience difficulties in entering or staying in the labour market may be important as well. For example, ALMP and lifelong learning opportunities may enhance the skills of employees with lower educational levels and promote their (re-) integration into the labour market [24]. In this study we test if national policies, in terms of protective and integrative policies, reduce socio-economic differences in work stress.

In the following analyses, we use the respondents' highest educational degree, as defined by the International Standard Classification of Education (ISCED, see Methods for details). This measure explicitly considers national variations in educational systems, therefore allowing for cross-country comparisons. Educational qualification is an important determinant of adult life chances as it provides resources and capabilities required for successful labour market integration. This includes individuals' occupational position in working life and labour market disadvantage [26]. In addition, people with lower levels of education may profit specifically from the labour market policies described above, and thus, this indicator appears appropriate for our analyses. For example, by improving the level of qualification through further education, people with lower educational classification may profit in particular. Education has the advantage of being relatively stable throughout the life course [27], and of being closely associated with a range of health conditions [28,29]. Moreover, educational qualification may moderate the effects of distinct labour market policies on work and employment [26].

Methods

Measures

Work stress.

Work stress was assessed by a short battery of items derived from (a) the Job Content Questionnaire measuring the demand-control model [35] and (b) from the effort-reward imbalance model questionnaire [36]. Given the constraints of a multi-disciplinary approach the inclusion of the full questionnaires was not possible in both surveys. Therefore abbreviated measures were used. The items were selected on the basis of factor loadings on respective original scales. In case of the demand-control model, the measurement was restricted to the control dimension based on evidence that the predictive power of control exceeded the power of demand in some studies [37]. Control was measured by the sum score of two Likert-scale items. The response categories of these items range from 1 to 4, with higher scores indicating lower control at work. The sum score of these two items varies between 2 and 8. To measure effort-reward imbalance, 2 items measuring ‘effort’ and 5 items assessing ‘reward’ at work were used. 'Effort-reward imbalance' was then calculated by dividing the sum score of the 'effort' items (nominator) through the sum score of the 'reward' items (adjusted for number of items; denominator). This ratio of the effort and reward items results in a sum score ranging from 0.25 to 4 where higher values are related to higher levels of work stress.

Education.

Our measure of education is based on respondents' highest educational degree that we regrouped according to the International Standard Classification of Education (ISCED). This classification was developed to improve the comparability of educational attainment between different countries. Following previous procedures [28], levels of classifications were regrouped into three categories: low (level 0–2; pre-primary, primary, or lower secondary education), medium (level 3; secondary or post-secondary education), and high educational attainment (level 4–6; first and second stage of tertiary education). This is justified by the fact that the selected categories represent significant levels of educational attainments with important consequences for occupational positions in working life [38].

Additional measures.

We included age, sex, employment status (self-employed vs. employed) and work time (full-time (>35 hours per week) versus part-time) as additional variables in the analyses.

Policy indicators.

Rather than using existing welfare state typologies for our analyses, we use four policy indicators of national labour market policies, each provided from official sources on a coherent and comparable basis. As described in the introduction, these indicators cover two relevant dimensions of labour market policies (two for each dimensions), that is, ‘integrative’ and ‘protective’ labour market policies.

In case of ‘protective’ policies, the first measure is the so-called ‘replacement rate’. It describes the expected net income in the period directly after job loss as a percentage of the net income before job loss. Second, we use one indicator provided by the OECD that summarizes the amount of a country’s labour market expenditures into ‘passive labour market policies’ (PLMP), expressed as percentage of GDP [39]. More specifically, PLMP is divided into two sorts of expenditures, first, in investments that aim to compensate individuals temporarily for loss of wage or salary (e.g. unemployment benefit), and second, expenditures that compensate and promote premature retirement of older workers with disadvantages on the labour market.

Turning to integrative policies, we use one indicator measuring the extent of lifelong learning possibilities in a country, and a second indicator measuring the amount of investments into ALMP. In case of lifelong learning, the indicator refers to older men and women (55 to 64) who stated that they received education or training in the last 12 months (in per cent). This information is provided by EUROSTAT and was collected in the ‘Adult Education Survey’ [40]. In case of ALMP, information is again expressed as percentage of GDP and comprises various policy measures of ALMP (usually classified into 6 different types of actions), in particular interventions that aim to promote labour market integration for disadvantaged groups. These four policy indicators are described in detail in Table 1, and we present country values in the supporting information (S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Policy indicators.

https://doi.org/10.1371/journal.pone.0121573.t001

Statistical analyses

Following a basic description of the sample (Table 2), we present two sets of analyses. The first set compares levels of work stress by educational qualification for each country. To do so, Fig. 1 presents average levels of work stress (for both work stress models) for each level of education, and we estimate multivariate linear regression models using work stress as dependent variable. More specifically, we test associations between education (broken down into dummy variables) and the two work stress measures (adjusted for sex, age-categories, employment status and work time) for each country separately. These findings are presented in Table 3, where we list the estimated coefficients for education using high education as reference category. In addition, to explore how educational differences are related to labour market policies, we study if effect sizes (as calculated within the regression models) are linked to the indicators of labour market policies, and we summarize findings in Figs. 2 and 3. Furthermore, using a subsample of the dataset with available information on occupational position in SHARE, we conduct sensitivity analyses and explore the role of occupational position within the association between education and work stress (results not presented in details). Thereby, we calculate stratified analyses (i.e. the association between education and work stress for each group of occupational position), and also estimate the association between education and work stress adjusting for occupational position. As an indicator of occupational position, we use the occupational skill level, which represents the broad hierarchical structure of the International Standard Classification of Occupation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Average levels of work stress (mean score) by education and country.

Note. Results are based on weighted data.

https://doi.org/10.1371/journal.pone.0121573.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Educational differences in work stress (low vs. high education) and lifelong learning (% of older worker in further education).

Note. Mean differences are adjusted for age, sex, self-employment and work time (based on Table 2).

https://doi.org/10.1371/journal.pone.0121573.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Educational differences in work stress (low vs. high education) and ALMP (expenditure into active labour market programmes).

Note. Mean differences are adjusted for age, sex, self-employment and work time (based on Table 2). Expenditures into active labour market policies (ALMP) are based on % of GDP (weighted by unemployment rate).

https://doi.org/10.1371/journal.pone.0121573.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Sample description (N = 13695).

https://doi.org/10.1371/journal.pone.0121573.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Associations between education and work stress scores by country: Results of linear regression models (unstandardized regression coefficients and p-values).

https://doi.org/10.1371/journal.pone.0121573.t003

In a second set of analyses, we combine all countries (pooled dataset), and we study the effects of the four policy indicators in more detail, using linear multilevel regression models (random intercept only). Given the multilevel structure of the data, we estimate a multilevel model with individuals (level 1) nested within countries (level 2). Using multilevel modelling allows for accurate adjustment for country affiliation, because the intercept is allowed to vary across countries. This is important for our analyses, because of previously reported country-variations of work stress [21,41]. In addition, we conducted likelihood ratio tests to compare the multilevel models to conventional linear regression models (with country dummies), and these tests revealed better model fits in all cases. In sum we estimate two models for each policy indicator (PI). In model 1 the association of the policy indicator with work stress is estimated, adjusting for all individual characteristics (sex, age-categories, employment status, work time and education). This procedure accounts for potential effects due to differential population composition across countries.

In model 2 we additionally include two cross-level interactions, one for medium education (denoted medium education * PI) and another one for low education (low education * PI).

With these two interaction terms we seek to answer our core research question, i.e. whether there are significant differences in the effect size of the policy indicator on work stress between the medium education group and the low education group respectively as compared to the high education group (reference category). As a formal test of significance, we also apply a likelihood ratio test (LR Test) that indicates whether the two interaction terms add explanatory power to the one displayed in model 1. The results of these analyses are displayed in Table 4, where we present the effects for each policy indicator and additionally the interactions with education. Given a rather complicated interpretation of the findings, we visualise them in Fig. 4. In this Figure, levels of work stress are predicted by educational category at different levels of policy indicators.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Predicted levels of work stress by education at different levels of policy indicators.

Note. Expenditures into active (ALMP) and passive labour market policies (PLMP) are weighted by unemployment rate. Results are based on Table 4, model 2.

https://doi.org/10.1371/journal.pone.0121573.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Association between policy indicators and work stress (model 1) and interactions between education and policy indicators (model 2): Results of random intercept linear multilevel regressions. Unstandardized regression coefficients (p-values).

https://doi.org/10.1371/journal.pone.0121573.t004

Results

Discussion

In this contribution differences in work stress in relation to educational attainment among older employees (50 to 64 years) were studied in a cross-country perspective, using data from two European surveys covering 16 countries (SHARE and ELSA, N = 13695). Three findings were of particular interest. First, we confirmed a rather robust social gradient of stressful work in terms of low control and effort-reward imbalance, where those with lower educational qualification had higher levels of stressful work. Second, in countries with well-developed labour market policies average levels of work stress were generally lower than in countries with less developed policies. Third, the steepness of educational differences in stressful work varied according to the extent of policy development such that differences tended to be smaller if national policies were well developed. This pattern was similar for both models of work stress. The associations between policy indicators and educational differences in stressful work were more consistent in case of ‘integrative’ vs. ‘protective’ policies.

While the first result corroborates previous findings it adds a new element as we use education instead of occupational position to investigate the social gradient of work stress [12,13,27]. As shown in previous studies, education is a core marker of SEP and has high significance for unequal life chances including occupational careers [26]. Along these lines, additional sensitivity analyses also explored the role of occupation in more details (as described in the methods section). In these analyses we found support of an association between education and work stress within each occupational position. Also, we found that effects of education on work stress are generally reduced, but remain significant after adjustment for occupational position in multivariate models. These findings lend important support in favor for an indirect effect of education through occupational position, but it also suggests that the effect of education is only partly due to occupational position.

Our second finding is in accordance with previously documented associations of average quality of work of a country’s working population with the extent to which distinct indicators of national social and labour market policies are implemented [21,22]. However, our analyses are based on two conceptually distinct sets of policy measures, labeled ‘integrative’ and ‘protective’ policies. This approach is in line with recent developments within welfare state literature [41,42,43]. It offers the opportunity of exploring whether the former, more active, integrative policies are more responsive to the needs of socially disadvantaged groups than is the case for ‘protective’ policies [42]. This latter assumption has been tested with regard to educational differences in the amount of stressful work where our third finding demonstrates smaller social gradients of stressful work under conditions of well developed ‘integrative’ labour market policies. One possible explanation of this finding is that ‘integrative’ labour market policies enable individuals to find better jobs and therefore improve working conditions [44]. This was especially evident in case of the indicator ‘lifelong learning’. To our knowledge, this hypothesis has not been tested so far. Our findings suggest that efforts to further qualify workers even in later phases of their career and help to successfully integrate them into the labour market may buffer the negative consequences of a low formal education gained in adolescence or early adulthood. Yet, while less important in reducing educational differences, our findings nevertheless show that ‘protective’ policies matter as well, in particular in reducing the overall level of work related stress. Similarly a recent study also showed that ‘protective’ policies may mitigate the effect of work stress on health [23].

This study has several limitations. First, our measurement of work stress was restricted to short assessments of single components of two established theoretical models, demand- control and effort-reward imbalance. ELSA and SHARE, the two studies used in our analyses, only include the control scale of the demand control model. It is therefore not possible to define job strain, the combination of high demands and low control, in our analyses. A stronger case would have been provided if the full original questionnaires were applied and if additional aspects of stressful work were incorporated [45,46]. Second, our choice of indicators of national labour and social policies can be challenged both in terms of coverage of relevant policy domains and in terms of accuracy of reflecting the specific quality of national policy developments. Third, although our study includes a variety of countries with different historical and political background, a larger number of countries would have strengthened the case and would correspond more appropriately to the methodological requirements of applying multilevel statistical analysis. Fourth, the cross-sectional design of the study prevents any conclusion concerning the direction of observed statistical effects. For instance, we do not know whether the implementation of a productive national policy measure is followed by a subsequent reduction in work stress or in an attenuation of its social gradient. The study population consists of of men and women aged 50 to 64. This age group is very sensitive to the healthy worker effect which could have affected the results [47], and therefore, future studies may replicate our findings to other age groups as well. Finally, while the overall sample size of this study is quite large, it is rather low in some of the countries included. On the one side, it is probable that this may affect the composition of our sample and thus the overall level of work stress therefore calling for further validation. Yet, on the other side, it seems less probable that low response rates additionally affect the associations under study and its strength.

Despite these limitations the study has several strengths. It is based on two prominent, internationally established models of stressful work. It identifies national social and labour policies at the level of quantified, cross-nationally comparable indicators. Moreover, our choice of indicators reflects a recent conceptual shift in welfare state research from ‘protective’ to ‘integrative’ policies activating and integrating disadvantaged groups of the labour market [42]. In this context we tested an innovative research hypothesis and found preliminary empirical support. Finally, the availability of a large comparative data set from two European studies enabled us to study variations of national policies and their relevance for average work stress levels in these countries and their social gradient. Previous research found preliminary evidence that distinct labour policies can mitigate the effects of stressful work stress on mental health at the level of respective populations [23].

If the findings of this study represent robust trends in further analyses, they may have important policy implications. More specifically, our main result suggests that distinct ‘integrative’ labour market policies exert more beneficial effects (in terms of lowering average levels of work stress) among socially disadvantaged groups than is the case for passive or ‘protective’ policies. This is eventually due to the fact that these policies target the needs of populations suffering from labour market disadvantage more effectively than other policies [20]. Therefore they may offer a promising entry point for policies that aim at reducing social inequalities in work and employment and their adverse effects on health and well-being [48].

In conclusion, this study illustrates that people with lower educational qualification are more likely to experience work-related stress in a range of European countries, and that specific national labour policies may contribute to a reduction of the social gradient of stressful work in respective countries. Given the negative consequences of work-related stress on mental and physical health, our findings can direct political attempts towards reducing social inequalities in the health of working populations.

Supporting Information

Acknowledgments

This paper uses data from SHARE wave 4 release 1.1.1, as of March 28th 2013 (DOI: 10.6103/SHARE.w4.111). The SHARE data collection has been primarily funded by the European Commission through the 5th Framework Programme (project QLK6-CT-2001-00360 in the thematic programme Quality of Life), through the 6th Framework Programme (projects SHARE-I3, RII-CT-2006-062193, COMPARE, CIT5- CT-2005-028857, and SHARELIFE, CIT4-CT-2006-028812) and through the 7th Framework Programme (SHARE-PREP, N° 211909, SHARE-LEAP, N° 227822 and SHARE M4, N° 261982). Additional funding from the U.S. National Institute on Aging (U01 AG09740-13S2, P01 AG005842, P01 AG08291, P30 AG12815, R21 AG025169, Y1-AG-4553-01, IAG BSR06-11 and OGHA 04–064) and the German Ministry of Education and Research as well as from various national sources is gratefully acknowledged (see www.share-project.org for a full list of funding institutions).

Data from ELSA were made available through the UK Data Archive (UKDA). ELSA was developed by a team of researchers based at the National Centre for Social Research, University College London and the Institute for Fiscal Studies. The data were collected by the National Centre for Social Research. The funding is provided by the National Institute of Aging in the United States, and a consortium of UK government departments coordinated by the Office for National Statistics. The developers and funders of ELSA and the Archive do not bear any responsibility for the analyses or interpretations presented here.