Background
Maintaining work ability among older employees is an important goal of national policies in rapidly ageing societies in Europe and beyond. Physical and mental health are key determinants of work ability [
1,
2]. In recent years, a growing impact of mental disorders, particularly depression, on work ability in terms of sick leave and disability was documented in several countries [
3,
4]. In addition to established risk factors of depression [
5], exposure to a chronically stressful work environment increases the probability of developing depressive disorders, especially so if stressful work is measured by the demand-control or the effort-reward imbalance model [
6,
7]. The former concept posits that jobs defined by high demands in combination with low decision latitude or task control are stressful [
8], whereas the latter model is based on the notion of failed contractual reciprocity between efforts spent and rewards received at work, where rewards include money, promotion prospects, job security, and esteem [
9]. Taken together, both work stress models cover different, but equally relevant aspects of the workplace, where lack of control and reward frustration matter most. It is of theoretical and practical interest to know whether distinct national labour and social policies have an influence on work related health problems. Conceptually, these policies may be of importance in at least two ways: First, they may exert an influence on the prevalence of a stressful work environment. Second, they can modify the effect of stressful work on health and well-being [
10]. The first assumption has been supported previously and existing evidence (based on comparative European data) indicates that national active labour market policies (ALMP) are related to better working conditions, in particular those policies that promote further education and workplace training among older people [
11,
12]. However, the evidence for the second assumption is still limited [
13‐
15]. This limitation is partly due to a lack of cross-national studies, but also to the problem of how to define and measure relevant aspects of labour and social policies. While ALMP may promote psychosocial working conditions in general, one may assume that their impact on the health-adverse consequences may be different. For instance, aspects of employment protection may be more important in this case.
In this contribution we set out to overcome these limitations by studying the following research questions: (1) Do we observe significant associations of indicators of a stressful work environment with depressive symptoms across a variety of countries (12 European countries and the United States)? (2) Can we observe protective effects of distinct favourable national labour and social policies on the strength of this association (modification of effect of stressful work on depressive symptoms)?
With the first research question we address the paucity of available cross-national studies, in particular longitudinal investigations as the one reported here. With the second research question we propose to focus on core aspects of national labour and social policies which may represent protective resources in our context (see below). Rather than relying on established typologies of national welfare regimes [
16,
17] we maintain that the following more specific policy measures are better suited to reflect protective policy effects: (1) the amount of the state’s investments in active labour market policies (ALMP), (2) the degree of employment protection provided by the state, and (3) the degree of distributive justice as reflected, e.g. in the amount of income inequality.
The first measure may protect workers against the threat of being excluded from a core social role in adult life [
18], whereas the second measure protects those who are at risk of being excluded due to job loss [
19]. With the third measure a relevant collective sense of fairness is identified which may mitigate stressful experience of inequity at work [
20]. All three aspects of national labour and social policies exert their effect on wellbeing of employees by reducing the amount of threat experienced in case of job instability, forced retirement, major income shocks, degradation, or loss of job autonomy. In terms of theories of stressful experience, these threats to occupational status affect workers mental and physical health as they undermine essential feelings of continued control and reward at work [
21]. As the notions of control and reward are embedded in the two work stress models mentioned, our conceptual approach enables us to link distinct macro-structural contexts with individual-level experience of work and health (for measurement see Methods).
Impact of stressful work on depressive symptoms as well as potential protective resources provided by national labour and social policies are of particular relevance in view of an ageing workforce, as mentioned above. Therefore in this contribution we analyse our two research questions by referring to three longitudinal surveys of older employees (50 to 64 years) with similar study design and well comparable measurements of core variables.
Results
An overview of the longitudinal sample is presented in Table
1 together with the percentage of people developing increased depressive symptoms between 2004 and 2006. The total sample consist of slightly more men (3176) compared to women (2474) and the large majority (81%) was younger than 60 at baseline assessment. Variations of depressive symptoms are found for sex (higher among women), age (higher among younger people), socioeconomic position (higher among those in lower positions), work time (higher among those working part time), self-reported chronic conditions (higher among persons with stroke and diabetes) and functional limitations (higher among persons with at least one limitation in activities of daily living). Moreover, we observe a higher percentage of depressive symptoms among employees experiencing work-related stress compared to the remaining group.
Table 1
Description of measures and sample (N = 5650)
Sex | Male | 56.2 | 3176 | 6.2 | |
| Female | 43.8 | 2474 | 11.4 | .000 |
Age group (2004) | 50-54 years | 40.1 | 2263 | 9.6 | |
| 55-59 years | 40.8 | 2304 | 8.4 | |
| 60-64 years | 19.2 | 1083 | 6.2 | .004 |
Effort-reward Imbalance | Yes | 30.0 | 1696 | 11.1 | |
| No | 70.0 | 3954 | 7.3 | .000 |
Low work control | Yes | 29.8 | 1683 | 11.0 | |
| No | 70.2 | 3967 | 7.4 | .000 |
Income | Low | 31.0 | 1744 | 11.0 | |
| Medium | 34.1 | 1926 | 9.9 | |
| High | 35.0 | 1980 | 7.1 | .000 |
Education | Low | 26.8 | 1512 | 9.5 | |
| Medium | 38.0 | 2144 | 8.9 | |
| High | 35.3 | 1994 | 7.1 | .023 |
Employment status | Self-employed | 17.2 | 970 | 7.2 | |
| Employed | 82.8 | 4680 | 8.7 | .126 |
Work time | Part-time | 26.9 | 1517 | 9.6 | |
| Full-time | 73.2 | 4113 | 8.1 | .072 |
Heart disease | Yes | 6.5 | 367 | 9.3 | |
| No | 93.5 | 5283 | 8.4 | .567 |
High blood pressure | Yes | 25.0 | 1414 | 8.8 | |
| No | 75.0 | 4236 | 8.4 | .629 |
Stroke | Yes | 0.9 | 51 | 19.6 | |
| No | 99.1 | 5599 | 8.4 | .004 |
Diabetes | Yes | 5.0 | 281 | 11.0 | |
| No | 95.0 | 5369 | 8.3 | .112 |
≥1 Limitation in activities of daily living | Yes | 3.0 | 170 | 14.7 | |
| No | 67.0 | 5480 | 8.3 | .003 |
Country | Sweden | 11.7 | 660 | 9.6 | |
| Denmark | 6.4 | 360 | 9.7 | |
| Germany | 6.2 | 349 | 11.8 | |
| Netherlands | 7.2 | 407 | 6.4 | |
| Belgium | 8.3 | 468 | 11.3 | |
| France | 5.7 | 321 | 13.7 | |
| Switzerland | 3.5 | 195 | 7.2 | |
| Austria | 3.0 | 168 | 8.3 | |
| Italy | 3.6 | 204 | 15.7 | |
| Spain | 3.1 | 174 | 13.2 | |
| Greece | 7.0 | 397 | 1.8 | |
| England | 24.1 | 1360 | 5.8 | |
| USA | 10.4 | 587 | 8.0 | .000 |
Total | | | 5650 | 8.5 | |
These findings were confirmed in multivariate analyses, as presented in Table
2. The risk of experiencing newly manifested depressive symptoms at follow-up is significantly higher among women, in the younger age group, among persons with limitations of daily living, among persons who had a stroke and in the low income group. Importantly, in case of both models of work stress we see strong associations between work stress and incident depressive symptoms (Hypotheses 1).
Table 2
Associations of work stress with risk of elevated depressive symptoms at follow up: results of multilevel estimates (odds ratios and 95% confidence intervals)
Fixed parameters | | | |
Effort-reward Imbalance | Yes | 1.55 (1.27-1.89) | |
| No(Ref.) | | |
Low work control | Yes | | 1.46 (1.19-1.79) |
| No(Ref.) | | |
Sex | Female | 2.01 (1.63-2.48) | 2.00 (1.63-2.47) |
| Male (Ref.) | | |
Age group (2004) | 50-54 years | 1.49 (1.11-2.00) | 1.50 (1.11-2.01) |
| 55-59 years | 1.30 (0.96-1.75) | 1.31 (0.97-1.76) |
| 60-64 years (Ref.) | | |
Income | Low | 1.50 (1.15-1.95) | 1.48 (1.14-1.93) |
| Medium | 1.35 (1.08-1.69) | 1.35 (1.08-1.69) |
| High (Ref.) | | |
Education | Low | 1.13 (0.87-1.47) | 1.09 (0.84-1.43) |
| Medium | 1.11 (0.87-1.40) | 1.09 (0.86-1.38) |
| High (Ref.) | | |
Employment status | Employed | 1.04 (0.79-1.37) | 1.00 (0.76-1.31) |
| Self-employed (Ref.) | | |
Work time | Full-time | 0.99 (0.79-1.25) | 1.05 (0.84-1.32) |
| Part-time (Ref.) | | |
Heart disease | Yes | 1.29 (0.88-1.90) | 1.31 (0.89-1.92) |
| No (Ref.) | | |
High blood pressure | Yes | 1.05 (0.84-1.32) | 1.04 (0.83-1.30) |
| No (Ref.) | | |
Stroke | Yes | 2.73 (1.32-5.67) | 2.76 (1.33-5.72) |
| No (Ref.) | | |
Diabetes | Yes | 1.41 (0.94-2.11) | 1.41 (0.94-2.11) |
| No (Ref.) | | |
≥1 Limitation in activities of daily living | Yes | 1.88 (1.19-2.96) | 1.94 (1.23-3.05) |
| No (Ref.) | | |
Random parameters | | | |
Sigma u | | 0.46 (0.28-0.76) | 0.46 (0.28-0.76) |
Rho | | 0.06 | 0.06 |
Log likelihood | | −1558.22 | −1560.86 |
BIC | | 3263.31 | 3268.58 |
AIC | | 3150.44 | 3155.71 |
In addition, Table
2 displays significant, but small between-country variations of depressive symptoms, with an intra-class correlation (‘rho’) of 0.06. This indicated that only 6% of the total variations in depressive symptoms are related to differences between countries, and that – vice versa - most of the variations are related to differences between individuals.
To answer our second research question (effect-modification) Table
3 first displays the distribution of the macro indicators under study. Concerning investments into ALMP, rehabilitative services and income maintenance, we observe some convergence between the three indicators within single countries or among groups of countries with similar social and labour policies (e.g. high ALMP expenditures in Scandinavian countries, low expenditures in England and the USA). A similar pattern is observed with regard to union density, with the exception of England. While these findings partly correspond to existing welfare state regimes (e.g. ‘social-democratic’ versus ‘liberal’), [
16] less consistent associations are observed regarding ‘lifelong learning’ and ‘income inequality’. (e. g. relatively high income inequality in Germany or France, i.e. in countries not traditionally considered ‘liberal’ welfare states).
Table 3
Labour/Social policy indicators by country (rank order)
Sweden | 0.98 (2) | 0.22 (4) | 61 (1) | 1.29 (7) | 78.1 (1) | 0.24 (2) |
Denmark | 1.37 (1) | 0.30 (2) | 29 (5) | 1.94 (3) | 71.7 (2) | 0.23 (1) |
Germany | 0.84 (5) | 0.15 (5) | 28 (6) | 2.27 (1) | 22.2 (8) | 0.30 (5) |
Netherlands | 0.89 (3) | 0.56 (1) | 29 (5) | 2.09 (2) | 21.3 (9) | 0.27 (3) |
Belgium | 0.87 (4) | 0.12 (6) | 29 (5) | 1.56 (5) | 53.1 (3) | 0.27 (3) |
France | 0.72 (6) | 0.06 (8) | 16 (9) | 1.63 (4) | 7.8 (13) | 0.28 (4) |
Switzerland | 0.64 (7) | 0.25 (3) | 45 (2) | 1.03 (9) | 19.6 (10) | 0.27 (3) |
Austria | 0.44 (10) | 0.04 (9) | 25 (7) | 1.12 (8) | 34.4 (4) | 0.27 (3) |
Italy | 0.54 (9) | 0.00 (12) | 12 (10) | 0.64 (10) | 34.1 (5) | 0.35 (8) |
Spain | 0.63 (8) | 0.07 (7) | 17 (8) | 1.46 (6) | 15.5 (11) | 0.31 (6) |
Greece | 0.14 (11) | 0.00 (12) | 5 (11) | 0.40 (11) | 24.5 (7) | 0.31 (6) |
England | 0.06 (13) | 0.01 (11) | 37 (4) | 0.19 (13) | 29.4 (6) | 0.34 (7) |
USA | 0.11 (12) | 0.03 (10) | 40 (3) | 0.27 (12) | 12.0 (12) | 0.37 (9) |
In Figure
1 we can explore visually the different effect sizes of work stress on depressive symptoms by the six macro indicators – each dichotomized and labelled ‘protective’ or ‘non-protective’ as described in the method section. In case of low work control odds ratios are generally similar in the two groups. In contrast, we see that the effect sizes between effort-reward imbalance and depressive symptoms are generally stronger in a ‘non-protective’ policy context. Finally, to formally test whether effect size are significantly different between protective and non-protective contexts, we additionally calculated main effects of work stress on depressive symptoms together with interaction terms. Results are given in Table
4 showing estimated odds ratios for the main effects of work stress and interactions. The interaction term indicates whether (and to what extent) the association is significantly higher in the group of participants working in a ‘non-protective’ policy context. For instance, in case of effort-reward imbalance the interaction ‘Poor working conditions * Low ALMP’ means that the odds ratio in the ‘non-protective’ policy context (OR 2.13) is 1.77 higher compared to the ‘protective’ policy context (OR 1.20) [
40]. In case of the effort-reward imbalance model four out of six interaction terms are statistically significant and the main effects are non-significant in these latter cases. This indicates that health-adverse effects of work stress are restricted to ‘non-protective’ contexts and supports the notion of a modifying effect of distinct ‘protective’ policies on the strength of associations of work stress with depressive symptoms. The hypotheses were not supported for ‘union density’ and ‘lifelong learning’. In case of low control at work no significant interactions are observed, though, main effects are significant in five of six cases. In other words, effects of low control on depressive symptoms are observed in all cases, with no differences between protective and non-protective contexts.
Table 4
Modification of the effect of work stress on depressive symptoms by six labour/Social policy indicators
Poor working conditions (main effect) | 1.20 (0.92-1.57) | 1.37 (1.05-1.81) |
Poor working conditions * Low ALMP | 1.77 (1.19-2.64) | 1.14 (0.76-1.71) |
Poor working conditions (main effect) | 1.03 (0.77-1.39) | 1.32 (0.97-1.77) |
Poor working conditions * Low ALMP rehabilitation | 2.13 (1.43-3.18) | 1.22 (0.81-1.82) |
Poor working conditions (main effect) | 1.39 (1.11-1.75) | 1.50 (1.19-1.89) |
Poor working conditions * Low Lifelong learning | 1.55 (0.98-2.45) | 0.89 (0.56-1.43) |
Poor working conditions (main effect) | 1.21 (0.93-1.57) | 1.36 (1.03-1.78) |
Poor working conditions * Low unemployment benefit | 1.80 (1.20-2.68) | 1.18 (0.79-1.78) |
Poor working conditions (main effect) | 1.34 (1.03-1.74) | 1.48 (1.13-1.93) |
Poor working conditions * Low union density | 1.40 (0.94-2.09) | 0.97 (0.64-1.45) |
Poor working conditions (main effect) | 1.26 (0.97-1.63) | 1.35 (1.03-1.76) |
Poor working conditions * High income inequality | 1.67 (1.12-2.49) | 1.21 (0.81-1.82) |
Discussion
This study provides new evidence on two research questions. First, in a cross-national study including 5650 working men and women aged 50 to 64 years from 13 countries, we find significantly increased odds ratios of depressive symptoms at two-year follow-up among participants experiencing work-related stress in terms of effort-reward imbalance and low job control. Effects based on multilevel analysis are adjusted for relevant confounders.
Second, in case of the effort-reward imbalance model, four out of six indicators of a ‘non-protective’ labour or social policy at national level modify the effect of work stress on depressive symptoms, with significantly stronger odds ratios compared to those observed in ‘protective’ policy contexts. Importantly, to our knowledge, this is the first study demonstrating that distinct labour and social policies can buffer the effect of work stress on depressive symptoms. The health threatening stress response of effort-reward imbalance might be less pronounced if effort-reward imbalance is experienced in a ‘protective’ policy context. The reward dimension of the effort-reward imbalance concept consist of the subdimensions ‘esteem’, ‘salary’, ‘job promotion and job security’. One may assume for example that the threatening effect of job insecurity is less severe if protective policies (e.g. unemployment benefit) exist. In case of low control main effects on depressive symptoms were significant, with no differences by contexts (no significant interactions). One interpretation is that the effect of low control is rather independent of the contexts, and that – while important for the general level of low control (e.g. [
11,
12]) – national policies matter less once a person is exposed to adverse working conditions.
It may be premature to interpret these findings in the frame of a protective role of distinct welfare state policies on the working populations health and well being. Yet, it is important to note that the effort-reward imbalance model puts its focus on threats to the work role in terms of low wage or salary, low esteem or appreciation, poor promotion prospects and low job security in response to high efforts spent at work. These threats are particularly harmful if experienced under challenging macro-economic conditions of elevated levels of unemployment, forced job mobility and wage cuts [
41,
42].
This study has several limitations. First, in an effort to compare data obtained from 3 surveys of older employees covering 13 countries, the available measures of our core variables represent short versions of original scales and, in case of depressive symptoms, are restricted to standardized self-assessed questionnaires. Despite satisfying psychometric properties of these scales, improved measurements [
3] are recommended for future studies. For example the studies used in our analyses only include the control scale of the demand control model. Second, the availability of macro-structural labour and social policy indicators at national level was restricted to easily accessible online data bases provided by OECD. These indicators are still relatively crude, and quality of data may vary to some extent across countries. In addition, the number of countries included in this study is still relatively small although they represent a fair spectrum of economically advanced nations. Third, we lost a fraction of our sample due to non-response and missing data (effort-reward imbalance 8.0%; low control 4.8%; depressive symptoms 1.8%). However, additional analyses revealed only minor evidence of systematic bias. It should also be mentioned that the sample size in some countries was rather small, thus limiting the robustness of some analyses.
One general problem of longitudinal surveys is sample attrition. In the SHARE study the attrition rate between wave 1 and wave 2 is 27.9% [
43]. ELSA and HRS have lower attrition rates. This could have affected our results. However, the attrition rate was only slightly higher for employees with low control and there was no higher attrition rate for people with effort-reward imbalance. These limitations are balanced by several strengths. First, we were able to use comparable standardized measures of main variables of interest taken from three pioneering epidemiological studies on economic, social and health-related characteristics of ageing populations in Northern, Western and Southern Europe and the United States of America, SHARE, ELSA, and HRS. The respective sample was large enough to conduct multivariate statistical analyses with appropriate confounder control. Second, consistent associations of two theoretically grounded measures of work-related psychosocial stress with newly occurring depressive symptoms were observed. Third, we applied an innovative approach towards estimating a modifying role of national welfare policies by selecting three types of macro-structural indicators reflecting (1) active labour market policies, (2) employment protection by the state, and (3) distributive justice in terms of income inequality. These indicators are thought to represent protective resources against the threats of psychosocial stress at work, thus mitigating adverse effects on workers’ mental health [
11].
In view of the challenging occupational public health problem of depression in rapidly ageing societies the findings of this study, if supported by further evidence, may have important policy implications. Improved efforts of national labour and social policies to provide protective resources against the threats of stressful work to their workforce can contribute to a reduction of harmful effects on their mental health, especially so in times of unrestrained neoliberal policies and related financial crises.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All authors read and approved the final manuscript. TL, MW and JS drafted the manuscript. TL and MW conducted the empirical data analysis. ND contributed to the study design, revised the manuscript and contributed to its final version.