at SciVerse ScienceDirect

Social Science & Medicine 92 (2013) 114e123

Contents lists available

Social Science & Medicine

journal homepage: www.elsevier .com/locate/socscimed

Worker health is good for the economy: Union density andpsychosocial safety climate as determinants of country differencesin worker health and productivity in 31 European countries

Maureen F. Dollard*, Daniel Y. NeserCentre for Applied Psychological Research, School of Psychology, Social Work and Social Policy, University of South Australia, Magill Campus,Adelaide, Australia

a r t i c l e i n f o

Article history:Available online 21 May 2013

Keywords:EuropeWork stressUnion densityPsychosocial safety climateCross-national differences in workerself-reported healthGross domestic product

* Corresponding author. Tel.: þ61 434187253.E-mail address: Maureen.dollard@unisa.edu.au (M

0277-9536/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.socscimed.2013.04.028

a b s t r a c t

Work stress is recognized globally as a social determinant of worker health. Therefore we exploredwhether work stress related factors explained national differences in health and productivity (grossdomestic product (GDP)). We proposed a national worker health productivity model whereby macromarket power factors (i.e. union density), influence national worker health and GDP via work psycho-social factors and income inequality. We combined five different data sets canvasing 31 wealthy Euro-pean countries. Aggregated worker self-reported health accounted for 13 per cent of the variance innational life expectancy and in national gross domestic product (GDP). The most important factorsexplaining worker self-reported health and GDP between nations were two levels of labor protection,macro-level (union density), and organizational-level (psychosocial safety climate, PSC, i.e. the extent ofmanagement concern for worker psychological health). The majority of countries with the highest levelsof union density and PSC (i.e., workplace protections) were Social Democratic in nature (i.e., Sweden,Finland, Denmark, Norway). Results support a type of society explanation that social and economicfactors (e.g., welfare regimes, work related policies) in concert with political power agents at a nationallevel explain in part national differences in workplace protection (PSC) that are important for workerhealth and productivity. Attention should be given across all countries, to national policies to improveworker health, by bolstering national and local democratic processes and representation to address andimplement policies for psychosocial risk factors for work stress, bullying and violence. Results suggestworker health is good for the economy, and should be considered in national health and productivityaccounting. Eroding unionism may not be good for worker health or the economy either.

� 2013 Elsevier Ltd. All rights reserved.

Introduction

The new European policy framework for Health 2020 valueshealth as a human right, and intends to tackle social determinants ofhealth to improve public health (World Health Organization, 2011).An important social determinant of health recognized globally iswork related stress (Commission for the Social Determinants ofHealth, 2008). Work stress refers to “when the demands of thework exceed the employee’s ability to cope with or control them”

(European Survey on New and Emerging Risks e Psychosocial Risks(ESENER) 2009a, p. 26). Work stress represents a “huge cost” interms of the public health disease burden and worker health andproductivity (European Agency for Safety and Health at Work

.F. Dollard).

All rights reserved.

(EASHW), 2009). Work stress contributes about 5e10 per cent tothe total disease burden of depression, and 16 per cent to the totalburden of cardiovascular disease, equating to 2.5 million deaths peryear (Prüss-Üstün&Corvalán, 2006). Stress-related illnesses such asdepression and cardiovascular disease are forecast to be the leadingcauses of the global disease burden by 2020 (Murray & Lopez,1996).

In 2005, on average, 22per centofworkers in 27 EUmember statesexperienced stress (European Foundation for the Improvement ofLiving and Working Conditions, 2006). Work stress costs are nation-ally significant with workplace bullying costing 1.5 per cent of grossdomesticproduct (GDP)or£17.65billion in theUK(Giga,Hoel,&Lewis,2008), andwork stress illnesses costingaroundV20billionperannumacross the EU15 (EASHW, 2009). Given the significance of theproblemthe aim of this research is to explore the central role of workin explaining national differences in levels of worker health and GDP.

If worker health is nationally important, it is crucial to under-stand its antecedents. In the current study we examine the

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123 115

influence of psychosocial factors at work, defined as the job design(e.g. work pressure, job control), organization, and management ofwork that causes stress (Cox, Griffiths, & Rial-Gonzalez, 2000).Beyond commonly explored job design factors we also investigateworkplace protection factors, the role of labor unions, psychosocialsafety climate (i.e. policies and procedures for worker psychologicalhealth and safety, Dollard & Bakker, 2010) and job redesign forworker health and productivity.

Increasingly researchers frame work stress as a problem withmultilevel causes (Kang, Staniford, Dollard, & Kompier, 2008). Mostresearch focuses on within-organizational factors. Yet there aremany important external or macro-level factors that impinge onthe workplace (Dollard, Osborne, & Manning, 2012a), including thetype of society implied by welfare state regimes, and nationalpolicy and regulation approaches to occupational safety and health(OSH). However these external factors are not readily explored inwithin-country studies. In this study, we propose that nationalfactors such as the broader political context, influence work con-ditions, and how and why organizations tackle psychosocial factorsat work. We examine the antecedents of worker health and itsimpact on GDP in 31 European countries. We draw on recentframeworks (Benach, Muntaner, & Santana, 2007; Navarro et al.,2006) and empirical work that links labor policies and welfarestate regimes to poor work quality and older worker depressivehealth symptoms (Dragano, Siegrist, & Wahrendorf, 2011). Inparticular we focus on two levels of labor protection, macro-level(union density), and organizational-level (psychosocial safetyclimate).

Significance of worker health

A healthy workforce is likely to have a significant impact onnational life expectancy and national productivity estimated interms of GDP. Work stress theory suggests that stress compro-mises worker health through an erosion of energy. Since outputsat work require energy inputs, performance is likely compromisedwhen health is not optimal. Further, workers experiencingstressful conditions may reciprocate by reducing commitment andengagement and in sequence reduce outputs (Xanthopoulou,Bakker, Demerouti, & Schaufeli, 2009). Among restaurantworkers, working conditions and worker health precede engage-ment and when workers are more engaged financial returns arehigher (Xanthopoulou et al., 2009). Moreover a longitudinal meta-analytic study of 7939 business units found that engagement isrelated to greater productivity and profit, with highly engagedunits returning increased profits of $80,000 to $120,000 permonth (Harter, Schmidt, & Hayes, 2002).

Worker health is also related to the cost of production. Whenaccidents, errors and turnover occur as a result of stressful condi-tions, production costs go up because of the associated costs relatedto sick leave, compensation and replacement costs of labor.Research from Australia shows that depression costs employersapproximately AUD$8 billion per annumdue to productivity loss (.5per cent GDP) because of sickness absence and presenteeism (i.e.,reduced performance at work) (McTernan, Dollard, & LaMontagne,2013).

Hypothesis 1. Worker self-reported health is positively associatedwith (a) life expectancy and (b) GDP.

Antecedents to worker health

Work plays a central role in many people’s lives. The averageperson in the 27 EU member states works around 61, 295 hours or10.6 per cent of a lifetime (Volger-Ludwig, 2009). Exposure to

stressful work conditions is likely to have an aggregate effect onworker health at a national level. Substantial evidence has nowaccumulated demonstrating that psychosocial factors such as highjob demands and low job control (Karasek & Theorell, 1990) arerelated to high blood pressure (Rau, 2004), and cardiovasculardisease (Kivimäki et al., 2012). Additionally, there is substantialevidence linking violence and bullying to physical and mentalhealth problems (Black, 2008; Leka & Jain, 2010). In this study weexamine psychosocial factors that relate to job design in the broadterms of job quality - the higher the quality the lower the psy-chosocial risks.

In addition to psychosocial stressors there are several workplaceprotective factors that play a role that have been ignored untilrecently. Psychosocial safety climate (PSC) is aworkplace protectivefactor that reflects the will of management to prevent and respondto stressful conditions. Psychosocial safety climate concerns howmanagement values worker psychological health, commits to andsupports psychological health protection, and prioritizes the psy-chological health of workers over profit and productivity (Hall,Dollard, & Coward, 2010). Psychosocial safety climate theory pro-poses that PSC is the “cause of the causes” of common psychosocialrisks (Dollard, 2012), and is a pre-eminent psychosocial risk factor(Dollard & Bakker, 2010). In a high PSC contextmanagers will have arange of policies, practices and procedures in place to ensure thatwork conditions are not too demanding for workers, that resourcesare adequate tomanage demands, and that overt psychosocial riskssuch as bullying, and violence and more subtle forms of aggressionlike incivility (Cortina, Magley, Williams, & Langhout, 2001), andmicroaggressions (Wing Sue, 2010), are not tolerated (Bond,Tuckey, & Dollard, 2010; Law, Dollard, Tuckey, & Dormann, 2011).In addition to having a preventative role, PSC may have a bufferingrole. Psychosocial safety climate may act as a safety signal to em-ployees indicating when it is safe to utilize personal resources (e.g.,coping strategies) and/or organizational resources (e.g., utilize au-tonomy) to cope with job demands (Dollard, Tuckey, & Dormann,2012c). In a practical sense, workers may not be inclined toreport bullying, or more subtle forms of incivility, or seek sup-portive resources when PSC is low (Dollard et al., 2012c).

Empirical research shows support for the dual roles of PSC. First,in relation to its preventative role, several longitudinal studiesshow that PSC negatively predicts psychosocial risk factors (e.g.emotional demands, bullying, harassment), that in turn are posi-tively related to psychological health problems (Bond et al., 2010;Dollard & Bakker, 2010, Dollard et al., 2012b; Idris & Dollard, 2011;Idris, Dollard, Coward, & Dormann, 2012; Law, et al., 2011). Second,in relation to the buffering role, several longitudinal studies havefound that PSC moderates the effects of demands and bullying onpsychological health outcomes (Bond et al., 2010; Dollard & Bakker,2010; Dollard, et al., 2012c; Law et al., 2011).

The pervasive effects of PSC are demonstrated in researchwherePSC assessed by one group of workers can predict work conditions(e.g., workload, control) and psychological strain at a later time indifferent workers in the same work unit (Dollard, et al., 2012b).Results combining lagged PSC data from different sources withinthe same study provide a strong test of the fundamental idea ofclimate as a property of the organization, and independent of theindividual. These empirical and theoretical examples suggest thatthe application of PSC in this research is important to characterizean essential aspect of the workplace setting that relates to workerhealth protection.

In this study, consistent with previous operationalizations, wemeasure PSC in terms of policies, practices and procedures forpsychosocial risks (stress, bullying and violence), and whetherthere is participation and consultation of employees in relation totaking measures to deal with psychosocial factors (Dollard &

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123116

Bakker, 2010). Although we have been talking in terms of organi-zational PSC, we expect that pressure through labor market pol-icies, procedures and regulations for worker healthwould influenceorganizational PSC, which in sum, is seen as differences in PSC atthe national level.

Finally job redesign actions are important to also consider asprotective factors because they intend to improve aspects such aswork-time arrangements, the organization of work, and the wayconflict is resolved, and should be related to better worker health(Karasek & Theorell, 1990). Summation of these job redesign ac-tions may provide an indicator of the importance of these at anational level.

Workplace factors at the individual, work unit (Dollard, et al.,2012c), and organizational level (Law et al., 2011) have been usedto determine worker health. Untested to date is the hypothesis thataggregate national levels of workplace factors (themselves affectedby national factors) determine differences in worker health statusbetween countries. Yet there is evidence of differences betweenEuropean countries on work conditions and worker healthand reason to assume they could be linked at the national level(Eurofound, 2011a).

Hypothesis 2. Psychosocial safety climate, job redesign, andquality psychosocial work conditions, are positively related toworker self-reported health all aggregated to a national level.

External influences on worker health

Organizations function within a larger context. To enhance ourunderstanding of the contextual origins of working conditions andworker health and their links to external political, social, and eco-nomic variables, several macro-models have been proposed(Benach et al., 2007; Navarro et al., 2006). These models integratepower relations, labor markets, and welfare regime interactionswith employment conditions, working conditions, and healthoutcomes. Beginning with the larger institutional context, powerrelations between the major social actors in the market (unions,corporations, institutions), political parties and societal groups,determine labor market (work conditions policies) and welfareregimes (social policies) and the consequent level of active policieson employment and social protection (Benach et al., 2007) (seeFig. 1).

These contextual factors cluster to characterize nations, andvarious typologies have been developed to distinguish countries by

Fig. 1. The National Worker Health Productivity Model; correlates of worker self-reporte

political ideologies and welfare regimes. Welfare regimes are alsohighly related to labor market policies such as worker health andsafety legislation (Benach et al., 2007; Epsing-Anderson, 1990).Here we draw on the work of Epsing-Anderson (1990; 1999). SocialDemocratic countries such as Norway, Sweden, Denmark, andFinland have a strong welfare system, and citizens are less depen-dent on the labor market (i.e. having a job) to ensure a reasonablestandard of living; Liberal or neo-liberal countries such as Ireland,United Kingdom (UK), United States, and Australia have weakwelfare regimes, are the most labor market dependent, and there isan emphasis on means testing and income testing to access wel-fare; the Conservative countries (Germany, France, Netherlands,Belgium) have intermediate welfare regimes and rely more heavilyon family for security, and on insurance based schemes (Arts &Gelissen, 2002; Epsing-Anderson, 1990; Coburn, 2004). The impli-cations of this for the workplace is that for instance in SocialDemocratic countries (e.g. Sweden) we expect more protections,stronger trade unions (Coburn, 2004) and the quality of psycho-social work conditions and health and safety responses (laborpolicies) and workplace protection such as PSC to be stronger, thanin either the Conservative or Liberal countries. Social Democraticcountries would also have more workplace health and safetyregulation and democratic frameworks that would promoteemployee participation (Coburn, 2004). Bringing together types ofsociety perspectives, and the integrative macro-models (Benachet al., 2007; Navarro et al., 2006) we propose that these social,economic and political factors at a national level affect work con-ditions and workplace protections and in turn national workerhealth and GDP. For instance Dragano et al., (2011) found that laborpolicy, economy-related macro factors (e.g., indicated by old-ageemployment rate), and welfare regime typology were related tovariations in older workers perceptions of work quality. Further thelink betweenwork conditions and health varied by welfare regime,and was higher in Liberal regimes compared with Social Demo-cratic regimes.

Union density is a macro-level indicator of power relations thatarguably have worker health as a central agenda. The Europeanmodel of industrial relations includes collective representation ofemployee rights and employee participation rights. At a nationallevel, trade union density, or the proportion of workers who aremembers of trade unions, is an indicator of the role of these or-ganizations in the regulation of employment and work conditions(Eurofound, 2011a). For instanceworker unions potentially increaseresources to employees via wages and benefits, compensation, pay

d health and gross domestic product in 31 European countries using five data sets.

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123 117

parity and workplace protection, all aspects that relate to workerhealth. There is evidence that politics and policies associatedwith union density predict health at the national level (Navarroet al. 2006). Further, via worker representation there is increasedpotential for worker participation and control in the workplace toalter and shape work conditions that also relate to worker health.Union density likely influences workplace union activity and in-creases government labor policy enforcement above the role ofhealth and safety committees (Weil, 1999). Here we propose uniondensity is effectual through its influence on workplace protectionand work conditions.

Hypothesis 3a. Union density is positively related to workplaceprotection (PSC, job redesign) and quality work conditions that inturn are related to worker health and GDP.

Union density also overlaps with welfare state regime types(Benach et al., 2007) and may also indicate influence over redis-tributive policies (Navarro et al., 2006). Via its linkwith bothwelfareregime and labor market policies, union density may relate toworker health because of material deprivation and economicinequality. Through the life course workers may be exposed topoorer educational opportunities, or employment insecurity withvariable degrees of protection such as income security (unemploy-ment) benefits, depending on the welfare state and the pressure ofpolitical interest groups (i.e., unions). Benach et al., (2007) reportthat a greater decline in the unionization rate in the United Statescompared to Canada largely accounts for the divergent growth inwage inequality between the two countries. The link between in-come inequality and population health has been shown empirically(Marmot & Wilkinson, 2001) and discussed theoretically (Lynchet al., 2001) but not specifically in relation to worker health.

Hypothesis 3b. Union density is negatively related to incomeinequality that in turn is related to worker health and GDP.

In sum we propose that worker health at a national level de-pends on these macro factors (indicated by union density) that mayaffect health via two pathways. The first is via workplace factorsand the second is via income inequality.

Method

Data were obtained for 31 European countries including the 27EU Member States (Austria; Belgium; Bulgaria; Cyprus; Czech Re-public; Denmark; Estonia; Finland; France; Germany; Greece;Hungary; Ireland; Italy; Latvia; Lithuania; Luxembourg; Malta;Netherlands; Poland; Portugal; Romania; Slovakia; Slovenia; Spain;Sweden; UK; Switzerland) plus Croatia, Turkey, and Norway. Allcountries had total purchasing power parity (PPP) converted toGDP per capita greater than international $10,000.

Data analysis was undertaken at the University of SouthAustralia, Australia in 2012. Since the study used secondary de-identified data, ethics approval was not sought beyond what isdocumented by the various data sources. Data were derived fromdifferent sources and matched at the country level.

Population size by country was determined using the 2009 PennWorld Tables 7.0.

GDP was assessed using total PPP converted to GDP per capita(2009 Penn World Tables 7.0).

Income inequalitywas assessed using the GINI coefficient, whichassesses family income inequality with values ranging from0 ¼ equality to 1 ¼ inequality. We used the GINI % from the period2005e2009 (CIA World Fact Book, 2011).

Quality work 2009 (managers) was assessed via interviews withN ¼ 28 649 senior OSH managers (ESENER, 2009a, 2009b). Man-agers were asked, if each of the following issues is a “major concern,

some concern or no concern at all in your establishment”: (a) work-related stress, defined to participants as “when the demands of thework exceed the employee’s ability to cope with or control them”

(MM 200.5); (b) violence or threat of violence, defined as “whenone or more workers or managers are threatened, assaulted orabused by clients, patients or pupils” (MM 200.6); and (c) bullyingor harassment, defined as “when one or more workers or managersare abused, humiliated or assaulted by colleagues or superiors”(MM 200.7). Responses were recoded as major concern ¼ 0, someconcern ¼ 1, no concern ¼ 2, not applicable ¼ missing, and theemployer weight was used. Alpha was .81.

Quality work 2009 (employee representatives)was assessed usingESENER data via interviews with N ¼ 7 226 employee representa-tives responsible for OSH in the organizations (ESENER, 2009a).They were asked the mirror questions as per the managers. Alphawas .81.

Quality work 2010 (workers in paid employment). An indicativemeasure of quality work was job control assessed using 10 itemsfrom the European Working Conditions Survey (EWCS) 2010 e.g.“does your job involve learning new things” (q49f0). Responseswere reverse scored to 1 ¼ never, 5 ¼ always, high scores nowindicating high control. Alpha was .81. As noted later other in-dicators such as job demands, and supports were also investigatedbut were not related to health nationally.

Union density refers to the percentage of the workforce that areunion members; national levels were assessed using combineddata from the OECD (2009), Hall-Jones (2010), and the Eurofound(2011b) for Croatia. There was a positive correlation between theESENER measure of shop floor union representation and uniondensity, r ¼ .73, p ¼ .001.

Psychosocial safety climate. Using the ESENER data (ESENER,2009a) OSH managers were asked five questions that representedbest, procedures to deal with psychosocial risks, and consultationand participation in the resolution of psychosocial risks in theworkplace. Questions were: “Does your establishment have a pro-cedure to deal with, (1) work-related stress (MM 250); (2) bullyingor harassment (MM 251); (3) work-related violence (MM 252)?”(reversed to No ¼ 0, Yes ¼ 1, not an issue ¼ system missing); (4)“What about the role of employees: Have they been consultedregarding measures to deal with psychosocial risks (MM 266)?”and (5) “Are employees encouraged to participate actively in theimplementation and evaluation of the measures?” (MM 267)(recoded to Don’t know ¼ system missing, No ¼ 0, Yes ¼ 1). Alphawas .87.

Job redesignwas assessed using four items from the ESENER data(ESENER, 2009a). OSHmanagers were asked “In the past 3 years hasyour establishment used any of the followingmeasures to deal withpsychosocial risks”: (1) “changed the waywork is organized?” (MM253.1); (2) “a redesign of thework area?” (MM253.2); (3) “set-up ofa conflict resolution procedure?” (MM 253.4); and (4) “changedworking time arrangements?” (MM 253.60) (recoded to Don’tknow ¼ system missing, No ¼ 0, Yes ¼ 1). Alpha was .93.

Worker self-reported general healthwas assessed using the EWCS(2010) data where workers in paid employment (N ¼ 34, 841) wereasked to rate their general health (q68) on a scale from 1 to 5,reverse scored for this study, high scores indicating high health;1 ¼ very bad, 2 ¼ bad, 3 ¼ fair, 4 ¼ good, 5 ¼ very good. Data wereweighted for design, stratification and population (w5_all_e). ForSwitzerland we interposed data from the European Social Survey(2010). Our health measure is a specific outcome measure (c.f. lifeexpectancy), relevant for an age-specific, cohort-specific (worker)group, and our selection of work specific psychosocial factors(Lynch, Davey-Smith, et. al., 2001).

Life expectancy at birth. This refers to the average years of lifeexpected for a child at birth if the current national mortality rates

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123118

remained constant during their lives. It is expressed in terms ofyears (United Nations, 2011). We used the life expectancy at birth2005e2010 UN measure (United Nations, 2011).

Fig. 2. Worker self-reported health by psychosocial safety climate by union density in31 European countries.

Statistical analyses

We used IBM SPSS 19 software for correlational and linear andhierarchical regression. To test mediated paths we used structuralequation modeling (SEM) and IBM SPSS Amos 20 software. Eachvariable was modeled with a residual term. To assess model fit weused the Root Mean Square Error of Approximation (RMSEA), thec2 goodness-of-fit statistic, the Goodness of Fit Index (GFI), theComparative Fit Index (CFI), the Non-Normed Fit Index (NNFI), andthe Akaike Information Criterion (AIC). For a good model fit, c2should be �.05; relative fit indices (CFI and NNFI) (Hu & Bentler,1999) and GFI values �.95 (Hoyle, 1995); RMSEA values should beclose to zero; and smaller levels of AIC indicate a better fit(Schermelleh-Engel, Moosbrugger, & Mueller, 2003). We testedstandardized indirect effects (mediation) directly via bootstrappingand the bias corrected percentile method to estimate confidenceintervals (CI). We log transformed the GDP measure to correct forskew due to unusually high GDP in Luxembourg.

Results

First, we examined the variables from the ESENER and EWCSdata sets that were collected at the individual level to calculate theper cent of variance due to country level factors in the data.Between-country variance explained 22 per cent of the variance inwork quality as reported by managers, 9 per cent in job control asreported by workers, 23 per cent in PSC, and 13 per cent in jobredesign. Between-country differences also explained 6 per cent ofthe variance in worker health (F (29, 34725) ¼ 46.37, p ¼ .0001).This preliminary analysis provides some support for the decision toinvestigate sources of these between-country differences.

Table 1 reports the correlations between the study variables.Union density and workplace protection measures (PSC, job rede-sign) were all significantly positively related to quality of work as

Table 1Correlations between variables among 31 European countries.

2009GDPf

p-value Incomeinequality

p-value Qualitywork(managers)

p-value Quawo(wocon

Societal attributesIncome inequality2005e2009a

�0.45 0.01

Quality work (psychosocial)Quality work 2009(OSH managers)b

�0.02 0.91 �0.38 0.04

Quality work 2010(workers)c

0.35 0.06 �0.38 0.04 0.08 0.67

Workplace protective factorsUnion density 2009d 0.43 0.02 �0.59 0.001 �0.19 0.30 0.4Psychosocial safetyclimate 2009(OSH managers)b

0.39 0.03 �0.22 0.23 �0.10 0.61 0.5

Job redesign 2009(OSH managers)b

0.00 0.99 0.03 0.89 0.40 0.03 0.3

HealthLife expectancy2005e2010e

1.00 0.0001 �0.44 0.02 0.02 0.92 0.2

Self-reportedhealth 2010c

0.58 0.001 -0.17 0.36 -0.10 0.60 -0.0

Notes. N ¼ 31. Superscripts denote different data sets: aGINI CIA World Fact Book 201Nations, 2011; fPenn World Tables 7.0, 2009. Exp., expectancy; OSH, occupational safetyspace as correlations closely followed those of Quality work 2009 (OSH managers) and t

reported by workers. In countries with high levels on these pro-tection measures, the quality of work was also reported morefavorably. Job redesign was also positively associated with workquality as reported by managers. Quality of work however was notrelated to health outcomes or GDP, but PSC and union density were(see Fig. 2). Income inequality and GDP were negatively relatedand both were related to union density (negatively and positivelyrespectively). As expected worker health was positively associatedwith life expectancy at a national level. When controlling forincome inequality and union density and work factors, workerhealth accounts for 13 per cent of the variance in life expectancy(Beta ¼ .44, p ¼ .03) supporting Hypothesis 1a.

Hypothesis 1b proposed that worker self-reported health ispositively associated with GDP. As shown in Table 2, Model 1, therelationship was significant and positive, Beta ¼ .58, p ¼ .001, thevariables sharing 34 per cent of their variance. After controlling forother variables worker health explained 13 per cent of the variancein GDP (step 2) (Table 2).

lityrkrkers)trol

p-value Uniondensity

p-value PSC p-value Jobredesign

p-value Lifeexp.

p-value

9 0.0080 0.005 0.56 0.001

1 0.03 0.28 0.15 0.58 0.001

0 0.39 0.43 0.02 0.39 0.03 0.00 0.99

2 0.92 0.44 0.02 0.53 0.003 0.13 0.48 0.57 0.001

1; bESENER; cEuropean Working Conditions Survey (2010); dOECD (2009), eUnitedand health. Quality work 2009 (worker representatives) not included here due tohese were intercorrelated r ¼ .92, p ¼ .0001; GDP is log transformed.

Table 2Regression models with worker self-reported health and gross domestic product asoutcomes in 31 European countries.

Model 1 Gross domestic product (PPP) R2

change(%)

B Std. error Beta t p-Value

(Constant)a 1.70 0.05 34.45 0.0001 34Worker self-reported health 0.05 0.01 0.58 3.87 0.001(Constant)a 1.78 0.05 33.32 0.0001 43Population 0.00 0.00 0.26 1.62 0.12Income inequality 0.00 0.00 -0.36 -2.08 0.05Psychosocial safety climate 0.00 0.00 0.04 0.20 0.85Union density 0.00 0.00 0.13 0.59 0.56Worker self-reported health 0.04 0.01 0.44 2.59 0.02

F (5, 25) ¼ 5.51, p ¼ .001, Adj R2 ¼ .52

Model 2 Worker self-reported health R2

change(%)

B Std. error Beta t p-Value

(Constant)a 3.86 0.07 54.52 0.0001Union density 0.01 0.002 0.41 2.42 0.02 17(Constant)a 3.67 0.11 33.58 0.0001Union density 0.002 0.002 0.18 0.93 0.36Psychosocial safety climate 0.11 0.05 0.42 2.22 0.04 12

F (2, 28) ¼ 5.77, p ¼ .008, Adj R2 ¼ .24 R2

Model 3(Constant)a 3.66 0.33 11.09 0.0001 18Income inequality 0.006 0.009 0.13 0.61 0.55Union density 0.005 0.002 0.48 2.30 0.03

F (2, 28) ¼ 3.05, p ¼ .06, Adj R2 ¼ .12

a New step in model.

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123 119

Hypothesis 2 proposed that work factors related to countrydifferences in worker health status. As shown in Table 1 only PSCwas significant. Hypothesis 2 is supported in relation to PSC.

Hypothesis 3a and 3b will be tested more fully using SEM. Butfirstwehadapreliminary look atmodeling effects usinghierarchicallinear regression. Hypothesis 3a proposed a mediation processwhereby union density related to worker health via PSC. First weconfirmed the direct effect of union density to worker health. Asshown in Table 2, Model 2 step 1, union density was significantlypositively related toworkerhealth, Beta¼ .41,p¼ .02.NextweaddedPSC to the model which was significantly related to health,Beta ¼ .42, p ¼ .04. Consistent with a mediation process the rela-tionship between union density and health reduced to zero in thisstep. In relation to Hypothesis 3b proposing that union density isrelated to worker health via income inequality, we already knowfromModel2 that uniondensity is related toworkerhealth.Next,wemodeled both union density and income inequality together, andonly union density was related to worker health. Hypothesis 3b isnot supported. As shown inTable 1 income inequality is correlated tolife expectancy but not to the specific worker health measure.

To test the full Hypothesis 3a regarding the mediated link be-tween union density, worker health and GDP via workplace factorswe used SEM. In this model we also included the relationship be-tween union density and income inequality and a covariation be-tween income inequality and GDP (or absolute income) as these areinterrelated as suggested by several authors (Lynch et al., 2001).The fit of the model was good, Chi square ¼ 3.63, df ¼ 5, p ¼ .60,GFI ¼ .96, NFI ¼ .93, CFI ¼ 1.00, RMSEA ¼ .000, AIC ¼ 23.63.

In terms of Hypothesis 3a union density was significantly indi-rectly positively related to worker health, Beta ¼ .29, CI (.07, .47),p ¼ .01, and GDP, Beta ¼ .17, CI (.04, .30), p ¼ .01; and PSC wasindirectly related to GDP, Beta¼ .31, CI (.10, .49), p¼ .01. Hypothesis3a is supported with union density working through PSC to posi-tively affect health and GDP. The standardised effects betweenvariables is shown in Fig.1, the NationalWorker Health ProductivityModel.

A final point is that in the model as specified here there is nodirect reverse relationship from GDP back to health, or a mediatedrelationship back through union density. We found that the modelexplained 30% of the variance in PSC, 27% of the variance in health,and 36% of the variance in GDP, and 28% of the variance in incomeinequality.

Given the significance of PSC we provide a graphic representa-tion of the correlation with health by country (Fig. 2) and levels bycountry since these data are not readily known (Fig. 3).

Given the importance of PSC and union representation weconverted PSC scores to percentages as per union density thenweighted these factors according to total effects (indirect anddirect) Beta weights for union density (.25) and PSC (.49) onworkerhealth in the final model (Fig. 2), and then added them together(see Table 3). We then examined the results in relation to theclassification of nations following Epsing-Andersen (Arts &Gelissen, 2002; Coburn, 2004).

Discussion

Given the widespread health and economic burden of workstress, and using work stress theory, we reasoned that worker self-reported health that is due to workplace factors, aggregated at thenational level, would be related to GDP. Based on types of societyand macro-level models we also considered that power relations,welfare state regimes, and labor market policy influences asexemplified by union density would affect worker health status viaworkplace psychosocial factors and income inequality. We useddata from 31 European countries to capture these between-countryeffects.

This is the first study to propose and find a direct link betweenworker health and life expectancy and also GDP. Supporting thesignificance of worker health for national health we found thatworker health accounted for 13 per cent of the variance in life ex-pectancy at a national level. In accord with work stress theory, acentral finding of our study, is that national levels of worker healthare positively related to national levels of GDP, and account for 13per cent of its variance. A further innovation of our research was toidentify specific work related factors as potential determinants ofworker health and productivity differences between countries. It isimportant to note that it was not the quality of work conditions, butrather, workplace protective factors at both the macro (uniondensity) and organizational level (PSC) that were most importantfor worker health (partly supporting Hypothesis 2). We used avariety of measures of quality (bullying, stress, and violence; jobdemands, control and support) from different sources (worker, OSHmanager, worker representatives) but none of the measuresshowed significant links to national worker health. Indirect effectanalysis showed that union density was related to PSC that insequence related to worker health and GDP in support of Hypoth-esis 3a.

We also considered the possible impact on worker health fromunion density via income inequality. Union density had an impor-tant negative link to income inequality. However the incomeinequality hypothesis as proposed by Wilkinson (1997) linking in-come inequality to nation health status was not supported in thiscase where worker health was the criterion.

For worker health our main finding is that national health in-equalities have their basis in national power and structural factorsimplied by union density that give rise to concrete psychosocialconditions, individual resources, public resources and workingconditions that could affect worker health (see also Lynch et al.,2004). This paper has uncovered a process whereby external fac-tors influence worker health; in this case via PSC-workplace pro-tection policies and practices.

Fig. 3. Psychosocial safety climate by European country.

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123120

When we looked at the socio-political characteristics of thesocieties that give rise to both high levels of union density and PSC,all of the four nations referred to as Social Democratic by Epsing-Andersen, Sweden, Norway, Finland, and Denmark, were rankedin the top five of the 31 countries. The Conservative countries(Belgium, The Netherlands, Italy, France, Germany, Switzerland),were dispersed and were ranked sixth and tenth and also appearedin the bottom tertile of all the nations (Germany twenty second;France twenty ninth). Of the two identified liberal countries,Ireland and the UK, both were ranked in the top ten, and wereranked fourth and seventh. This pattern supports the propositionthat the nature of society, its fundamental ideology regarding labormarket legislation, policies and welfare regime, shaped by and inconcert with power actors such as unions and management giverise to important protective structures and functions within theworkplace that in turn affect worker health. We expected the bestconditions would be evident in Social Democratic societies. It wassurprising that the Neo-liberal countries were ranked somewhatmore highly than the Conservative countries, mainly due to PSCscores rather than union density. A possible explanation is that PSC

Table 3Top ten European countries by psychosocial safety climate and union density, by nation

Rank Nation PSC Nation Union density

1 Sweden 86 Sweden 712 Ireland 85 Finland 703 Finland 81 Denmark 694 United Kingdom 80 Cyprus 685 Denmark 71 Malta 596 Norway 70 Norway 547 Belgium 70 Belgium 538 Croatia 70 Luxembourg 469 The Netherlands 70 Slovenia 4510 Romania 54 Austria 40

Note: PSC, psychosocial safety climate; eno typology, based on Epsing-Andersen typolo

is also being driven by internal professional management needsalongside any union influence. To shed further light on drivers forPSC we looked at OSH managers explanations in the ESENER data.Aggregated at a national level themain reasons OSHmanagers gavefor dealing with psychosocial risks, in order, were legal re-quirements (63 per cent), and requests from employees or theirrepresentatives (45 per cent) (this accords well with our uniondensity finding). An additional reason for psychosocial risk man-agement that goes to the professional management explanationwas high absenteeism rates (reported by 18 per cent of managers).This internal stimulus was statistically related to PSC (r ¼ .68, p ¼.001). Perhaps because of concern for loss of productivity this in-ternal problem is a trigger for a stronger PSC. Management will andunion representation have been previously highlighted by Dollardand Karasek (2010) as key drivers for PSC. Further, national policyand strategies may be influential in the neo-liberal countries. In thepast decade the UK and less prominently Ireland, have beenworking to improve national dialog and interventions regardingpsychosocial factors at work, and several strategies such as theManagement Standards have been implemented by the Health and

type.

Nation PSC þ union density# Nation type

Sweden 54 Social DemocraticFinland 52 Social DemocraticDenmark 48 Social DemocraticIreland 44 LiberalNorway 44 Social DemocraticBelgium 43 ConservativeUnited Kingdom 41 LiberalCroatia 39 e

Malta 36 e

The Netherlands 35 Conservative

gy (Coburn, 2004).

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123 121

Safety Executive (2012). These may now be having an effect interms of policies and practices in the workplace. When risks forwork related stress are being managed and controlled throughManagement Standards a strong PSC is built. It is important then tostress that good practice policies accompanied by practical guid-ance can be beneficial in various country context and this shouldrepresent a priority going forward.

Importantly our measure of PSC was reported by OSHmanagers,and was therefore beyond individual worker perceptions yet wasstill related to worker health. We propose therefore that this factoralong with union density reflect concrete circumstances forworkers that have serious implications for worker health. Psycho-social safety climate reflects management care for worker healthand well-being via stress prevention procedures. Previous researchhas shown that PSC has important effects on health via its effect ona range of work conditions, and work behaviors (e.g. bullying). Weunsurprisingly note in this study that PSC, union density, and jobredesign are all positively related to quality work conditions re-ported by workers. The fact that PSC is a pre-eminent organiza-tional level determinant of many of these lower level psychosocialfactors may explain its salience here above any specific work con-dition (Dollard & Bakker, 2010). Recall that these results are at thenational level; at the individual level work conditions wereconsistently correlated with worker health.

Limitations

Although we made directional assumptions, the non-experimental design severely limits adequate tests of direction-ality. It is important to note though that in Fig. 1 an additionalreverse path from GDP to worker health is not significant, givingus confidence of a stronger clockwise rather than counter-clockwise process. Nevertheless modeling GDP as an outcomeof worker health is an innovation of this study and againstconventional economic approaches where GDP is modeled priorto national health. Future research could directly test the direct,reverse and reciprocal relations between the two using repeatedmeasures. Also the power to detect some effects may have beenreduced because of the relatively small number of countriesavailable for analysis.

A particular problem in using self-report health categorical datais cross-population comparability (Murray, Tandon, Salomon,Mathers, & Sadana, 2002). Beyond reliability and validity is aconcern about how individuals use categorical response scales. Ourresults showed uniform responses across the 31 countries forinstance on the bad and very bad category of the health measurewith variations of only 0 per cent to 5 per cent, providing confi-dence that the measurewas not affected by countries differences inresponse tendencies.

Our results pertain towealthy European countries, andmay varywhen other international data sets are used. An issue for thisresearch is the ecological fallacy problem (Shively, 1969). The da-tabases, ESENER and EWCS, were of data collected at the individuallevel (workers or OSH managers), that were then aggregated to thecountry level. The ecological fallacy is to infer that observationsfound at the national or population level apply at the individuallevel. Because of the great variability of the observations within thepopulation, we cannot predict individual level results withincountries based on knowing about the national levels. We cannotsay, for instance, that individuals in workplaces in Sweden havehigh levels of PSC; or that all individual in workplaces in Norwayhave high levels of union density. We can only talk in terms ofaverages.

Our research focused on union density and we made assump-tions about the kinds of policies that might influence workplace

factors. Additional research may explore more directly specific la-bor policies and their direct effect on worker health (see Draganoet al., 2011), and go further to look at external policy influence oninternal workplace policy development. Future research mayexplore and leverage internal triggers for better managementpractices as suggested here (i.e. high absenteeism rates, workerscompensation rates).

Although we assumed job types were constant across countries,it is possible that more stressful jobs have emerged in less wealthynations, and this explains national health differences. Evidenceagainst this is that we found that job quality was not the mainworkplace factor explaining worker health differences by nation.Rather it was variability in workplace protections between coun-tries rather than job quality that was most important. Finally, wedid not investigate the impact of more subtle psychosocial factorssuch as incivility and microaggressions an additional line ofresearch for the future.

ESENER researchers have recently derived an 8 item OSH_-psycho index as a measure of the scope of the management ofpsychosocial risks (Van Stolk, Staetsky, Hassan, & Woo Kim, 2012)that overlaps in content with our PSC measure. The two scales arehighly correlated, r ¼ .85, p ¼ .001 at the individual level, and atthe country level, r ¼ .91, p ¼ .001. The OSH_psycho index washowever not as strongly related to worker health (r ¼ .30, p ¼ .10),life expectancy (r ¼ .33, p ¼ .08), or GDP (r ¼ .30, p ¼ .10), at anational level compared with PSC. Our study shows how the indexcan be slightly modified so that it can be conceptualized theo-retically and practically along the lines of PSC theory. Strongerconclusions can then be drawn based on prior evidence of ex-pected effects.

Implications

The results point to the importance of national ideology, policyand power (i.e., union density) to promote workplace PSC - themanagement of, and reaction to, psychosocial risk factors at work-as an important strategy to build national health and productivity.The PSC of organizations (i.e., specific policies for managing psy-chosocial risks) is a system feature of organizations that preventsstressful work conditions and/or ameliorates the health impacts ofhazardous work and has public health implications. An implicationof the research is that national approaches should require themeasurement and monitoring of PSC in the workplace. Promotingdemocratic processes that enable participation and representationis important as these are related to better work conditions, otherprotective factors, and better health. Specific national in-terventions focused on management awareness and training onrisk management and control, as specified by PSC, would be veryworthwhile.

We measured both union density and union representation onthe shop floor. The external factor, union density, was moreimportant, suggesting additional leverage can be brought to workhealth via external lobbying and macro policy development. Re-sults suggest that eroding unionism as is evident in neo-liberaleconomies may not be good for worker health or the economyeither.

Theoretically our research supports extending frameworks ofworker health to include macro-level factors as well as GDP.Additionally a healthy workforce also means that, on aggregate, theworkforce itself is earning a wage higher than an unhealthyworkforce. Moremoney in the hands of workersmeans that there ismore money in the hands of consumers and consumption spendingincreases. Finally work stress researchers need to look at macrolevel factors both within the organization and external to it toadequately model the multi-source origins of work stress.

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123122

Conclusion

A healthy workforce is good for the economy. The observationthat worker self-reported health is related to GDP and life expec-tancy assessed at a national level underscores the importance of thework context for national health and productivity status. Twolevels of labor protection, macro-level (union density), and orga-nizational-level (PSC) were most important for worker health. Theprominence of union density and its influence onworkplace PSC forworker health raises the possibility that external forces can gofurther toward improving working circumstances (e.g. additionallegislation, guidance). Given the costs associated with work stress,the implementation of the recent international policy agenda onsocial determinants of health, to expand participatory OSH policyand programs to address work-related stress issues, is crucial(Commission on Social Determinants of Health, 2008; Leka & Jain,2010).

Acknowledgment

EWCS and ESENER data accessed under registrationwith the UKData Service of the University of Essex. ESENER data, http://www.esds.ac.uk/findingData/snDescription.asp?sn¼6446&key¼esener,established by the European Agency for Safety and Health at Work(EU-OSHA) and the UK Data Archive. EU-OSHA and the UK DataArchive bear no responsibility for the further analysis or interpre-tation of the data. The paper benefitted by helpful comments fromRobert Karasek, John Lynch, Helge Hvid, Michelle Tuckey, ValerieO’Keefe, Carolyn Boyd, Tessa Bailey, Stavroula Leka and MikaelaOwen.

References

Arts, W., & Gelissen, J. (2002). Three worlds of welfare capitalism or more? A state ofthe art report. Journal of European Social Policy, 12, 137e158.

Benach, J., Muntaner, C., & Santana, V. (2007). Employment conditions and healthinequalities. Employment Conditions Knowledge Network. Final Report of WHOCommission on Social Determinants of Health. Available from: http://www.who.int/social_determinants/themes/employmentconditions/en/. Retrieved26.07.2011.

Black, C. (2008). Review of the health of Britain’s working age population: working fora healthier tomorrow. London: TSO.

Bond, S. A., Tuckey, M. R., & Dollard, M. F. (2010). Psychosocial safety climate,workplace bullying, and symptoms of posttraumatic stress. OrganizationDevelopment Journal, 28, 37e56.

CIA World Fact Book 2011, www.cia.gov/library/publications/the-world-factbook.Commission on Social Determinants of Health. (2008). Closing the gap in a gener-

ation: health equity through action on the social determinants of health. FinalReport of the Commission on Social Determinants of Health. Geneva: WorldHealth Organization.

Coburn, D. (2004). Beyond the income inequality hypothesis: class, neo-liberalism,and health inequalities. Social Science & Medicine, 58, 41e56.

Cortina, L. M., Magley, V. J., Williams, J., & Langhout, R. (2001). Incivility in theworkplace: incidence and impact. Journal of Occupational Health Psychology, 1,64e80.

Cox, T., Griffiths, A., & Rial-Gonzalez, E. (2000). Research on work-related stress.Luxembourg: Office for Official Publications of the European Communities,European Agency for Safety and Health at Work.

Dollard, M. F. (2012). Psychosocial safety climate: a lead indicator of workplacepsychological health and engagement and a precursor to intervention success.In C. Biron, M. Karanika-Murray, & C. Cooper (Eds.), Improving organizationalinterventions for stress and well-being interventions: addressing process andcontext (pp. 77e101). London: Routledge.

Dollard, M. F., & Bakker, A. B. (2010). Psychosocial safety climate as a precursorto conducive work environments, psychological health problems, andemployee engagement. Journal of Occupational and Organisational Psychology,83, 579e599.

Dollard, M. F., & Karasek, R. (2010). Building psychosocial safety climate: evaluationof a socially coordinated PAR risk management stress prevention study. InJ. Houdmont, & S. Leka (Eds.), Contemporary occupational health psychology:Global perspectives on research and practice (pp. 208e234) Chichester: WileyBlackwell.

Dollard, M. F., Osborne, K., & Manning, I. (2012a). Organizationeenvironmentadaptation: a macro-level shift in modeling work distress and morale. Journal ofOrganizational Behavior, http://dx.doi.org/10.1002/job.1821.

Dollard, M. F., Opie, T., Lenthall, S., Wakerman, J., Knight, S., Dunn, S., et al. (2012b).Psychosocial safety climate as an antecedent to work characteristics and psy-chological strain: a multilevel model. Work & Stress, 26, 385e404.

Dollard, M. F., Tuckey, M. R., & Dormann, C. (2012c). Psychosocial safety climatemoderates the demand-resource interaction in predicting work stress. AccidentAnalysis and Prevention, 45, 694e704.

Dragano, N., Siegrist, J., & Wahrendorf, M. (2011). Welfare regimes, labour policiesand unhealthy psychosocial working conditions: a comparative study with 9917older employees from 12 European countries. Journal of Epidemiology andCommunity Health, 65, 793e799.

Esping-Andersen, G. (1990). The three worlds of welfare capitalism. Oxford: Polity Press.Esping-Andersen, G. (1999). Social foundations of postindustrial economies. Oxford:

Oxford University Press.Eurofound. (2011a). Industrial relations context. Eurofound. www.eurofound.europa.

eu/areas/industrialrelations/dictionary/dictionary0.htm.Eurofound. (2011b). Croatia: industrial relations profile. Dublin: Eurofound. www.

eurofound.europa.eu/eiro/country/croatia.pdf.European Agency for Safety and Health at Work. (2009). OSH in figures: stress at

work e facts and figures. Luxembourg: Office for Official Publications of theEuropean Communities. http://osha.europa.eu/en/publications/reports/TE-81-08-478-EN-C_OSH_in_figures_stress_at_work.

European Foundation for the Improvement of Living and Working Conditions.(2006). Fourth European Working Conditions Survey. Luxembourg: Office forOfficial Publications of the European Communities. http://www.eurofound.europa.eu/ewco/surveys/EWCS2005/index.htm.

European Social Survey 2010. Edition 1.0 http://www.europeansocialsurvey.org/index.php.

European Working Conditions Survey (EWCS). (2010). European Working ConditionsObservatory. www.eurofound.europa.eu/surveys/ewcs/2010/.

European Survey on New and Emerging Risks e Psychosocial Risks (ESENER).(2009a). UK Data Archive Study Number 6446, Management Questionnaire. www.esds.ac.uk/findingData/snDescription.asp?sn¼6446&;key¼esener.

European Survey on New and Emerging Risks e Psychosocial Risks (ESENER).(2009b). UK Data Archive Study Number 6446. www.esds.ac.uk/findingData/snDescription.asp?sn¼6446&;key¼esener.

Giga, S. I., Hoel, H., & Lewis, D. (2008). The costs of workplace bullying. www.unitetheunion.org/Docs/Research - cost of bullying.doc Accessed 9.03.2012.

Hall, G. B., Dollard, M. F., & Coward, J. (2010). Psychosocial safety climate: devel-opment of the PSC-12. International Journal of Stress Management, 4, 353e383.

Hall-Jones, P. (2010). Unionism and economic performance. http://www.newunionism.net/library/member%20contributions/news/Unionism%20and%20Economic%20Performance.htm.

Harter, J. K., Schmidt, F. L., & Hayes, T. L. (2002). Business-unit-level relationshipbetween employee satisfaction, employee engagement, and business out-comes: a meta-analysis. Journal of Applied Psychology, 87, 268e279.

Health and Safety Executive. (2012). What are the Management Standards. http://www.hse.gov.uk/stress/standards.

Hu, L., & Bentler, P. (1999). Cutoff criteria for fit indexes in covariance structure analysis:conventionalcriteriaversusnewalternatives.StructuralEquationModelling,6,1e55.

Idris, M. A., & Dollard, M. F. (2011). Psychosocial safety climate, work conditions, andemotions in the workplace: a Malaysian population-based work stress study.International Journal of Stress Management, 18, 324e347.

Idris, M. A., Dollard, M. F., Coward, J., & Dormann, C. (2012). Psychosocial safetyclimate: conceptual distinctiveness and effect on job demands and workerpsychological health. Safety Science, 50, 19e28.

Kang, S. Y., Staniford, A., Dollard, M. F., & Kompier, M. (2008). Knowledge devel-opment and content in occupational health psychology: a systematic analysisof the Journal of Occupational Health Psychology, and Work & Stress, 1996e2006. In J. Houdmont, & S. Leka (Eds.), Occupational health psychology: Euro-pean perspectives on research, education and practice, Vol. 3 (pp. 27e63). Maia,Portugal: ISMAI Publishers.

Karasek, R. A., & Theorell, T. (1990). Healthy work: stress, productivity and thereconstruction of working life. New York: Basic Books.

Kivimäki, M. J., Nyberg, S. T., Batty, G. D., Fransson, E. I., Hiekkila, K., Alfredsson, L.,et al. (2012). Job strain as a risk factor for coronary heart disease: a collaborativemeta-analysis of individual participant data. The Lancet1e7. online.

Law, R., Dollard, M. F., Tuckey, M. R., & Dormann, C. (2011). Psychosocial safetyclimate as a lead indicator of workplace bullying and harassment, job resources,psychological health and employee engagement. Accident Analysis and Preven-tion, 43, 1782e1793.

Leka, S., & Jain, A. (2010). Health impact of psychosocial hazards at work: an overview.Geneva: World Health Organization.

Lynch, J., Davey Smith, G., Hillemeier, M., Shaw, M., Raghunathan, T., & Kaplan, G.(2001). Income inequality, the psychosocial environment, and health: com-parison of wealthy nations. Lancet, 358, 194e200.

Lynch, J., Davey Smith, G., Harper, S., Hillemeier, M., Ross, N., Kaplan, G. A., et al.(2004). Is income inequality a determinant of population health? Part 1. Asystematic review. The Milbank Quarterly, 82.

Marmot, M., & Wilkinson, R. G. (2001). Psychosocial and material pathways in therelation between income and health: a response to Lynch et al. British MedicalJournal, 322, 1233e1236.

McTernan, W. P., Dollard, M. F., & LaMontagne, A. D. (2013). Clinical and sub-clinical depression in Australian workplaces: an economic cost analysis ofdepression-related productivity loss attributable to job-strain and bullying.Work & Stress.

M.F. Dollard, D.Y. Neser / Social Science & Medicine 92 (2013) 114e123 123

Murray, C. J. L., & Lopez, A. D. (1996). In The global burden of disease. A comprehensiveassessment of mortality and disability from diseases, injuries and risk factors in1990 and projected to 2020. Cambridge, MA: Harvard School of Public Health.

Murray, C. J. L., Tandon, A., Salomon, J. A., Mathers, C. D., & Sadana, R. (2002). Cross-pop-ulation comparability of evidence for health policy evidence for health policy. GlobalProgramme on Evidence for Health Policy Discussion Paper No. 46. World HealthOrganization http://www.who.int/healthinfo/paper46.pdf. Downloaded 12.03.2012.

Navarro, V., Muntaner, C., Borrell, C., Benach, J., Quiroga, A., Rodriguez-Sanz, M.,et al. (2006). Politics and health outcomes. Lancet, 368, 1033e1037.

OECD. (2009). Statistics on Trade Union Density. Paris, France: OECD. http://stats.oecd.org/Index.aspx?DataSetCode¼UN_DEN.

Penn World Tables 7.0 http://pwt.econ.upenn.edu/php_site/pwt_index.php.Prüss-Üstün, A., & Corvalán, C. (2006). Preventing disease through healthy environments.

Towards an estimate of the environmental burden of disease. Geneva: World HealthOrganization.

Rau, R. (2004). Job strain or healthy work: a question of task design. Journal ofOccupational Health Psychology, 9, 322e338.

Schermelleh-Engel, K., Moosbrugger, H., & Muller, H. (2003). Evaluating the fit ofstructural equation models: tests of significance and descriptive goodness-of-fitmeasures. Methods of Psychological Research Online, 8, 23e74.

Shively, W. P. (1969). "Ecological" inference: the use of aggregate data to studyindividuals. The American Political Science Review, 63, 1183e1196.

Health 2020: vision, values, main directions and approaches; The new European policyfor health. (2011). Regional Committee for Europe, World Health Organization

Regional Office for Europe. http://www.euro.who.int/__data/assets/pdf_file/0007/147724/wd09E_Health2020_111332.pdf Accessed 27.03.2012.

United Nations. (2011). World Population Prospects, the 2010 Revision. New York:United Nations.

Volger-Ludwig, K. (2009). Monitoring the duration of active working lifein the European Union final report, European Commission, Employment, SocialAffairs and Equal Opportunities. http://ec.europa.eu/social/main.jsp?langId¼en&catId¼89&newsId¼652&furtherNews¼yes. Downloaded 9.03.2012.

Van Stolk, C., Staetsky, L., Hassan, E., & Woo Kim, Chong (2012). Management ofpsychosocial risks at work: an analysis of the findings of the European Survey ofEnterprises on New and Emerging Risks (ESENER). In European Agency for Safetyand Health at Work: . Luxembourg: Publications Office of the European Union.

Wilkinson, R. (1997). Health inequalities: relative or absolute material standards?British Medical Journal, 314, 591e595.

Weil, D. (1999). Are mandated health and safety committees substitutes foror supplements to labor unions? Industrial and Labor Relations Review, 52,339e360.

Wing Sue, D. (2010). Microaggressions in everyday life: race, gender, and sexualorientation. New Jersey: John Wiley & Sons Inc.

Xanthopoulou, D., Bakker, A. B., Demerouti, E., & Schaufeli, W. B. (2009). Workengagement and financial returns: a diary study on the role of job and per-sonal resources. Journal of Occupational and Organisational Psychology, 82,183e200.