good jobs, good pay, good health? - university of sheffield · involve physiological arousal that...
TRANSCRIPT
1
Good jobs, good pay, good health? The direct effects of job quality on health dynamics among older European workers
Golo Henseke LLAKES Research Centre, UCL Institute of Education,
First Draft, This version: 15/06/2015
Abstract:
Longer working lives require sustainable jobs. By applying an amended methodology from Adams et
al. (2003, 2004) to a dataset with longitudinal information on work and health for older people in 15
continental European countries, I provide new evidence on differential effects of job quality on a
range of physical and mental health outcomes. Job quality indicators are grounded in a recently
proposed multi-disciplinary approach that combines financial and non-financial aspects into a
comprehensive concept. To better control for section into jobs I utilise data on childhood
circumstances and healthiness – a novelty in this literature strand. My findings suggest significant
effects of intrinsic job quality on the occurrence of musculoskeletal disorders, incident depressive
symptoms, and the onset of poor health. A cross-national comparison reveals the most limited
health effects in Southern Europe and the widest range of job quality effects on health in
Scandinavia. Potential policy initiates to improve population health through intrinsic job quality will
need to take the existing institutional context into account.
Keywords: job quality, older workers, health dynamics, childhood circumstances, cross-national
comparison
JEL: I14, J81, C22
1 Introduction Despite improving morbidity and mortality, an almost universal access to health care, health and
safety regulations, and an overall decline in accidents at work, around a quarter of Europeans
believe their jobs to be pathogenic (Barnay 2014). The total costs from major work-related health
disorders are estimated to amount to about 4.5% of total GDP in the EU (EU-OSHA 2014). Extended
working lives, growing wage inequality, a largely unbroken trend to more intense work (Green et al.
2013), and increasingly common precarious employment relations (Standing 2011) have an impact
on job quality and potentially add to work-related health hazards. Almost 30% of older workers in
low quality jobs suffered from fair or poor health, compared to only 12% in high quality jobs in
Continental Europe (Figure 1). Impaired health is a main reasons for early retirement. In this study, I
analyse how job quality contributes to health inequalities in a general population of people aged 50+
in European countries. A better understanding of the potential health hazards of work will help to
inform on sustainable jobs to extend working lives in times of ageing populations in the different
policy contexts across Europe.
There is a rich body of multi-disciplinary research that explores the influence of work and working
conditions on health and well-being (Barnay 2014; Bassanini & Caroli 2014). Hazards at work,
material well-being, and psychosocial facets of job quality correlate with a range of stress-related
2
physical and mental health disorders. This affects workers of every age, but evidence suggest
growing vulnerabilities to work-related health hazards over the life-course (M. K. Jones et al. 2013).
Emerging health inequalities are, however, moderated and shaped by the wider national policy
context (Bambra et al. 2014).
Figure 1: Self-reported health by intrinsic job quality amongst older European workers, 2004-2013
Base: Employed labour force aged between 50-74 years Austria, Belgium, Czech Republic, Denmark, Estonia, France,
Germany, Italy, Netherlands, Poland, Slovenia, Spain, Sweden, and Switzerland. Source: SHARE W1, W2, W4, W5, own
calculations
Despites decades of research in this field there is debate as to what extent the observed associations
reflect causal mechanisms. Workers do not randomly select into jobs but chose them based on their
skills, health, and tastes. The labour market positioning at any given time is the result of decisions
over the career, during childhood and in adolescence. A growing economic research strand has
started to deal with these selection mechanisms in order to identify the direct effects of work on
health in greater detail. However, this literature has so far looked mainly at general health
outcomes.
By analysing the effect of job quality on the health dynamics over time in a cross-national population
of older workers, my work makes several contributions to the growing body of economic research in
this area. Firstly, I utilise a modern, multi-disciplinary concept of job quality to analyse the different
pathways through which work can affect the health of older workers. Going beyond psychosocial job
quality allows me to analyse the effect of related, but often ignored, dimensions of job quality such
as earnings and job security. Secondly, unlike most previous work, I consider detailed health
outcomes covering both physical health, mental health disorders and functional disabilities rather
than a summary indicator of health or one specific disorder to better trace comorbidities. And
finally, I adopt a longitudinal model of the incidence of new health conditions that together with rich
0.280.27
0.32
0.24
0.40
0.15
0.48
0.12
0.1
.2.3
.4.5
Self-R
ep
ort
ed
Hea
lth
1st Quartile 2nd Quartile 3rd Quartile 4th Quartile
Excellent/ Very Good Fair/ Poor
3
individual background information including data on childhood circumstances, can help to shed light
on the health effects of good and bad work.
Drawing on data from the Survey of Health and Retirement in Europe (SHARE), I can follow the life
situations of people aged 50 and older and their partners between 2004 and 2012 across 15
European countries. The survey provides not only biannually collected longitudinal data on health
and work, but also unique retrospective information on childhood circumstances and healthiness.
In all, my results suggest that causal effects on job quality cannot be ruled out. Psychosocial job
quality predicts the occurrence of musculoskeletal disorders, incident depressive symptoms and the
onset of bad health. Women seem more physically susceptible to poor intrinsic job quality, whilst
men struggle more with mental health outcomes. Job security and earnings do not predict health
innovations in the pooled data. The estimates suggests cross-national differences in the subsequent
health inequalities by working conditions. Distinguishing by type of welfare states reveals that
differences in job quality have the smallest direct health effects in southern European countries and
the most diverse direct impacts in Scandinavia. A finding that clearly contrasts with the established
cross-country patterns in income and skills inequality. Improving the intrinsic job quality by one
standard deviation is predicted to reduce the onset of work-related health outcomes. The estimated
average effects are similar to giving up a risky health behaviour trait in the pooled sample. But the
there are differences across the existing policy context – a point specifically relevant for EU wide
policy initiatives.
The remainder of this study starts with a review of the existing literature on working conditions and
health. Section three describes the conceptual model and the empirical implementation. The data
and variables are introduced in section four. Section five summarises the findings from the empirical
analysis, followed by discussion and conclusions in section 6.
2 Previous Literature Long-term exposure to adverse working conditions is thought to affect health mainly through
continued physiological stress responses. Even relatively minor environmental and intrinsic stressors
involve physiological arousal that leads to a wear and tear on the body, called allostatic load
(McEwen & Seeman 1999). Stressors are not restricted to jobs strains, but can also include stress
due to material deprivation, or hazardous environments. In the short-term, these bodily response
help to cope with stress, but when exposed to repeated or persistent stressors, the accumulation of
allostatic load causes dysregulation in the body and leads to health problems. High blood pressure
develops into hypertension and eventually coronary heart disease. Repeated muscle strain develops
into chronic pain (Gruenewald et al. 2012; Karlamangla et al. 2002; Seeman et al. 2001; Steptoe &
Kivimäki 2012). As consequence, chronic stress lowers the body’s resources to adapt and eventually
increases frailty and the risk of diseases and disorders (Seeman et al. 2014; Ganster & Rosen 2013).
Research on health effects of working conditions is commonly grounded in psychosocial theories of
work-related stress. Seminal conceptual contributions include Karasek (1979), Johnson et al. (1989),
and Siegrist (1996). Situations of job strain (combination of high demands with low control), iso-
strain (job strain in conjunction with a lack of social support) or effort-reward imbalance (lack of
reciprocity between work effort and received reward) are theorized to impact health and subjective
well-being. Empirical findings, mostly from epidemiological, sociological and psychological research,
largely confirm this conclusion for a range of health outcomes such as cardiovascular disease,
musculoskeletal disorders, gastrointestinal problems, disrupted immune response, body mass, or
psychological well-being including fatigue, depressive symptoms and emotional exhaustion (Backé et
4
al. 2012; Häusser et al. 2010; Kivimäki et al. 2012; Niedhammer et al. 2014; Nieuwenhuijsen et al.
2010; Nixon et al. 2011; Vanroelen et al. 2009; Schütte et al. 2014).
Older workers are found to be more susceptibility to job-related stress (Fletcher et al. 2011 in US
data; Davies et al. 2014 for UK data; M. K. Jones et al. 2013 for European data; Ravesteijn et al. 2013
for Germand data; for a general review Bohle et al. 2009). When exposed to job-strain, older
workers face among others alleviated risks of developing depressive symptoms, disabilities, or
increasing frailty (Reinhardt et al. 2013; Siegrist et al. 2012; Kalousova & Mendes de Leon 2015).
Ravesteijn et al (2013) calculates that an increase in physical jobs demands by one standard
deviation over the ages 60-64 is equivalent to ageing 14 additional months, similarly working in a job
with low-control has equivalent health effects as ageing 2 years in German data. Tentative evidence
suggests differential dynamics of health by job type: cumulative exposure to adverse working
conditions seems to wear out people faster. Health inequalities thus widen over the working life
(Case & Deaton 2005; Gueorguieva et al. 2009; Kjellsson 2013).
More recently, job insecurity (Sverke et al. 2006) and precarious work (Benach et al. 2014; Benach &
Muntaner 2007) have attracted the attention of scholars as further sources of work-related stress.
Flexible work arrangements have progressively become more common in developed countries and
are advocated, in the form of downshifting, bridge jobs or gradual retirement, as key to extend
working lives beyond current normal retirement ages (Barnay 2014; Van Vuuren 2014). Downshifting
is thought to reduce job demands and help people to achieve a good work-life balance (Kennedy et
al. 2013). But older workers, particularly women, job-seekers and people from lower socio-economic
background, risk instead to move into precarious job roles that constitute own health hazards
(Benach & Muntaner 2007; Bohle et al. 2009; Loretto & Vickerstaff 2015). The adverse consequences
on physiological health and mental well-being of chronic job insecurity and precarious employment
are well documented (Caroli & Godard 2013; Green 2011; Llena-Nozal 2009; Virtanen et al. 2013;
Aerden et al. 2015).
Major differences in the quality of health care systems, labour market institutions, and work
organisation across European countries shape the wider context of work with potential impacts on
the degree of health inequalities. Existing empirical findings suggest that the more generous welfare
states in Scandinavian countries, for instance, correlate simultaneously with better population
health and less strenuous working conditions (Dragano et al. 2010; Lunau et al. 2015). A common
typology differentiates between Scandinavian, Corporatist, Liberal, Southern, and Eastern European
welfare states – ordered by their generosity and the level of social protection against adversaries
(e.g., Eikemo et al. 2008; Bambra 2011). But even the generous welfare states of the Nordic
countries with relatively small inequalities in income and education produce persistent, sizeable
health inequities (Brennenstuhl et al. 2012; Bergqvist et al. 2013; Mackenbach 2012). The few
existing studies on cross-country variations in health inequalities by job quality point to similar
conclusions (Gupta & Kristensen 2008; Bambra et al. 2014).
Building on (Bustillo et al. 2011) and the multi-disciplinary insights into the health-effect of working
conditions, Green and Mostafa (2012) and Green et al. (2013) propose a comprehensive concept of
objective job quality. Besides intrinsic, psychosocial aspects, their approach distinguishes
additionally between earnings, job prospects, and working time quality. Earnings capture the
material variation in job quality. Job prospects combine job security and prospects for future
development. Intrinsic job quality includes most of the commonly explored psychosocial stressors
and workplace hazards grouped into four subscales: skill use and discretion, social environment,
physical environment, and work intensity. Working time quality encapsulates job features that
contribute to a good work-life balance. The authors successfully validate their concept against a
5
range of health and subjective well-being outcomes in cross-sectional data. So far, the concept has
not been used yet to study the onset of health conditions.
Though decades of research have explored the link between job quality and health, the evidence on
the causal effects particularly in a general populations of older workers has remained scarce. Two
key selection mechanisms hamper the inference of work-related health effects from observational
data (Ravesteijn et al. 2013). Firstly, there is a strong socio-economic gradient with respect to job
quality and health (Macmillan & Smith 2007; Stowasser et al. 2012; Galama & van Kippersluis 2013).
Workers choose specific jobs depending on to their own abilities, educational achievement, health,
and, potentially, taste. As a consequence people form better off backgrounds tend to be healthier
and employed in higher quality jobs (Landsbergis et al. 2014; Moncada et al. 2010). In fact, job
quality differences explain a substantial fraction of the socioeconomic health inequalities over the
life course (Bauer et al. 2009; Kaikkonen et al. 2009).
Socioeconomic inequalities in health and employment over the life-course can be traced to
childhood conditions (Mazzonna 2014; Case et al. 2005; Smith 2009b). Socioeconomic circumstances
during childhood and potentially even in utero determine the level of acquired skills, healthiness and
education with subsequent effects on the labour market trajectories and the history of adverse
working conditions and labour market disadvantage (Dragano & Wahrendorf 2014; Jerrim 2014;
Macmillan et al. 2013). Working conditions in the first job after full-time education are indeed found
to impact health later in life (Fletcher 2012).
Further, older workers potentially downshift or become economically inactive in reaction to their
health status (García-Gómez 2011; Disney et al. 2006; García-Gómez et al. 2013; Bound et al. 2010).
This health related selection in labour market positioning introduces a feedback of reverse causation
– declining health can lead to lower payer, lower pay can contribute to diminished health. How these
selection mechanisms are dealt with in the empirical analysis, matters for the estimated health
effects of work (Davies et al. 2014; M. K. Jones et al. 2013).
Secondly, the theory of compensating wage differentials suggests a trade-off between pay and risky
working conditions. Economic theory suggest that higher pay is thought to compensate workers for
otherwise not insured job hazards (Pouliakas & Theodossiou 2013 for review). Workers could use
the earnings premium of high risk jobs to offset work-related health impairments (e.g., Case &
Deaton 2005) or, more generally, change health-related behaviour in response to job quality. A
sedentary job might stimulate to exercise after work. High levels of work-related stress could lead to
increased alcohol consumption or an unhealthy diet. Initial blue collar employment seems, for
instance, to lead to more risky health behaviour (Kelly et al. 2012).
Two basic econometric strategies are commonly applied to deal with selection bias: instrumental
variables or individual effects to controls for unobserved heterogeneity. Scholars have proposed a
range of external instrumental variables such as the number of health and safety regulations across
countries (Cottini 2012; Cottini & Lucifora 2013), industry and country-specific stringency of
employment protection (Caroli & Godard 2013), variations in the local unemployment rate at time
of job choice (Kelly et al. 2012), or organisational measures of high performances human resource
(M. K. Jones et al. 2013). But IV strategies, though promising, remain hard to realise in this context.
Good external instruments are rare, most seem to have insufficient statistical power, and even if
they instruments strongly correlated with working conditions, effects would be identified for local,
not clearly, defined populations. Cottini and Lucifora (2013) and Jones et al. (2013), for instance, can
report a larger effect of working conditions on health outcome using IV, whereas the point estimates
in the remaining IV-studies decline compared to the baseline findings.
6
Models with time invariant individual effects have been the method of choice for most scholars to
achieve consistent estimates. Dynamic panel data estimator with lagged dependent variable as
additional regressor to account for the persistency in health are commonly applied in the literature
(Butterworth et al. 2011; Fletcher et al. 2011; Gupta & Kristensen 2008; Llena-Nozal 2009; Robone et
al. 2011). Alternatively, fixed effects specifications or models in first differences are used to cancel
out time invariant individual effects (Ravesteijn et al. 2013; Green 2011; Llena-Nozal 2009). Fletcher
(2012) uses sibling fixed effects to control for common family background characteristics.
Adjusting for selection and state-dependent health changes the estimated effect sizes substantially.
Compared to baseline estimations without adjustments coefficients on job quality drop and on
occasion become insignificant. Clearly, selection and persistency in health explain most of the
association between job quality and health. However, in most of the existing studies health
dynamics are restricted to a first-order Markov process. The influence of longer lags is ignored.
Further information on initial conditions is usually gleaned from data at survey baseline. A practice
that has received criticism (Smith 2009a).
For the current study, I follow in the footsteps of the latter literature strand but add pre-labour
market information on healthiness and abilities. I adopt an identification design that was initially
developed to detect causal non-effect of SES on health (Adams et al. 2003; Adams et al. 2004). The
highly cited study, henceforth HWWA, applied a Granger causality framework to a longitudinal
survey of elderly American to test for the absence of direct SES impacts on the onset of a range of
health outcomes conditional on previous health. HWWA main advances were the exploration of
health innovations rather than levels and the insight that non-causal effects in a statistical
framework might imply absence of “true” causality. The authors propose parameter invariance as a
necessary condition for causal processes.
HWWA stimulated a huge controversy. SES was found to have no impact on the incidence of most
health outcomes apart from mental health conditions. Most criticism was levied against the
treatment (or lack thereof) of unobserved heterogeneity, the test of non-causality including the
proposed condition of time, and the assumption that health dynamics can be satisfyingly described
by a first-order Markov process.
More recently Stowasser et al. (2014; 2012) have revisited the approach. By adding childhood
information to the model and allowing for longer health lags, Stowasser et al. (2014) tackle two
areas of criticism: lack of rich health dynamics and omission of individual effects that potentially
confound the relation between SES and health. Their methodology provides a valuable starting point
for my analysis of job quality effects on health.
3 Methodology Drawing from the existing literature, I assess the effects of job quality on new health events
conditional on previous health and individual effects that potentially correlate with health
outcomes, labour supply and job quality. In doing so, I can account for at least part of the selection
bias that confounds estimations of job quality effects on health. Further by analysing the effect of
job-quality on health across several European countries I can test for parameter invariance under
several different policy regimes. The latter test might provide valuable insights into the (non-
)existence of cross-country differences in job-quality effects on health. It can also be informative
about the role of public policy as cause of health inequalities in a mature population.
Adapting the methodology in HWWA and the related literature, I specify the following dynamic
model of health incidences:
7
𝑓(𝐻𝐼𝑖𝑡𝑗
|𝐻𝐼𝑖𝑡𝑘<𝑗
, 𝐻𝑖𝑡−1, 𝐽𝑄𝑖𝑡−1, 𝑋𝑖𝑡−1, 𝜗𝑖)
The model describes the incidence of health condition 𝑗 for individual 𝑖 conditional on past job
quality (𝐽𝑄𝑖𝑡−1), the instantaneous effects from concurrent health shocks 𝐻𝐼𝑖𝑡𝑘<𝑗
, past health
(𝐻𝑖𝑡−1), a set of socio-demographic variables (𝑋𝑖𝑡−1), and a time constant individual effect 𝜗𝑖. More
specifically:
JOB QUALITY: 𝐽𝑄𝑖𝑡−1
Following Green and Mostafa (2012), I consider earnings, intrinsic job quality, and job security at 𝑡 −
1 as indicators of overall job quality in 𝑡 − 1. These are the variables of interest. I expect that good
jobs protect people’s health and lower the risk to develop medical conditions. Additionally I include
information on job tenure and working hours which might affect health transitions and job quality.
PAST HEALTH, 𝐻𝑖𝑡−1
Health is highly persistence over time (Contoyannis et al. 2004; Jones et al. 2012). Including past
health in the empirical model controls for the state dependency (e.g. depressive symptoms in the
past make it more likely to develop depressive symptoms in the future) and comorbidities between
health conditions (e.g. increased risk of heart attack with history of hypertension; relation between
physical and mental health). The vector of past health conditions includes health behaviour,
measures of physical as well as mental conditions, and information on general health and functional
disabilities.
CONCURRENT HEALTH INNOVATIONS, 𝐻𝐼𝑖𝑡𝑘<𝑗
The onset of health conditions could result from other concurrent health shocks. To capture the
effect of contemporaneous health shock, I impose a unidirectional instantaneous causal chain of
health incidences flowing from potentially life-threatening acute conditions, to chronic
cardiovascular conditions, to musculoskeletal disorder, to depressive symptoms, functional
disabilities, health-related behaviour, and, finally, self-assessed health. The assumed causal chain
follows propositions in HWWA. Adams et al. (2004) report validity tests.
SOCIO-DEMOGRAPHIC COVARIATES, 𝑋𝑖𝑡−1
A vector of time varying and time constant covariates with potential influences on health and job
quality. It includes age, age square, a dummy variable to indicate whether he respondent has
reached the early retirement age, gender, indicator if born abroad, current marital status, cognitive
abilities, and country of residence.
INDIVIDUAL EFFECTS: CHILDHOOD AND FAMILY BACKGROUND
Let the individual effect 𝜗𝑖 be a representation of innate healthiness, behaviour, and abilities with
effects on both the labour market positioning and the trajectory of health. To approximate the
variable, I combine retrospective childhood information on healthiness, parental socioeconomic
status, educational attainment and academic performance. Further I include information on
premature parental mortality to capture potential genetic risk, the respondent’s smoking history to
control for past health-related behaviour and height which has been shown to reflect healthiness
and to correlate with labour market outcomes. In addition to dealing with individual heterogeneity,
measures of childhood health also enrich the possible dynamics in health transitions.
8
The proposed dynamic model is comparable to Grossman’s specification of health as durable capital
good (Grossman 1972; Grossman 2000). Within this human capital framework, job quality can be
interpreted as a health (dis-)investment whose level is chosen by the individual to maximise life-
cycle utility (Fletcher et al. 2011; e.g., Case & Deaton 2005).
The health innovations are assumed to follow a Poisson distribution.
𝐻𝐼𝑖𝑡𝑗
= 𝑒ln(𝐸𝑡−1,𝑡)+𝛽0+∑ 𝐻𝐼𝑖𝑡𝑘𝑘<𝑗
𝑘=1 𝛼1𝑘+𝐻𝑖𝑡−1
′ 𝛼2+𝐽𝑄𝑖𝑡−1′ 𝛼3+𝑋𝑖𝑡−1
′ 𝛼4+𝜗𝑖𝛼5+𝑢𝑖𝑡
For chronic conditions, e.g. hypertension, diabetes or disability, 𝐻𝐼𝑖𝑡𝑗
indicates the onset, whereas for
acute conditions, e.g. heart attack or stroke, it indicates a new occurrence. The events are either
binary, e.g. occurrence of depression, or a count of multiple events, e.g., onset of additional
cardiovascular risk factors. The models are estimated using Poisson regression. In conjuncture with
the length of exposure, 𝐸𝑡−1,𝑡, measured by the duration between survey interviews, Poisson
regression is equivalent to a proportional hazard survival model. Because strict exogeneity is unlikely
to hold, e.g. downshifting in the follow-up period in response to an acute health incidence, I pool the
data and do not use a dedicated panel estimator.
To test for potential heterogeneous effects across time and countries, I include two sets of
interaction terms between lagged job quality and, firstly, country dummies and, secondly, period
dummies. Wald tests of joint significance are used to test against the null hypothesis of
homogenous parameters.
Despite best efforts it is conceivable that unobserved factors, such as taste and inter-temporal
preference correlate with the variables of interest and health dynamics. Further, with longer time
periods between observations the assumed causal chain of health innovations becomes more
restrictive and ignored job quality shocks could start to manifest in health differences. To deal with
potential model misspecification, violations of the Poisson process, and serial correlation in the
outcomes within individuals, I estimate the regressions using a Huber/White/Sandwich variance
estimator. Significant coefficients on lagged job quality measures might best be interpreted to
predict but not “structurally” cause the onset of new health conditions.
4 Data, variables and summary statistics
4.1 Dataset I use data from the Survey of Health, Ageing and Retirement in Europe. SHARE is a large longitudinal
representative probability sample that provides comprehensive and cross-nationally comparable
data on health, socioeconomic characteristics, and the labour market status including job quality
instruments of people aged 50 years and older and their partners (Börsch-Supan et al. 2005; Börsch-
Supan et al. 2013). SHARE was launched in 2004 and has been conducted in a biannual rhythm. To
date, the survey has collected information on more than 110,000 individuals from 20 European
countries and Israel over five waves. The first wave was fielded in 2004 in 11 northern, central and
southern European countries. For the second wave in 2006/2007, Poland and the Czech Republic
joined the survey. The third wave, SHARELIFE, supplemented the concurrent information from the
previous waves with retrospective life history data. The fifth and most recent wave covers the yeas
2012/2013. It includes along concurrent information also key retrospective data on childhood
circumstances. For more information on the data collection and survey design see the technical
reports (Börsch-Supan & Jürges 2005; Börsch-Supan et al. 2008; Schröder 2011; Malter & Börsch-
Supan 2013; Malter & Borsch-Supan 2015).
9
The analysis requires longitudinal information on a wide range of covariates for people who were at
work at baseline. I make use of data from all five available waves for people aged 50-74 years. To
study health transition conditional, amongst others, on childhood information, I restrict the sample
to countries that have contributed to at least two consecutive waves and participated in waves three
or five. Wave three (SHARELIFE) contains details life history data, but only a very limited selection of
concurrent health measures and is therefore not directly available for the analysis of health
transitions. In total, my sample comprises of about 19,700 observations from roughly 14,100
different respondents in 14 countries with complete information on all variables. The countries
group into four distinct welfare state regimes: Denmark and Sweden belong to the Scandinavian
cluster, the central and western European countries Austria, Belgium, France, Germany,
Netherlands, and Switzerland represent Corporatist welfare states, Greece, Italy, and Spain are
examples of Southern European welfare states, and the Czech Republic, Estonia, Poland, and
Slovenia are classified as Eastern European welfare states. The exact sample sizes will differ across
regression models.
The average length between interviews was 2.8 years including the gap between waves 2 and 4. I’ll
conduct invariance tests to establish whether the estimated effect sizes shift between waves.
4.2 Variables
4.2.1 Job Quality SHARE includes measure of psychosocial working conditions and job security from the Job Content
Questionnaire (Karasek et al. 1998) and the effort-reward imbalance questionnaire (Siegrist et al.
2004), employment contract, pay and occupations. Following the concept outlined in Green et al
(2013) and Green and Mostafa (2012), I group the items into three key aspects of job quality:
intrinsic job quality, job security, and earnings. Scores are calculated in the complete pooled cross-
sectional dataset (N = 43,500) to guarantee sufficient observations in each category. Once derived,
the intrinsic job quality and job security scales are z-standardized.
Intrinsic job quality combines categorical items on skills and discretion with social support, physical
environment and work intensity. The included items cover the ISCO skill level, opportunity to
develop new skills, discretion over how to do the work, support in difficult situations, recognition of
work, adequate salary, job prospects, physical work demands, and time pressure. The psychosocial
measures are coded on Likert-scales. ISCO skill levels distinguish between four ordinal ranked groups
(academic, tertiary non-academic, upper secondary, lower secondary or below) which map to the
occupational major groups. I use multiple correspondence analysis (MCA) to derive the item weights
needed to calculate a scale of intrinsic job quality. MCA can be viewed as a generalization of
principal component analysis for categorical variables. The first two dimensions explain at least 82.5
% of the total inertia. After an inspection of the calculated weights, dimension one returns noise,
whilst dimension two retains the intrinsic job quality. The weights of the index follow plausible
patterns: opportunity for skill development, high skill level, recognition, social support and
particularly adequate salary contribute strongly to high levels of intrinsic job quality.
The index of job security is constructed from a Likert-item on perceived job insecurity and the
employment terms: permanent, temporary employment, or self-employment. Combing both
variables using MCA, explains at least at least 95.5 % of the total inertia in the first dimension. Again
the weights are intuitively plausible. Temporary employment diminishes job security. Employees in
secure, permanent position receive the highest index value.
SHARE contains a range of labour income questions, but the items suffer from non-response and
reporting error. Therefore, I use imputed values that are provided as data supplement to the survey
10
waves. SHARE uses a multiple imputation procedure drawing on longitudinal information, a set of
concurrent variables and unfolding brackets to predict a range of plausible annual monetary values.
In the analysis, I use the first plausible value of annual net income from employment or self-
employment. Variables are converted to average monthly earning using information on the number
of months worked in the job and transformed to German prices in 2005 by purchasing power parity
conversion rates. In total, average monthly net earnings were around EUR 1,712 in the sample (Table
1). Approximately 13% of the income values were imputed.
Table 1: Job Quality
Variable Obs. Mean Std. Dev. Min Max
Intrinsic Job Quality 19707 -0.071 0.939 -4.288 1.815
Job Prospects 19707 0.053 0.922 -3.734 0.858
Real Net Pay 19707 1712.07 2186.06 -324.42 62812.70
Imputed Pay 19707 0.132 0.338 0 1
The distribution of job quality facets show sensible patterns across individuals and countries (see
Table 2). People with tertiary education score highest in all three dimensions, whilst workers with
lower secondary education or below found themselves in less secure jobs of lower quality with less
pay. Similarly, average intrinsic job quality declines as we move from Scandinavian to Eastern
European welfare state regimes. Net pay and job security peak in Corporatist regimes. Perhaps the
result of protective on insider focused labour markets. Job quality was generally lowest in Eastern
Europe. Job quality in southern European countries was located between the Corporatist and
Eastern European welfare states. The low average value of real pay in Scandinavia is partly resulting
from the high prevalence of part-time jobs. But even if differences in hours worked are account for
average net pay remains highest in Corporatist countries.
Table 2: Distribution of job quality
Intrinsic Job Quality Job Prospects Real Net Pay
Educational Attainment
Lower Secondary -0.335 -0.066 EUR 1,187.66
Upper Secondary -0.164 0.028 EUR 1,587.18
Tertiary 0.275 0.188 EUR 2,321.10
Welfare State Regime
Scandinavian 0.225 0.079 EUR 1,243.24
Corporatist -0.039 0.127 EUR 1,984.68
Southern -0.203 -0.027 EUR 1,423.34
Eastern -0.275 -0.522 EUR 505.68
Total -0.071 0.053 EUR 1,712.07
4.2.2 Health Indicators The full set of health indicates is given in Table 3 and health interventions are summarised in Table 4.
Some of the constructed variables may require further clarification. Acute conditions is a summary
score of previous diagnosis of heart attacks and strokes. Cardiovascular risk factor combines
previous diagnosis of hypertension, high blood cholesterol and diabetes. The index of
musculoskeletal disorders measures mostly mobility limitations resulting from impaired functioning
11
of the musculoskeletal system. It combines information on joint pain with 10 items on physical
difficulties such as walking, sitting, climbing stairs, or carrying and lifting. In wave 5, the item on joint
pain was split up and more thoroughly explored at the expense of some intertemporal consistency.
Depressive symptoms are measured on the EURO-D scale from 12 self-reported items. EURO-D has
been validated as a cross-national scale of mental health in elderly populations (Copeland et al.
2004; Prince et al. 1999). Values of four and above are indicative of major clinical depressions
(Castro-Costa et al. 2007; Larraga et al. 2006). The summary index of risky health behaviour captures
whether a respondent engages either never or only occasionally in moderate physical activity,
consumes alcohol daily or almost daily, and currently smokes. Functional disabilities measure the
difficulties a person has with self-care (ADL) and household management (IADL).
The occurrence of acute, potentially life threatening diseases is inferred from a dedicated item
battery. In cases respondent died after a heart attack or stroke, the cause of death from available
end of life interviews is coded as an acute incidence. The constructed binary variable indicates
whether any acute incidence has occurred or not. Innovations in cardiovascular risk and
musculoskeletal disorder code the onset of any new condition and limitation, respectively. Incident
depression indicates the development of severe depressive symptoms. Innovation in risky health
behaviour captures the development of potentially unhealthy behavioural habits. The onset of
functional disability is one if a respondent developed at least one ADL or IADL between survey waves
and zero otherwise. Similarly, the occurrence of poor or fair health captures the transition from
good or better health into less than good health. All health measures are self-reported.
Table 3: Prevalence of Health Conditions among older European workers
Variable Obs. Mean Std. Dev. Min Max
Acute Conditions 19707 0.059 0.241 0 2
Cardiovascular Risk Factors 19707 0.481 0.695 0 3
Musculoskeletal Disorders 19707 1.091 1.453 0 10
EURO-D Score 19707 1.988 1.957 0 12
Obese (BMI>=30) 19707 0.150 0.357 0 1
Underweight (BMI<18.5) 19707 0.007 0.083 0 1
BMI Missing 19707 0.014 0.118 0 1
Risky Health Behaviour 19707 0.546 0.686 0 3
Functional Disabilities 19707 0.098 0.585 0 13
Self-Reported Health
Poor/ Fair 19707 0.188 0.391 0 1
Good 19707 0.440 0.496 0 1
Very Good/ Excellent 19707 0.372 0.483 0 1
Table 4: Health Innovations
Variable Obs. Mean Std. Dev. Min Max
Acute health incident 19707 0.018 0.132 0 1
Cardiovascular Risk 19707 0.181 0.451 0 3
Musculoskeletal Disorders 19707 0.540 1.054 0 10
Incident Depression 19707 0.094 0.292 0 1
Risky Behaviour 19707 0.111 0.322 0 2
Functional disability 19705 0.051 0.220 0 1
Onset poor/ fair health 19703 0.098 0.298 0 1
12
4.2.3 Childhood and Family Background, socio-demographics, and further job
characteristics The full list of included covariates is given in Table 5. The derived variables were constructed as
follows.
The score of cognitive ability combines self-repots on writing and reading proficiency with test data
on orientation, memory functioning, verbal fluency, and numeracy. The categorical variables are
aggregated using weights derived from multiple correspondence analysis. The resulting index clearly
correlates with broad educational categories (spearman rank correlation coefficient of .47). The
strength of the correlation is similar to values obtained from the OECD Survey of Adult Skills using
more comprehensive cognitive skills assessments.
Premature parental mortality is an average of potential years of life lost. The variables averages the
number of potential years of life lost due to mortality before the age of 70 across both parents. If a
parent is still alive or deceased after the age of 70, his or her contribution is 0. If a parent died at the
age of 55, his or her potential years of life lost is 15. The reference age of 70 is in line with current
OECD values (OECD 2011).
Information on health and the living situation during childhood are collected retrospectively. For the
purpose of the survey, childhood was defined to include ages 0 to 15. Childhood health can be
thought to captures initial healthiness. Previous research has confirmed the validity of the measures
(Havari & Mazzonna 2011). The health conditions index is an average over the occurrence of a range
of acute and chronic health conditions such as respiratory problems including asthma, chronic ear
problems, difficulties seeing with eyeglasses, migraines, or psychiatric problems.
Self-reported academic position at age 10 combines two binary items which are one if respondent
reported above average performance in mathematics and languages relative to others at the age of
10.
The adjusted total annual net household income is the available household income minus the
individual income from employment/ self-employment. It captures the effect of non-labour income
on health and labour supply decisions. I use the first plausible imputed value to derive the variable.
Early retirement eligibility ages are taken from the Mutual Information System on Social Protection
(MISSOC). I derive a dummy variable that distinguishes between people above and below the age
threshold. The variable is thought to capture differences in the propensity to leave the labour force
and thus in the length of exposure.
Table 5: Summary Statistics Covariates
Variable Obs. Mean Std. Dev. Min Max
Educational Level
Lower Secondary (ISCED0,1,2) 19707 0.270 0.444 0 1
Upper Secondary (ISCED3,4) 19707 0.412 0.492 0 1
Tertiary (ISCED5,6) 19707 0.317 0.465 0 1
Cognitive Ability Score 19707 0.377 0.750 -6.03 1.52
Premature Parental Mortality 19707 2.951 5.537 0 46.5
Adult Height (in cm) 19707 170.3 9.2 100 203
Ever smoked 19707 0.537 0.499 0 1
Self-Reported Health during childhood
13
Poor/ Fair/ Volatile 19707 0.084 0.277 0 1
Good 19707 0.274 0.446 0 1
Very Good/ Excellent 19707 0.643 0.479 0 1
Health Conditions Index 19707 0.017 0.038 0 0.4
Missed School 1+ Month 19707 0.133 0.339 0 1
Self-Assessed Academic Position (age=10) 19707 0.755 0.792 0 2
1+ Shelf of books (age=10) 19707 0.700 0.458 0 1
Rooms per household member (age=10) 19707 0.792 0.394 0 11.25
Job tenure 19707 20.9 13.0 0.5 62.5
Average working Hours
0-19 hrs 19707 0.124 0.330 0 1
20-34 hrs 19707 0.185 0.388 0 1
35-42 hrs 19707 0.402 0.490 0 1
42-54 hrs 19707 0.202 0.401 0 1
55+ hrs 19707 0.087 0.282 0 1
Adjusted Total Annual Net HH Income 19707 23205.06 40501.1 0 2930197
Imputation Flag - Household Income 19707 0.625 0.484 0 1
Marital Status
Cohabiting 19707 0.059 0.236 0 1
Separate/ Divorced 19707 0.118 0.323 0 1
Widowed 19707 0.746 0.435 0 1
Never Married 19707 0.077 0.266 0 1
Age 19707 56.7 4.4 50 74
Above Early Retirement Age 19707 0.455 0.498 0 1
Female 19707 0.452 0.498 0 1
Foreign-born 19707 0.074 0.261 0 1
5 Results
5.1 General
5.1.1 Health inequalities by job quality Table 6 reports the risk ratios for the prevalence of health conditions by job quality in the population
of older workers.
Table 6: Predicted Risk Ratios by Bad vs. Good Jobs
Health Outcome RR SE
Acute Conditions 1.1733 0.1009
Cardiovascular Risk Factors 1.1247 0.0347 ***
Musculoskeletal Disorders 1.3242 0.0304 ***
EURO-D Score 1.2473 0.0279 ***
Obese (BMI>=30) 1.3225 0.0646 ***
Risky Health Behaviour 1.0507 0.0293
Functional Disabilities 1.3607 0.1206 **
Poor/ Fair Self-Reported Health 1.2798 0.0517 ***
14
Note: * p<0.05, ** p<0.01, *** p<0.001. Risk ratio compare the predicted prevalence of health conditions among people in bad (1st
quartile of job quality indicators) vs. people in good jobs (3nd quartile of job quality indicators). Derived from sets of Poisson regressions
with age, age^2, gender, period and a full set of country dummies. Regression use the provided cross-sectional survey weights.
There are clear and apparent health inequalities by job type among older workers in Europe.
Workers in low quality jobs are significantly more likely to suffer from a range of health conditions
than people in high quality jobs. The notable exemptions are acute conditions and risky health
behaviour. These general differences in the prevalence of ill-health are well documented in the
related literature. The findings confirm the commonly acknowledged association between health
and working conditions, but how much of these inequalities can be actually attributed to job quality
differences is less clear.
5.1.2 Incidence of health conditions Table 7 summaries the estimated average marginal effects of job quality facets on the incidence
rates of health innovations conditional on the full set of covariates. Average monthly earnings are
broken down into five groups that roughly correspond to the income quintiles in the sample.
Table 7: Average marginal effects of job quality on the incidence of health conditions
Variable Acute Cardio muscle Mental Behaviour Disability Poor Health
Intrinsic Job Quality 0.0002 -0.0038 * -0.0199 *** -0.0042 *** 0.0015 -0.0004 -0.0054 ***
0.0005 0.00 0.0035 0.00 0.0012 0.00 0.0011
Job Security 0.0001 0.00 -0.0013 0.00 0.0007 0.00 -0.0014
0.0004 0.00 0.0032 0.00 0.0012 0.00 0.0011
Earnings
EUR <200 0.0006 0.00 -0.0134 0.00 -0.0041 0.00 -0.0053
0.0017 0.00 0.0117 0.00 0.0038 0.00 0.0041
EUR 200-950 -0.0008 0.00 -0.0122 0.00 -0.0041 0.00 -0.005
0.0017 0.00 0.0122 0.00 0.0041 0.00 0.0041
EUR 950-1500 (Ref.) (Ref.) (Ref.) (Ref.) (Ref.) (Ref.) (Ref.)
EUR 1500-2400 0.002 0.01 -0.0193 0.00 -0.0013 -0.01 -0.0063
0.0023 0.01 0.0138 0.00 0.0045 0.00 0.0052
EUR >=2400 0.0031 0.0042 -0.0228 0.0012 -0.0047 0.0027 -0.0108 *
0.0049 0.0067 0.0152 0.0048 0.0054 0.0048 0.0055
Time Invariance 0.417 0.809 0.00 0.727 0.895 0.45 0.447
Country Invariance 0.000 0.000 0.00 0.000 0.001 0.00 0.000
Joint Effect 0.821 0.153 0.00 0.001 0.704 0.46 0.000
N 20124 20035 20035 19983 19983 18726 16400
Note: * p<0.05, ** p<0.01, *** p<0.001. Average marginal effects derived from Poisson regressions using the full set of covariates.
Musculoskeletal disorders, incident depressive symptoms and poor health link mostly clearly to job
quality. Acute conditions, changes in risky health behaviour or the occurrence of functional
limitations in contrast are not directly predicted by job quality differences in the pooled sample. The
findings also suggest an effect of intrinsic job quality on the development of cardiovascular risk
factors but the influence is not strong enough to translate into a statistically significant combined
effect of job quality on this conditions.
Intrinsic job quality is the most active facet of job quality. An improvement of intrinsic job quality by
a standard deviation for 1000 people would have prevented the development of approximately four
new cardiovascular risk factors, around 20 musculoskeletal disorders, four incident depressions, and
five complaints about poor health within a year of exposure. The health effects are comparable to
15
the predicted consequences of 1000 people giving up an unhealthy behavioural trait (e.g., smoking,
physical inactivity, daily drinking)
Job security and earnings are on their own not associated with health innovations. Further
investigation shows that this not due to strong links between quality, pay and security. Even when
entered as single indicator security and pay largely fail to predict changes in health.
Gender differences exist, but are small and do not alter the main conclusions (see Table 9 in the
appendix). Intrinsic job quality emerges again as the most significant health-related dimension of
job quality. The findings suggest women are more susceptible to physical health impairments,
whereas men are more likely to struggle with mental health as a result of poor intrinsic job quality.
There are few significant health effects of earning, but the pattern is inconclusive. Job security does
again not directly predict health innovations. Further, there is still no evidence for job quality effects
on the onset of acute conditions, functional disabilities or changes in unfavourable health behaviour.
In all, allowing for differential effects of job quality by gender confirms the previous findings.
Health shocks might cause people to leave the survey before follow-up; in extreme cases, because of
death after a fatal heart attack or stroke. However even non-fatal health shocks might reduce the
propensity to participate in the survey. In that case, non-response bias might occur as the collected
data stems from more healthy people. This is a general problem for health-related inferences form
observational data and not specific to my data set. The rich list of included covariates in the
empirical model should account for systematic non-response, but potential bias could still be
present. Applying the methodology developed in Verbeek and Nijman (1992) and outlined in Jones
et al. (2013), I test for the presence of non-random attrition. The test results suggest systematic non-
response is not a systematic issues in my analysis. Acute health shocks, as expected, correlate
statistically significantly with response behaviour, but the tests do not systematically reject the null
hypotheses of attrition at random among the remaining outcomes. Table 10 in the appendix
summarises the p-values. The estimated (non-)effects can thus not be attributed to non-response
bias.
Overall, job quality clearly predicts the onset of further musculoskeletal disorders, incident
depression and the transition into poor health among older workers. If one believes my covariates
are sufficient to condition out selection, different paces of health risk accumulation across jobs can
account, at least in parts, for the observed inequalities in the prevalence of health conditions. Effects
slightly differ by gender: job quality predicts the onset of cardiovascular risk factors among women
but is unrelated with depression and the other way round for men. None of the remaining
conditions is directly related to job quality in the pooled sample. But health dynamics and the
development of comorbidities over time could lead to spill-overs from one condition to another. A
closer inspection of the estimations results suggests, for instance, that musculoskeletal disorders at
baseline predict the incidence of acute conditions and the development of functional disabilities at
follow-up. However, a prediction of long-term effects of job quality requires an etiologically more
fully defined model and is beyond the scope of this study.
But does non-effect of job quality on some health outcomes allow us to conclude the absence of
direct structural causal effects? No. Even though the estimations largely confirm time invariance,
there is still substantial cross-country heterogeneity in the health response to job quality. Even
outcomes that do not directly respond to job quality in the pooled sample, appear to react to work
characteristics in some of the included country. The question arises what country specificities drive
these unequal health effects of job quality. What public policies help to protect older workers from
the adverse consequences of poor job quality across Europe?
16
5.2 Country Differences in Health Inequalities
5.2.1 Welfare State Regimes Potential differences in health inequalities by welfare state types can provide first, tentative insights
into the role of public policies in moderating the risks from poor job quality. By replacing the country
dummies with a set of welfare state regime dummies in the estimation models, I can evaluate the
health effects of job quality across regimes. This multilevel fixed effects strategy accounts for
general, not work related differences in the onset of diseases across the regimes, whilst allowing for
differential effects of job quality on health. Table 8 displays the p-values of joint significance tests
of job quality effects and of cross-regime differences.
Table 8: Wald tests of jointly significant job quality effects on health outcome
Scandinavian Corporatist Southern Eastern Difference across
regimes
Acute health incident
Cardiovascular Risk * * Musculoskeletal Disorders * *** *** *
Incident Depression * * **
Risky Behaviour ** *
Functional disability
Onset poor/ fair health ** *** * * Note: * p<0.05, ** p<0.01, *** p<0.001, p-values from tests of jointly significant marginal effects of job quality within welfare state
regimes. Last column reports p-value from Wald tests of jointly homogenous job quality effects between welfare states.
The estimates confirm previous empirical studies in their inconsistency: health inequalities do not
adhere to the proposed ranking of welfare states. My findings suggest that job-related health
inequalities are least pronounced in Southern Europe (1 in 7 conditions), followed by countries with
a Corporatist or Eastern European type welfare state (3 in 7), and finally Scandinavia (5 in 7).
In southern Europe, job quality facets predict the development of cardiovascular risk factors (high
earnings), incident depression (intrinsic job quality), and the onset of poor health (intrinsic job
quality). All estimates are borderline significant and cumulate therefore into insignificant joint
effects of job quality, with the exemption of transitions into poor health.
In Corporatist welfare states, intrinsic job quality predicts innovations in cardiovascular risk factors,
musculoskeletal disorders, mental health and poor overall health. The marginal effect of intrinsic job
quality on the onset of musculoskeletal disorders is notably larger than in the pooled sample. There
is also a significant protective effect of high earnings on the occurrence of musculoskeletal disorders.
In all, job quality is related with changes in the major work-related health components:
musculoskeletal limitations, depression and poor health.
For Eastern European welfare states my findings suggest, if statistically significant, the quantitatively
largest health differences by intrinsic job quality. Marginal effects on the development of
cardiovascular risk factors, musculoskeletal disorders, and incident depression are highly statistically
significant and larger than in the other country groups. Given these clear and pronounced health
inequalities, the non-effect of intrinsic job quality on poor overall health is even more surprising.
Intrinsic job quality is like in the other regimes the main driver of health inequalities, but findings
additionally suggest significant protective effects of job security on musculoskeletal disorders and
from high earnings on the transition into poor health.
17
The findings for Scandinavia stand out. Like in the other countries, the estimates disclose protective
effects of intrinsic job quality on musculoskeletal disorders, mental health and poor general health.
But in addition there are significant beneficial effects of job security on the incidence of acute health
conditions and risky health behaviour. Further, it is the only country group where high earnings are
not found to be protective in at least one instance. Instead, I find that low earnings reduce the
occurrence of cardiovascular risk factors compared to a medium-paid job. As a result of the multiple
pathways of job quality on health, Scandinavia emerges as the region with the most diverse direct
health effects of job quality on health including on potentially fatal acute conditions.
Overall, the results support musculoskeletal disorders, depression and overall poor health as major
health outcomes of poor job quality. Job quality effects on cardiovascular risk factors are more
mixed across regimes and strongest in Eastern Europe. Intrinsic job quality emerges as the most
health active facet of job quality across welfare state regimes. The patterns for earnings and job
security are erratic and inconclusive. The onset of functional limitations is the only health outcomes
that remains statistically independent of job quality across regimes. But despite differences in
details, welfare state regimes can only account for some of the country heterogeneity in job quality
effects uncovered in the pooled analysis. Much of the differences remain unexplained.
6 Discussion and conclusions Sustainable jobs are essential for longer working lives in times of population ageing. In this study, I
have investigated the effects of job quality on a range of physical and mental health outcomes in a
general population of European workers aged 50 and above. The study makes use of a modern
comprehensive concept of job quality which combines earning, job security and intrinsic job quality
into a multidimensional measure of working conditions. Selection effects complicate the inference. I
adopt a methodology pioneered by Adams et al. (2003) and refined by Stowasser et al (2014) to
address reverse causality and endogenous selection into jobs.
Three key findings emerge. Firstly, causal effects of job quality on diverse health outcome cannot be
ruled out. Intrinsic, psychosocial, job quality is the most important job quality facet for subsequent
health innovations. It consistently predicts the occurrence of musculoskeletal disorders, incident
depressive symptoms and the onset of poor health. Job security and earnings predict health
outcomes in a few instances within specific policy context but not across the whole sample.
Secondly, whilst men are prone to develop psychiatric disorders in response to poor job quality,
women seem more physically vulnerable to job quality difference with alleviated occurrence of
cardiovascular risk factors. Thirdly, a differential analysis by welfare state types sees the broadest
range of job quality effects on health in the Nordic countries, whereas in Southern Europe job
quality is less key for the development of health inequalities. The quantitatively largest health
inequities are observed in Eastern European countries.
Overall the results largely confirm patterns found in the related literature on socioeconomic
positioning, working conditions and health inequalities. The inconsistent evidence on the role of
financial resources on the formation of health mirrors earlier empirical findings on the SES-health
nexus. On one hand, the non-effect of earning could be a testimony to the effectiveness of universal
public health care systems to achieve a decoupling of health from financial resources. On the other,
it could result from attenuation bias as earnings are measured with error and/ or approximate
health-related consumption only poorly. The limited influence of job security seems to contrast with
other findings in the literature. However, it is possible that job security is less of an issue for a
population close to retirement. Furthermore, previous research focused more broadly on mental
18
well-being, whereas the current study explores the effect on potentially clinical depressive
symptoms among other health outcomes.
The apparent inconsistency in predicted health inequalities across welfare states has caused some
stir among researchers in social epidemiology. So far studies have mainly investigated education-
related health inequities, but my findings suggest that the patterns carries over to job quality.
Several explanations have been put forward to illuminate this puzzle including stronger health-
related job selection in the Nordic countries or larger behavioural differences between job types. It
also conceivable that the systematic reduction of competing risk in highly developed countries with
generous social protection, increase the relative importance of job quality as predictor of health. If
job quality is partly hard wired into jobs by design, there will be limited scope for change.
This brings me to the potential policy implications. If my effects reflect causal effects, an
improvement in intrinsic job quality could reduce the onset of range of health disorders and
potentially help people to maintain employment to extend their working lives. The German
government has, for instance, been debating an anti-stress act proposed by the trade unions to
reduce psychosocial stressors at work. Over the working life, a reduction in stress might in fact
reduce wear and tear and reduce health inequalities. However my finding suggest that the change in
job quality would need to be quite substantial to realize shits in population health. Targeting other
population wide health risks such as obesity or risky health behaviour might prove to be more cost-
efficient. A careful assessment of costs and benefits of job quality improvements might be need to
provide a more conclusive answer.
Acknowledgment:
"This paper uses data from SHARE Wave 5 release 1.0.0, as of March 31st 2015 (DOI:
10.6103/SHARE.w5.100), SHARE Wave 4 release 1.1.1, as of March 28th 2013 (DOI:
10.6103/SHARE.w4.111), SHARE Waves 1 and 2 release 2.6.0, as of November 29th 2013 (DOI:
10.6103/SHARE.w1.260 and 10.6103/SHARE.w2.260) and SHARELIFE release 1.0.0, as of November
24th 2010 (DOI: 10.6103/SHARE.w3.100).
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)."
19
7 Appendix
Table 9: Job Quality Effects on Health Innovations by Gender
Variable Acute
Conditions CV Risks
Musculoskeletal Disorders
Depressive Symptoms
Risky Behaviour
Functional Disabilities
Poor/ Fair Health
Intrinsic Job Quality
M 0.0009 -0.002 -0.0188 *** -0.0065 *** 0.0004 -0.0001 -0.0046 **
0.0007 0.0021 0.0043 0.0011 0.0017 0.0009 0.0016
W -0.0003 -0.0054 ** -0.0211 *** -0.0019 0.0024 -0.0008 -0.0059 ***
0.0004 0.0019 0.0048 0.0015 0.0014 0.0011 0.0015
Job Security
M 0 0.0025 0.0065 -0.0002 0.001 -0.0012 -0.0009
0.0006 0.0022 0.0041 0.0012 0.0017 0.0008 0.0015
W 0.0001 0.0012 -0.0085 -0.0015 0.0005 0.0012 -0.0018
0.0004 0.0019 0.0044 0.0014 0.0014 0.0011 0.0014
EUR <200
M 0.0019 0.005 0.0088 0.0023 -0.0048 -0.0027 -0.0021
0.0025 0.0065 0.0142 0.0034 0.0053 0.0031 0.0052
W -0.0006 -0.0021 -0.0336 * 0.003 -0.0032 -0.0021 -0.008
0.0018 0.0053 0.0156 0.0046 0.0043 0.0035 0.0052
EUR 200-950
M -0.001 -0.0021 0.0078 0.0074 -0.0017 -0.0033 0.003
0.0025 0.0067 0.016 0.0042 0.0061 0.0033 0.0058
W -0.0007 0.006 -0.028 0.0017 -0.0049 -0.0001 -0.0104 *
0.0017 0.0055 0.0157 0.0048 0.0043 0.0036 0.005
EUR 950-1500
M (omitted) (omitted) (omitted) (omitted) (omitted) (omitted) (omitted)
20
W (omitted) (omitted) (omitted) (omitted) (omitted) (omitted) (omitted)
EUR 1500-2400
M 0.0039 0.01 -0.0099 0.006 -0.004 -0.0043 -0.0018
0.0031 0.0073 0.0146 0.0038 0.0059 0.0034 0.0059
W 0.0001 0.006 -0.027 0.0005 0.0017 -0.0073 -0.011
0.0021 0.0068 0.0192 0.0054 0.0053 0.0038 0.0064
EUR >=2400
M 0.0066 0.0051 -0.0161 0.0022 -0.0077 0.0003 -0.011
0.0072 0.0081 0.0154 0.0043 0.0065 0.0042 0.0058
W -0.0017 0.0024 -0.0207 0.0036 0.0011 0.006 -0.0068
0.0026 0.0079 0.0218 0.0072 0.0065 0.0069 0.0076
Gender Invariance 0.275 0.269 0.038 0.083 0.441 0.313 0.245
Time Invariance 0.428 0.797 0.002 0.742 0.894 0.458 0.454 Country Invariance 0.000 0.000 0.001 0.000 0.000 0.000 0.000 Joint Job Quality (Men) 0.437 0.479 0.000 0.000 0.913 0.555 0.011 Joint Job Quality (Women) 0.969 0.065 0.000 0.693 0.386 0.227 0.001
N 20124 20035 20035 19983 19983 18726 16400
Table 10: Test results for non-response bias
Test: participation in t+1 Test: total number of waves in the panel
Chi p Chi p
Acute health incident
22.366 0.000 5.097 0.024
21
Cardiovascular Risk
0.425 0.515 2.291 0.130
Musculoskeletal Disorders
3.434 0.064 6.171 0.013
Incident Depression
8.323 0.004 0.448 0.503
Risky Behaviour 0.388 0.534 0.038 0.846
Functional disability
1.330 0.248 0.672 0.412 )* Greece data exempt
Onset poor/ fair health
0.248 0.618 0.091 0.762
22
8 References
Adams, P. et al., 2003. Healthy, wealthy, and wise? Tests for direct causal paths between health and socioeconomic status. Journal of Econometrics, 112(1), pp.3–56.
Adams, P. et al., 2004. Healthy, Wealthy, and Wise? Tests for Direct Causal Paths between Health and Socioeconomic Status. In Chicago: University of Chicago Press.
Aerden, K. Van et al., 2015. The relationship between employment quality and work-related well-being in the European Labor Force. Journal of Vocational Behavior, 86, pp.66–76. Available at: http://dx.doi.org/10.1016/j.jvb.2014.11.001.
Backé, E.M. et al., 2012. The role of psychosocial stress at work for the development of cardiovascular diseases: A systematic review. International Archives of Occupational and Environmental Health, 85(1), pp.67–79.
Bambra, C. et al., 2014. Work, health, and welfare: the association between working conditions, welfare states, and self-reported general health in Europe. International journal of health services : planning, administration, evaluation, 44(1), pp.113–36. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24684087.
Bambra, C., 2011. Work, worklessness, and the political economy of health, Oxford University Press.
Barnay, T., 2014. Health, Work and Working Conditions. , (1148), p.32. Available at: http://www.oecd-ilibrary.org/economics/health-work-and-working-conditions_5jz0zb71xhmr-en.
Bassanini, A. & Caroli, E., 2014. Is Work Bad for Health? The Role of Constraint vs Choice. IZA Discussion Paper Series, 7891. Available at: http://www.econstor.eu/handle/10419/93285.
Bauer, G.F. et al., 2009. Socioeconomic status, working conditions and self-rated health in Switzerland: Explaining the gradient in men and women. International Journal of Public Health, 54(1), pp.23–30.
Benach, J. et al., 2014. Precarious employment: understanding an emerging social determinant of health. Annual review of public health, 35, pp.229–53. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24641559.
Benach, J. & Muntaner, C., 2007. Precarious employment and health: developing a research agenda. Journal of epidemiology and community health, 61(4), pp.276–277.
Bergqvist, K., Yngwe, M.A. & Lundberg, O., 2013. Understanding the role of welfare state characteristics for health and inequalities - an analytical review. BMC public health, 13, p.1234. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3909317&tool=pmcentrez&rendertype=abstract.
Bohle, P., Pitts, C. & Quinlan, M., 2009. Time to Call It Quits? The Safety and Health of Older Workers. International Journal of Health Services, 40(1), pp.23–41.
23
Börsch-Supan, A. et al., 2013. Data Resource Profile: the Survey of Health, Ageing and Retirement in Europe (SHARE). International journal of epidemiology, 42(4), pp.992–1001. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23778574 [Accessed February 3, 2014].
Börsch-Supan, A. et al. eds., 2008. First Results from the Survey of Health, Ageing and Retirement in Europe (2004-2007 ): Starting the Longitudinal Dimension, Mannheim: Manneheim Research Institute for the Economics of Aging (MEA).
Börsch-Supan, A., Hank, K. & Jürges, H., 2005. A new comprehensive and international view on ageing: introducing the “Survey of Health, Ageing and Retirement in Europe.” European Journal of Ageing, 2(4), pp.245–253. Available at: http://www.springerlink.com/index/10.1007/s10433-005-0014-9 [Accessed May 4, 2011].
Börsch-Supan, A. & Jürges, H. eds., 2005. The Survey of Health, Ageing and Retirement in Europe – Methodology, Mannheim: Mannheim Research Institute for the Economics of Aging (MEA).
Bound, J., Stinebrickner, T. & Waidmann, T., 2010. Health, economic resources and the work decisions of older men. Journal of Econometrics, 156(1), pp.106–129. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0304407609002097 [Accessed May 4, 2011].
Brennenstuhl, S., Quesnel-Vallee, a. & McDonough, P., 2012. Welfare regimes, population health and health inequalities: a research synthesis. Journal of Epidemiology & Community Health, 66(5), pp.397–409.
Bustillo, R.M. de B. et al., 2011. E pluribus unum? A critical survey of job quality indicators. Socio-Economic Review, 9(3), pp.447–475. Available at: http://ser.oxfordjournals.org/content/9/3/447.short [Accessed May 4, 2014].
Butterworth, P. et al., 2011. The psychosocial quality of work determines whether employment has benefits for mental health: results from a longitudinal national household panel survey. Occupational and Environmental Medicine, 68(11), pp.806–812.
Caroli, E. & Godard, M., 2013. Does Job Insecurity Deteriorate Health ? A Causal Approach for Europe. Paris School of Economics, Working Paper 2013 - 01, pp.0–30. Available at: http://hal-pse.archives-ouvertes.fr/docs/00/78/47/77/PDF/wp201301.pdf.
Case, A. & Deaton, A.S., 2005. Broken Down by Work and Sex: How Our Health Declines. In D. A. Wise, ed. Analyses in the Economics of Aging. University of Chicago Press, pp. 185–212. Available at: http://www.nber.org/chapters/c10361.
Case, A., Fertig, A. & Paxson, C., 2005. The lasting impact of childhood health and circumstance. Journal of health economics, 24(2), pp.365–89. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15721050 [Accessed March 15, 2013].
Castro-Costa, E. et al., 2007. Prevalence of depressive symptoms and syndromes in later life in ten European countries. The SHARE study. The British Journal of Psychiatry, 191(5), pp.393–401.
Contoyannis, P., Jones, A.M. & Rice, N., 2004. The dynamics of health in the British Household Panel Survey. Journal of Applied Econometrics, 19(4), pp.473–503. Available at: http://onlinelibrary.wiley.com/doi/10.1002/jae.755/full [Accessed July 13, 2014].
24
Copeland, J.R.M. et al., 2004. Depression among older people in Europe: the EURODEP studies. World psychiatry : official journal of the World Psychiatric Association (WPA), 3(1), pp.45–49.
Cottini, E., 2012. Is your job bad for your health? Explaining differences in health at work across gender. International Journal of Manpower, 33(3), pp.301–321. Available at: http://www.emeraldinsight.com/10.1108/01437721211234174.
Cottini, E. & Lucifora, C., 2013. Mental Health and Working Conditions in Europe. Industrial and Labor Relations Review, 66(July), pp.958–988. Available at: http://heinonlinebackup.com/hol-cgi-bin/get_pdf.cgi?handle=hein.journals/ialrr66§ion=49 [Accessed August 29, 2014].
Davies, R., Jones, M. & Lloyd-Williams, H., 2014. Age and Work-Related Health: Insights from the UK Labour Force Survey. British Journal of Industrial Relations, (Schils 2008).
Disney, R., Emmerson, C. & Wakefield, M., 2006. Ill health and retirement in Britain: a panel data-based analysis. Journal of Health Economics, 25(4), pp.621–49. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16678924 [Accessed September 29, 2010].
Dragano, N., Siegrist, J. & Wahrendorf, M., 2010. Welfare regimes, labour policies and unhealthy psychosocial working conditions: a comparative study with 9917 older employees from 12 European countries. Journal of epidemiology and community health.
Dragano, N. & Wahrendorf, M., 2014. Consistent health inequalities in Europe: the importance of labour market disadvantage. Journal of Epidemiology and Community Health , 68 (4 ), pp.293–294. Available at: http://jech.bmj.com/content/68/4/293.short.
Eikemo, T. a. et al., 2008. Health inequalities according to educational level in different welfare regimes: A comparison of 23 European countries. Sociology of Health and Illness, 30(4), pp.565–582.
EU-OSHA, 2014. Calculating the cost of work-related stress and psychosocial risks,
Fletcher, J.M., 2012. The Effects of First Occupation on Long Term Health Status: Evidence from the Wisconsin Longitudinal Study. Journal of Labor Research, 33(1), pp.49–75. Available at: http://link.springer.com/10.1007/s12122-011-9121-x [Accessed May 5, 2014].
Fletcher, J.M., Sindelar, J.L. & Yamaguchi, S., 2011. Cumulative Effects of Job Characteristics on Health. Health Economics, 20(5), pp.553–570.
Galama, T.J. & van Kippersluis, H., 2013. Health Inequalities through the Lens of Health-capital Theory: Issues, Solutions, and Future Directions. In P. R. Dias & O. O’Donnell, eds. Health and Inequality. Emerald Group Publishing Limited, pp. 263–284. Available at: http://dx.doi.org/10.1108/S1049-2585(2013)0000021013\nhttp://www.emeraldinsight.com/books.htm?chapterid=17103212.
Ganster, D.C. & Rosen, C.C., 2013. Work Stress and Employee Health: A Multidisciplinary Review, Available at: http://jom.sagepub.com/content/early/2013/02/19/0149206313475815.abstract\nhttp://jom.sagepub.com/cgi/doi/10.1177/0149206313475815.
25
García-Gómez, P., 2011. Institutions, health shocks and labour market outcomes across Europe. Journal of Health Economics, 30(1), pp.200–213.
García-Gómez, P. et al., 2013. Long Term and Spillover Effects of Health Shocks on Employment and Income. Journal of Human Resources, 48(4), pp.873–909.
Green, F. et al., 2013. Is job quality becoming more unequal? Industrial and Labor Relations Review, 66(4), pp.753–784.
Green, F., 2011. Unpacking the misery multiplier: How employability modifies the impacts of unemployment and job insecurity on life satisfaction and mental health. Journal of Health Economics, 30(2), pp.265–276. Available at: http://dx.doi.org/10.1016/j.jhealeco.2010.12.005.
Green, F. & Mostafa, T., 2012. Trends in Job Quality in Europe, Luxembourg: European Union. Available at: http://eprints.ioe.ac.uk/16320/ [Accessed May 30, 2014].
Grossman, M., 1972. On the concept of health capital and the demand for health. The Journal of Political Economy, 80(2), pp.223–255. Available at: http://www.jstor.org/stable/10.2307/1830580 [Accessed January 23, 2013].
Grossman, M., 2000. The Human Capital Model. Handbook of Health Economics, 1, pp.347–408.
Gruenewald, T.L. et al., 2012. History of socioeconomic disadvantage and allostatic load in later life. Social Science and Medicine, 74(1), pp.75–83. Available at: http://dx.doi.org/10.1016/j.socscimed.2011.09.037.
Gueorguieva, R. et al., 2009. The impact of occupation on self-rated health: cross-sectional and longitudinal evidence from the health and retirement survey. Journal of Gerontology: Social Sciences, 64B(1), pp.118–124. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2654983&tool=pmcentrez&rendertype=abstract [Accessed December 5, 2011].
Gupta, N.D. & Kristensen, N., 2008. Work environment satisfaction and employee health: panel evidence from Denmark, France and Spain, 1994–2001. The European Journal of Health Economics, 9(1), pp.51–61. Available at: http://link.springer.com/article/10.1007/s10198-007-0037-6 [Accessed August 26, 2014].
Häusser, J.A. et al., 2010. Ten years on: A review of recent research on the Job Demand–Control (-Support) model and psychological well-being, Available at: http://www.tandfonline.com/doi/abs/10.1080/02678371003683747 [Accessed July 4, 2011].
Havari, E. & Mazzonna, F., 2011. Can We Trust Older People’s Statement on Their Childhhod Circumstances? Evidence from SHARELIFE. SHARE Working Paper Series, (5).
Jerrim, J., 2014. The link between family backgrback and later life income: how does the UK compare to other countries?,
Johnson, J. V., Hall, E.M. & Theorell, T., 1989. Combined effects of job strain and social isolation on cardiovascular disease morbidity and mortality in a random sample of the Swedish male working population. Scandinavian journal of work, environment & health, 15(4), pp.271–279.
26
Jones, A.M. et al., 2013. Applied Health Economics 2 edition., London, New York: Routledge.
Jones, A.M., Rice, N. & Contoyannis, P., 2012. The dynamics of health. In A. M. Jones, ed. The Elgar Companion to Health Economics, 2nd Edition. Cheltenham, UK; Northampton, MA, USA: Edward Elgar Publishing, pp. 15–23.
Jones, M.K. et al., 2013. Work-related health risks in Europe: Are older workers more vulnerable? Social Science and Medicine, 88, pp.18–29. Available at: http://dx.doi.org/10.1016/j.socscimed.2013.03.027.
Kaikkonen, R. et al., 2009. Physical and psychosocial working conditions as explanations for occupational class inequalities in self-rated health. European Journal of Public Health, 19(5), pp.458–463.
Kalousova, L. & Mendes de Leon, C., 2015. Increase in frailty of older workers and retirees predicted by negative psychosocial working conditions on the job. Social Science & Medicine, 124, pp.275–283. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0277953614007874.
Karasek, R. et al., 1998. The Job Content Questionnaire (JCQ): an instrument for internationally comparative assessments of psychosocial job characteristics. Journal of occupational health psychology, 3(4), pp.322–55. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9805280.
Karasek, R. a, 1979. Job Demands , Job De- cision Latitude , and Mental Strain : Implica- tions for Job Redesign. Administrative Science Quarterly, 24(2), pp.285–308.
Karlamangla, A.S. et al., 2002. Allostatic load as a predictor of functional decline. MacArthur studies of successful aging. Journal of clinical epidemiology, 55(7), pp.696–710. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12160918.
Kelly, I.R. et al., 2012. The impact of early occupational choice on health behaviors. Review of Economics of the Household, 12(4), pp.737–770. Available at: http://link.springer.com/10.1007/s11150-012-9166-5 [Accessed May 5, 2014].
Kennedy, E.H., Krahn, H. & Krogman, N.T., 2013. Downshifting: An Exploration of Motivations, Quality of Life, and Environmental Practices. Sociological Forum, 28(4), pp.764–783.
Kivimäki, M. et al., 2012. Job strain as a risk factor for coronary heart disease: A collaborative meta-analysis of individual participant data. The Lancet, 380(9852), pp.1491–1497. Available at: http://dx.doi.org/10.1016/S0140-6736(12)60994-5.
Kjellsson, S., 2013. Accumulated occupational class and self-rated health. Can information on previous experience of class further our understanding of the social gradient in health? Social Science and Medicine, 81, pp.26–33. Available at: http://dx.doi.org/10.1016/j.socscimed.2013.01.006.
Landsbergis, P.A., Grzywacz, J.G. & LaMontagne, A.D., 2014. Work Organization, Job Insecurity, and Occupational Health Disparities. American Jounral of Industrial Medicine, 57(5), pp.495–515.
Larraga, L. et al., 2006. Validation of the Spanish version of the EURO-D Scale: an instrument for detecting depression in older people. International Journal of Geriatric Psychiatry, 21(12), pp.1199–1205.
27
Llena-Nozal, A., 2009. the Effect of Work Status and Working Conditions on Mental Health in Four Oecd Countries. National Institute Economic Review, 209(1), pp.72–87.
Loretto, W. & Vickerstaff, S., 2015. Gender, age and flexible working in later life. Work, Employment & Society. Available at: http://wes.sagepub.com/cgi/doi/10.1177/0950017014545267.
Lunau, T. et al., 2015. The Association between Education and Work Stress: Does the Policy Context Matter? Plos One, 10(3), p.e0121573. Available at: http://dx.plos.org/10.1371/journal.pone.0121573.
Mackenbach, J.P., 2012. The persistence of health inequalities in modern welfare states: The explanation of a paradox. Social Science and Medicine, 75(4), pp.761–769. Available at: http://dx.doi.org/10.1016/j.socscimed.2012.02.031.
Macmillan, L., Tyler, C. & Vignoles, A., 2013. Who gets the Top Jobs? The role of family background and networks in recent graduates ’ access to high status professions,
Macmillan, P. & Smith, I., 2007. Journal of Sports Economics.
Malter, F. & Borsch-Supan, A. eds., 2015. SHARE Wave 5: Innovations & Methodology, Munich: MEA. Max Planck Institute for Social Law and Social Policy.
Malter, F. & Börsch-Supan, A. eds., 2013. SHARE Wave 4: Innovations & Methodology, Munich: MEA, MEA, Max Planck Institute for Social Law and Social Policy.
Mazzonna, F., 2014. The long-lasting effects of family background: A European cross-country comparison. Economics of Education Review, 40, pp.25–42. Available at: http://dx.doi.org/10.1016/j.econedurev.2013.11.010.
McEwen, B.S. & Seeman, T., 1999. Protective and damaging effects of mediators of stress. Elaborating and testing the concepts of allostasis and allostatic load. Annals of the New York Academy of Sciences, 896, pp.30–47.
Moncada, S. et al., 2010. Psychosocial work environment and its association with socioeconomic status. A comparison of Spain and Denmark. Scandinavian journal of public health, 38(3 Suppl), pp.137–148.
Niedhammer, I. et al., 2014. Fractions of cardiovascular diseases and mental disorders attributable to psychosocial work factors in 31 countries in Europe. International Archives of Occupational and Environmental Health, 87, pp.403–411.
Nieuwenhuijsen, K., Bruinvels, D. & Frings-Dresen, M., 2010. Psychosocial work environment and stress-related disorders, a systematic review. Occupational medicine (Oxford, England), 60(4), pp.277–286.
Nixon, A.E. et al., 2011. Can work make you sick? A meta-analysis of the relationships between job stressors and physical symptoms. Work & Stress, 25(November), pp.1–22.
OECD, 2011. Health at a Glance 2011: OECD Indicators, OECD Publishing. Available at: http://dx.doi.org/10.1787/health_glance-2011-en.
28
Pouliakas, K. & Theodossiou, I., 2013. The economics of health and safety at work: An interdiciplinary review of the theory and policy. Journal of Economic Surveys, 27(1), pp.167–208.
Prince, M.J. et al., 1999. Depression symptoms in late life assessed using the EURO-D scale. Effect of age, gender and marital status in 14 European centres. The British Journal of Psychiatry, 174(4), pp.339–345.
Ravesteijn, B., van Kippersluis, H. & van Doorslaer, E., 2013. The Wear and Tear on Health: What is the Role of Occupation?, Available at: http://repub.eur.nl/pub/50276/ [Accessed June 14, 2014].
Reinhardt, J.D., Wahrendorf, M. & Siegrist, J., 2013. Socioeconomic position, psychosocial work environment and disability in an ageing workforce: a longitudinal analysis of SHARE data from 11 European countries. Occupational and environmental medicine, 70(3), pp.156–63. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23243100 [Accessed May 21, 2013].
Robone, S., Jones, A.M. & Rice, N., 2011. Contractual conditions, working conditions and their impact on health and well-being. The European Journal of Health Economics, 12(5), pp.429–444. Available at: http://link.springer.com/article/10.1007/s10198-010-0256-0 [Accessed August 26, 2014].
Schröder, M. ed., 2011. Retrospective Data Collection in the Survey of Health, Ageing and Retirement in Europe. SHARELIFE Methodology, Mannheim: Mannheim Research Institute for the Economics of Ageing (MEA).
Schütte, S. et al., 2014. Psychosocial working conditions and psychological well-being among employees in 34 European countries. International Archives of Occupational and Environmental Health, pp.1–11.
Seeman, M. et al., 2014. Social status and biological dysregulation: The “status syndrome” and allostatic load. Social Science & Medicine, 118, pp.143–151. Available at: http://linkinghub.elsevier.com/retrieve/pii/S027795361400519X.
Seeman, T.E. et al., 2001. Allostatic load as a marker of cumulative biological risk: MacArthur studies of successful aging. Proceedings of the National Academy of Sciences of the United States of America, 98(8), pp.4770–5. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=31909&tool=pmcentrez&rendertype=abstract.
Siegrist, J., 1996. Adverse health effects of high-effort/low-reward conditions. Journal of Occupational Health Psychology, 1(1), pp.27–41.
Siegrist, J. et al., 2012. Depressive symptoms and psychosocial stress at work among older employees in three continents. Globalization and Health, 8(1), p.27. Available at: Globalization and Health.
Siegrist, J. et al., 2004. The measurement of effort-reward imbalance at work: European comparisons. Social science & medicine (1982), 58(8), pp.1483–99. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14759692 [Accessed March 21, 2011].
Smith, J.P., 2009a. Reconstructing childhood health histories. Demography, 46(2), pp.387–403.
29
Smith, J.P., 2009b. The Impact of Childhood Health on Adult Labor Market Outcomes. Review of Economics and Statistics, 91(3), pp.478–489.
Standing, G., 2011. The Precariat: The New Dangerous Class, London, New York: Bloomsbury Academic.
Steptoe, A. & Kivimäki, M., 2012. Stress and cardiovascular disease. Nature Reviews Cardiology, 9(6), pp.360–370. Available at: http://dx.doi.org/10.1038/nrcardio.2012.45.
Stowasser, T. et al., 2012. “Healthy, Wealthy and Wise?” Revisited: An Analysis of the Causal Pathways from Socioeconomic Status to Health. In D. A. Wise, ed. Investigations in the Economics of Aging. University of Chicago Press, pp. 267–317. Available at: http://www.nber.org/chapters/c12443 [Accessed September 4, 2014].
Stowasser, T. et al., 2014. Understanding the SES gradient in health among the elderly: The role of childhood circumstances. Munich Discussion Paper, (2014-16). Available at: http://www.nber.org/chapters/c12976.pdf.
Sverke, M., Hellgren, J. & Näswall, K., 2006. Job insecurity A literature review. National Institute for Working Life, (1), p.32. Available at: www.arbetslivsinstitutet.se/saltsa.
Vanroelen, C., Levecque, K. & Louckx, F., 2009. Psychosocial working conditions and self-reported health in a representative sample of wage-earners: A test of the different hypotheses of the Demand-Control-Support-Model. International Archives of Occupational and Environmental Health, 82(3), pp.329–342.
Verbeek, M. & Nijman, T., 1992. Testing for selectivity bias in panel data models. International Economic Review, pp.681–703. Available at: http://www.jstor.org/stable/2527133 [Accessed August 29, 2014].
Virtanen, M. et al., 2013. Perceived job insecurity as a risk factor for incident coronary heart disease: systematic review and meta-analysis. BMJ (Clinical research ed.), 347(August), p.f4746. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3738256&tool=pmcentrez&rendertype=abstract.
Van Vuuren, D., 2014. Flexible retirement. Journal of Economic Surveys, 28(3), pp.573–593.