Depressed during the depression: has the economic crisis affected mental health inequalities in Europe? Findings from the European Social Survey (2014) special module on the determinants of health
European Journal of Public Health
Depressed during the depression: has the economic crisis affected mental health inequalities in Europe? Findings from the European Social Survey (2014) special module on the determinants of health
Nadine Reibling 0
Jason Beckfield 1
Tim Huijts 2
Alexander Schmidt-Catran 3
Katie H. Thomson 4
Claus Wendt 0
0 University of Siegen , Adolf-Reichwein-Str. 2, Siegen 57068 , Germany
1 Department of Sociology, Harvard University , 33 Kirkland Street, Cambridge, MA 02138 , USA
2 Wentworth College, University of York , Heslington, York YO10 5DD
3 University of Cologne, Institute of Sociology and Social Psychology , Albertus-Magnus-Platz 50923, Cologne
4 Institute of Health and Society, Faculty of Medical Sciences, Newcastle University , The Baddiley-Clark Building, Richardson Road, Newcastle upon Tyne, NE2 4AX , United Kingdom
European Journal of Public Health, Vol. 27, Supplement 1, 2017, 47-54 The Author 2017. Published by Oxford University Press on behalf of the European Public Health Association. All rights reserved. doi:10.1093/eurpub/ckw225 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Background: Economic crises constitute a shock to societies with potentially harmful effects to the mental health status of the population, including depressive symptoms, and existing health inequalities. Methods: With recent data from the European Social Survey (2006-14), this study investigates how the economic recession in Europe starting in 2007 has affected health inequalities in 21 European nations. Depressive feelings were measured with the CES-D eight-item depression scale. We tested for measurement invariance across different socio-economic groups. Results: Overall, depressive feelings have decreased between 2006 and 2014 except for Cyprus and Spain. Inequalities between persons whose household income depends mainly on public benefits and those who do not have decreased, while the development of depressive feelings was less favorable among the precariously employed and the inactive than among the persons employed with an unlimited work contract. There are no robust effects of the crisis measure on health inequalities. Conclusion: Negative implications for mental health (in terms of depressive feelings) have been limited to some of the most strongly affected countries, while in the majority of Europe persons have felt less depressed over the course of the recession. Health inequalities have persisted in most countries during this time with little influence of the recession. Particular attention should be paid to the mental health of the inactive and the precariously employed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ocial scientists have had a longstanding interest in the
implicaStions of economic crises for health,1 because such crises represent
a shock to the affected societies with profound changes in the general
social structure and thus, the social determinants of health. The
theoretical expectations of these crisis effects are not straightforward.
On the one hand, economic crises lead to increases in
unemployment and income loss as well as reduced government funds2 which
could harm population health due to the elevated stress levels and
the reduction in individual and societal coping resources (e.g.
reductions in healthcare spending).3 On the other hand, economic
downturns might reduce traffic, pollution and exposure to harmful
work conditions (including stress).4,5 Moreover, in times of crisis,
individuals have less money for harmful health behaviors (e.g.
smoking) and more time for health-promoting behaviors (e.g.
exercise).6 as well as their social networks, an important resource
for coping with stress.7 Finally, the cultural meaning of a time of
crisis might reduce the health impacts of negative life events such as
unemployment, because they can be interpreted as a collective
experience rather than personal failure.8 Of course, these negative
and positive effects can also cancel each other out, resulting in no
clear ?net effect? of economic crisis.
Empirical evidence from various economic crisis provide mixed
results of the relationship between macroeconomic changes and
health.2,4,9,10 The global financial crisis from 2007 has spurred new
interest and empirical material for the investigation of crisis effects.
Although some studies reported more mental health disorders,
alcohol abuse,11?13 a rise in poor self-reported health,13,14 more
infectious diseases, and increased suicide rates,13?15 other studies
found that health indicators and several health-promoting
behaviors remained stable or improved in the period of the
Even though the economic crisis is a macro-level shock to
societies as a whole, it may impact some groups more than others.
Thus, we should be equally concerned with the potential influence of
economic crisis on not only health, but also health inequality.18
Evidence from earlier economic downturns suggest that in the UK
and Japan recessions have exacerbated health inequalities, while in
Scandinavia inequalities have remained stable or even decreased
during this time.19 This demonstrates the importance of
crossnational comparative work on the effect of economic crisis on
health inequalities. To our knowledge two studies have investigated
the effect of the crisis on health inequalities in Europe:1 Abebe et al.3
using panel data from the EU-SILC from 2005 and 2011 find no
signs of a stronger effect of the crisis for self-rated health of the low
educated and unemployed,2 Buffel et al.17 used 2006 and 2012 data
from the European Social Survey (ESS) to investigate the effect of
the crisis on mental health for groups with different employment
status. They find that persons in marginal part-time employment
showed more depressive feelings in countries with a higher increase
in unemployment as well as several groups of inactive men and
This study investigates the impact of the crisis on mental health
inequalities in Europe?focusing on depressive feelings as one
important indicator for low mental health. Our study advances the
literature in several ways. First, by using three waves of the ESS
including the first health module from 2014, we provide the most
long-ranging perspective on the crisis so far. Having information from
2014 is crucial because the peak of the unemployment within the EU
was in 2013 and thus was not included in earlier analysis. Second,
previous analysis have focused on inequalities between educational
groups3,20 and groups with different employment status in the
working-age population.17 Although the unemployed and persons
with a precarious employment status are important and likely
candidates for stronger impacts of the crisis, the crisis in Europe has
also been strongly associated with (the threat of) austerity measures.
Therefore, we investigate if persons from households that strongly
depend on public benefits showed a stronger increase in depressive
feelings. Third, we examined how the crisis affects persons that are
economically vulnerable by looking at inequalities in depressive
symptoms across income groups.
The data used for this study are three waves of the ESS21?23 including
21 European countries. All countries contain pre-crisis measurement
from wave 3 (2006) and post-crisis-onset measurements from wave
6 (2012) and/or wave 7 (2014). The design of the ESS is based on
strict random probability sampling and samples are representative
for all persons aged 15 and over in the individual countries. All
interviews were conducted face-to-face. Our analysis sample
contains 106 158 respondents after we have deleted all respondents
with missing values (2% of the ESS sample). As the number of
missing cases is low, we expect bias resulting from listwise deletion
to be negligible.24
Depressive feelings were measured using the eight-item version of
the Center of Epidemiological Studies-Depression (CES-D) scale.25
Respondents are asked how often within the past week they (i) felt
depressed, (ii) felt everything they did as effort, (iii) had restless
sleep, (iv) were happy, (v) felt lonely, (vi) enjoyed life, (vii) felt
sad and (viii) could not get going. There are four response
categories: none or almost none of the time, some of the time,
most of the time, all or almost all of the time.
We estimated measurement models to test different hypotheses
about the dimensionality of this scale.26,27 The one-dimensional
(1D) model assumes that all eight items load together on one
common factor, which we would label ?generalized depression?.
The two-dimensional model (2D) differentiates two factors
?depressed affect? consisting of items 1, 4, 5, 6, 7 (see above) and
?somatic complaints? including items 2, 3 and 8. Two of the eight
items of the scale are reverse-worded which can lead to low
scalereliability and multi-factor solutions.28 One of the reasons for these
effects of reverse-worded items is careless respondent behaviour
when all questions are answered with the same response without
noticing the reverse-worded items. Such careless behaviour
exhibits response bias which can be identified through zero
variance around the mean score for individual respondents. Due
to the potential negative implications for scale consistency,28 we
excluded the 1551 observations (1.5% of the analytical sample)
with no variance on the eight depression items.
In addition, we used different modeling strategies for the two
reverse-worded items. First, we test a three-factor solution (3D)
that differentiates within the factor affect between ?depressive
affect? with the three negative items 1, 5 and 7 and ?positive affect?
including items 4 and 6. An alternative to account for the different
polarity of these items is the specification of a covariance between
the error terms of items 4 and 6 in the 1D or 2D solution. We
compare these five different models in a multi-group confirmatory
factor analysis framework in order to determine the best-fitting
solution for the ESS-data.
Three independent variables are used to identify social inequalities
in health in depressive feelings. First, information on ?employment
status? is used to identify vulnerable groups in the labour market such
as the unemployed and persons in precarious employment. Our
categorical measure distinguishes between five groups: (i) Employed
persons with unlimited contract, (ii) self-employed persons or
persons working for a family business, (iii) persons in precarious
employment (with no or a limited work contract), (iv) unemployed
persons and (v) inactive persons covering those who are in education,
retired, in community or military service, the (permanently) sick or
disabled, and homemakers. Second, for the independent variable
?subjective income? we relied on the question ?Which of the
descriptions on this card comes closest to how you feel about your
household?s income nowadays??(We use subjective rather than
objective income, because it has been measured consistently across
countries and time periods. The measurement of objective income has
changed over the waves in the ESS and there are many cross-national
and?temporal income definitions even in the most recent waves.
Therefore, while technically feasible a reliable and valid
harmonization of objective income is extremely challenging. Moreover, a higher
number of missing cases (20 compared with <1%) would require the
use of multiple imputation for the harmonized variable. Since we
include a larger number of waves and specifically analyse trends, we
are convinced that subjective income is the most adequate
operationalization of income groups in the context of the ESS
data.). The four response categories are: (i) living comfortably on
present income, (ii) coping on present income, (iii) finding it
difficult on present income and (iv) finding it very difficult on
present income. Due to low number of cases in some countries, we
combined the latter two categories into one. Subjective income is used
because the objective income measure changed between the waves and
a valid harmonization is not available. Third, particularly within
Southern Europe, the crisis has led to cuts of public benefits.
Therefore, we look at persons whose primary income sources are
based on public benefits compared with those with other primary
income sources. All models also include sex, age, and education as
control variables. Age is categorized in seven groups. Education, as an
important predictor of both mental health29 and employment status/
income, is measured in years.
In order to directly assess the impact of the crisis we use the
change in unemployment rate4,17,30 and GDP per capita between
the later waves and the pre-crisis year 2006. Since the impact of
the crisis may depend on the previous situation in the economy,
we also control for the pre-crisis GDP per capita and unemployment
level as time-invariant macro-indicators. The macro indicators were
extracted from the Eurostat database.
Appendix Table A1 reports descriptive statistics.
Previous studies have tested measurement invariance for the CES-D
8 scale with ESS data across men and women with the 2006 wave.26
The set-up of this study required an assessment of measurement
invariance across socio-economic groups as well as waves.
Multigroup Confirmatory factor analysis was estimated for each
socioeconomic group and the country indicator to assess the
cross-cultural invariance of the scale, and for each socioeconomic
group and the wave indicator to assess inter-temporal invariance of
the scale. The analyses were conducted with maximum-likelihood
estimation using STATA Version 13.1. Goodness-of-fit statistics
were used to assess and compare model specification. Chi-square
tests, the traditional way of assessing goodness-of-fit, are heavily
influenced by sample size and have proven not robust in large
samples like ours. Therefore, while we report chi2 values, we based
or assessment on more robust absolute fit indices
(Root-MeanSquare-Error of Approximation?RMSEA, Comparative Fit
Index?CFI, Tucker-Lewis-index (TLI).31
In a second step, we use the predicted latent depression scale as a
dependent variable in a multilevel growth curve analysis. Due to the
repeated cross-sectional design we specified our multi-level model by
considering the individuals (level 1) nested in 59 country years (level
2) nested in 21 countries (level 3). Since the time trends are of specific
interest for the assessment of the crisis effects, year dummies are
added to the model. Interactions between year dummies and
socioeconomic groups are used to assess trends in health inequalities.
Figure 1 Trends in depressive feelings (
) between 2006 and 2015, by country (sorted by increase in unemployment rate). Source: ESS
(2006, 2012, 2014), own calculations, weighted means, AT, Austria; BE, Belgium; CH, Switzerland; DE, Germany; FR, France; NL, Netherlands;
DK, Denmark; FI, Finland; NO, Norway; SE, Sweden; CY, Cyprus; ES, Spain; PT, Portugal; GB, Great Britain; IE, Ireland; BG, Bulgaria; EE,
Estonia; HU, Hungary; SI, Slovenia; SK, Slovakia.
Model fit and measurement invariance
Table 1 reports the model comparison of the different configurations
of the eight CES-D items. Due to the large number of observations,
all models show a significant chi2, but the three models that take the
positive phrasing of the two items ?enjoyed life? and ?were happy?
into account (models 2, 4, and 5) have a good model fit with the CFI
and the TLI above .95 and the RMSEA below 0.06. These indices also
indicate that models 4 and 5 fit the data better than model 2.
However, in line with an earlier analysis of the ESS data,24 the
different model dimensions are highly correlated. Thus, the
squared correlations (SCs) are higher than the average extracted
variance (AVE) which indicates low discriminant validity. As the
1D specification with correlated errors (model 2) still shows a
good fit to the data and is more parsimonious for the specification
and the presentation of results, we proceed with this model.
Next, we assessed measurement invariance across socioeconomic
groups by wave or country for the pooled sample (see Table 2). The
top of the left panel shows that both for the entire sample and for the
different socio-economic groups configural, metric and scalar
invariance can be assumed across waves, because all indices show
a good fit (RMSEA < 0.06, CFI > 0.95, TLI > 0.95). In many
configurations, the fit indices even indicate a better model fit when
factor loadings are assumed identical (metric invariance) or even
when item intercepts are constrained to be equal (scalar
invariance). Below are the invariance tests between countries and
socioeconomic groups. The indices indicate that there is both
configural and metric invariance across countries (and
socioeconomic groups) with either good or acceptable fit (RMSEA
< 0.08, CFI > 0.9, TLI > 0.9) to the data. However, none of the
models shows an acceptable fit when item intercepts are constrained
to be equal (scalar invariance). Since the aim of this study is the
analysis of time trends and associations, metric invariance is
sufficient. A further indication that time trends can be compared
across countries is that scalar invariance across years can be found
also when tested for each individual country (In Hungary, the
overall model fit cannot consistently be seen as acceptable.
However, both RMSE and CFI suggest a better model fit if factor
loadings and item intercepts are constrained.)
Source: ESS (2006, 2012, 2014), pooled sample, own calculations, SE, standard errors in parentheses, significance levels: P < 0.05, P < 0.01,
P < 0.001, models are controlling for sex, age in categories, and years of education; random effects for income, employment status and
income source are specified both on the country and the country-year level and are all significant.
Trends in depression during the economic crisis
Figure 1 shows trends in depressive feelings for the individual
countries as a whole and three vulnerable groups. In the vast
majority of the countries, the general population reported fewer
depressive feelings in 2012 than in 2006. Only the trends in
Cyprus and Spain indicate a rise in depressive feelings. Most
countries also show a further decline in depressive feelings
between 2012 and 2014. Three Nordic countries (Denmark,
Norway, and Sweden) as well as Portugal, and Slovenia have a
higher average of depressive feelings in 2014 than in 2012.
If we compare the trend of the vulnerable population groups with
the overall trend, we get a first impression of how inequalities
developed during the economic crisis. Inequality trends are rather
complex. In the four countries with the strongest increase in
unemployment rates (Portugal, Ireland, Cyprus and Spain) inequalities
seemed to have decreased between 2006 and 2012 and either
remained stable or increased again, thereafter. In a number of
countries particularly the unemployed seem to have caught up to
the general population average during the crisis (e.g. in Austria,
Sweden, Finland). However, in Hungary depressive feelings of the
unemployed have substantially risen between 2006 and 2012
compared with a declining trend of the overall population. In
sum, over the course of the recession, inequalities generally have
neither increased nor decreased, but there are indications that in
the ?crisis countries? inequalities initially might have become smaller.
In the multivariate three-level growth curve analysis (see Table 3),
we find that for the pooled sample depressive feelings decreased in
2012 ( = 0.046+) and even stronger in 2014 ( = 0.077 ).
Countries with a higher GDP per capita before the crisis exhibit a
lower level of depressive feelings ( = 0.008+). The impact of the
crisis operationalized by the change in unemployment rates shows a
positive association with depressive feelings ( = 0.008 ) (Model
1). This effect seems to be mainly due to changes in the composition
of income and employment groups, because the effect is no longer
significant in Model 2. In addition, sensitivity analysis excluding one
country at a time shows that the effect of the crisis is sensitive to the
inclusion of individual countries.
In Model 2, the individual-level variables were added. Significant
inequalities in depressive feelings are found for all three indicators.
Individuals with a lower subjective income show significantly more
depressive feelings ( = 0.122 ; = 0.441 ). Compared with
employed persons with an unlimited work contract, inactive ( =
0.099 ), unemployed ( = 0.139 ), and precariously employed
persons ( = 0.077 ) felt more depressed. Finally, even when
controlling for the income level, persons whose household income
stems primarily from public benefits report significantly more
depressive feelings ( = 0.098 ).
In order to investigate potential crisis effects on inequalities, we
estimate interactions both with the time trend to see if trends
developed differently across the social groups, and with the change
in unemployment rate (model 3). Only significant interactions are
reported. We find that differences in depressive feelings between
those who primarily rely on public benefits for their income and
those who do not have become smaller between 2006 and 2014,
because of a stronger downward trend of those with primarily
public income ( = 0.048 ). In contrast inequalities in
employment status (reference: employed with unlimited work
contract) have increased for the inactive ( = 0.036 ) and the
precariously employed ( = 0.038 ).
A direct test of the effect of the crisis is performed in model 4
which includes interactions between the change in unemployment
rate and the socio-economic variables. A larger increase in
unemployment is associated with larger inequalities between persons
relying mainly on public benefits and those who do not ( = 0.004+).
In contrast, a larger increase in unemployment is associated with less
inequality in depressive feelings between the employed and the
unemployed ( = 0.004+). However, both interactions are only
The 2007 banking crisis had a strong impact on the European
economy in the following years. As a result, both researchers and
policymakers are concerned about the implications of the crisis for
overall population health and health inequalities. Several studies
have documented negative health implications in the Southern
European countries that were both strongly hit by the crisis, and
implemented austerity measures during the recession.11,14,20,32
The aim of this article was to investigate if an impact of the crisis
on health can be found across Europe with most up-to-date data
(2006?14) that cover the whole development of the recession
including the peak in unemployment at 2013.
First, has the economic crisis affected the level of depressive
feelings in Europe? No. Our pan-European analysis indicates that
the economic crisis was not associated with a general increase in
depressive feelings, since both the overall trend showed a decline in
such feelings and the direct effect of crisis (change in unemployment)
was not robust. Second, has the economic crisis resulted in larger
health inequalities? Not generally. We found no systematic effect of
the recession on health inequalities across Europe, except that in the
crisis countries, some inequalities decreased rather than increased.
Thus, our study adds a longer-term perspective and confirms
Buffel et al.?s17conclusion that the crisis had no overall negative
effect on mental health in Europe. However, in contrast, we do
not find effects of the unemployment rate on inequalities between
groups with different employment status nor for the additional
vulnerable groups that we analysed (based on low subjective
income or public-benefit based income). Thus, our conclusion
with respect to inequalities are in line with the panel-data results
based on the EU-SILC(
) which also concluded that inequalities are
stable and have not been affected by the crisis.
There are some limitations to the data and analysis presented
here. First, the study used a research design with repeated
crosssectional data which is limited in the identification of causal
effects. However, a comparison of individual health trajectories is
not possible with the existing cross-national panel data (e.g.,
EUSILC).3 Second, the ESS sample unfortunately excludes several
countries that were strongly affected by the economic crisis, viz.
Greece, Iceland, and Romania. Third, bias may arise through
selective non-response, particularly as depressive feelings might
lower the probability of survey participation. Moreover, persons
with a lower socio-economic status are less likely to participate.
Since selective non-response might have been even stronger after
the crisis, we might underestimate health inequalities after the
recession, particularly in the countries that have been most
strongly affected by the crisis. Fourth, the subjective income
measure we use may not concur with objective income measures
and the crisis could lead to respondents overestimating financial
difficulties. If this was the case, we may not be able to measure
the full extent of health inequalities between income groups
during the peak of the crisis. Finally, the ESS includes only one
measure of depressive feelings before the crisis so that we do not
know how they developed up until then. Thus, we might miss a
deceleration of mental health improvements during the crisis.
Limitations acknowledged, we believe our study carries several
implications. For Europe as a whole, health inequalities present a
consistent public health issue independent of the economic
recession. Our study suggests that the health disadvantage of the
inactive and precariously employed has increased in the last
decade and thus interventions should focus particularly on these
Terje A. Eikemo, Clare Bambra and Tim Huijts led the design of the
ESS special module on the social determinants of health in
coordination with Rory Fitzgerald of the ESS.
This article is part of the HiNEWS project?Health Inequalities in
European Welfare States?funded by NORFACE (New
Opportunities for Research Funding Agency Cooperation in
Europe) Welfare State Futures programme (grant
reference:46214-110). For more details on NORFACE, see http://www.norface.
Conflicts of interest: None declared.
Source: ESS (2006, 2012, 2014), pooled sample, own calculations, weighted means and proportions.
1 Brenner MH. Fetal , Infant, and Maternal Mortality during Periods of Economic Instability . Int J Health Serv 1973 ; 3 : 145 - 59 .
2 Cutler DM , Knaul F , Lozano R , et al. Financial crisis, health outcomes and ageing: Mexico in the 1980s and 1990s . J Public Econ 2002 ; 84 : 279 - 303 .
3 Abebe DS , T?ge AG , Dahl E . Individual-level changes in self-rated health before and during the economic crisis in Europe . Int J Equity Health 2016 ; 15 : 1 .
4 Ruhm CJ . Are Recessions Good for Your Health? . Q J Econ 2000 ; 115 : 617 - 50 .
5 Ruhm CJ . Understanding the Relationship between Macroeconomic Conditions and Health . In: Jones AM, editor. The Elgar Companion to Health Economics. Cheltenham , UK: Edward Elgar Pub., 2012 .
6 A?sgeirsdo? ttir TL , Corman H , Noonan K , et al. Are Recessions Good for Your Health Behaviors? Impacts of the Economic Crisis in Iceland [Internet] . National Bureau of Economic Research; 2012 Juli [zitiert 30 . Ma?rz 2016 ]. Report No.: 18233 . Verfu?gbar unter: http://www.nber.org/papers/w18233
7 Stuckler D , Basu S , Suhrcke M , et al. The public health effect of economic crises and alternative policy responses in Europe: an empirical analysis . Lancet 2009 ; 374 : 315 - 23 .
8 Sharone O. Flawed System/Flawed Self: Job Searching and Unemployment Experiences . University of Chicago Press; 2013 . 239 S.
9 Friedman J , Thomas D. Psychological Health Before , During, and After an Economic Crisis: Results from Indonesia , 1993 - 2000 . World Bank Econ Rev 2009 ; 23 : 57 - 76 .
10 Borowy I. Similar but different: Health and economic crisis in 1990s Cuba and Russia . Soc Sci Med 2011 ; 72 : 1489 - 98 .
11 Gili M , Roca M , Basu S , et al. The mental health risks of economic crisis in Spain: evidence from primary care centres, 2006 and 2010 . Eur J Public Health 2013 ; 23 : 103 - 8 .
12 Bartoll X , Pale`ncia L , Malmusi D , et al. The evolution of mental health in Spain during the economic crisis . Eur J Public Health 2014 ; 24 : 415 - 8 .
13 Simou E , Koutsogeorgou E. Effects of the economic crisis on health and healthcare in Greece in the literature from 2009 to 2013: a systematic review . Health Policy 2014 ; 115 : 111 - 9 .
14 Kentikelenis A , Karanikolos M , Papanicolas I , et al. Health effects of financial crisis: omens of a Greek tragedy . Lancet 2011 ; 378 : 1457 - 8 .
15 Chang S-S , Stuckler D , Yip P , Gunnell D. Impact of 2008 global economic crisis on suicide: time trend study in 54 countries . BMJ 2013 ; 347 : f5239
16 Regidor E , Mart??nez D , Santos JM , et al. New findings do not support the neomaterialist theory of the relation between income inequality and infant mortality . Soc Sci Med 2012 ; 75 : 752 - 3 .
17 Buffel V , Van de Velde S , Bracke P. The mental health consequences of the economic crisis in Europe among the employed, the unemployed, and the non-employed . Soc Sci Res 2015 ; 54 : 263 - 88 .
18 Escolar-Pujolar A , Bacigalupe A , San Sebastian M. European economic crisis and health inequities: research challenges in an uncertain scenario . Int J Equity Health 2014 ; 13 : 59
19 Bambra C , Garthwaite K , Copeland A , Barr B. All in it together? Health Inequalities, Welfare Austerity and the Great Recession . In: Smith KE , Bambra C , Hill SE , editors. Health Inequalities: Critical Perspectives . Oxford: Oxford University Press, 2016 : 164 - 76 .
20 Zapata Moya AR , Buffel V , Navarro Ya?n?ez CJ, Bracke P. Social inequality in morbidity, framed within the current economic crisis in Spain . Int J Equity Health 2015 ; 14 : 131
21 ESS Round 3 . European Social Survey Round 3 Data. Data file edition 3 .5. Norwegian Social Science Data Services , Norway - Data Archive and distributor of ESS data for ESS ERIC , 2006 .
22 ESS Round 6 . European Social Survey Round 6 Data. Data file edition 2 .2. Norwegian Social Science Data Services , Norway - Data Archive and distributor of ESS data for ESS ERIC , 2012 .
23 ESS Round 7 . European Social Survey Round 7 Data. Data file edition 1 .0. Norwegian Social Science Data Services , Norway - Data Archive and distributor of ESS data for ESS ERIC , 2014 .
24 Barzi F , Woodward M. Imputations of missing values in practice: results from imputations of serum cholesterol in 28 Cohort studies . Am J Epidemiol 2004 ; 160 : 34 - 45 .
25 Radloff LS . The CES-D scale: a self-report depression scale for research in the general population . Appl Psychol Meas 1977 ; 1 : 385 - 401 .
26 Van de Velde S , Bracke P , Levecque K , Meuleman B . Gender differences in depression in 25 European countries after eliminating measurement bias in the CES-D 8 . Soc Sci Res 2010 ; 39 : 396 - 404 .
27 Riddle AS , Blais MR , Hess U . Static versus dynamic structural models of depression: the case of the CES-D . CIRANO Sci Ser 2002 ; 2002s - 37 .
28 Woods CM . Careless Responding to Reverse-Worded Items: Implications for Confirmatory Factor Analysis . J Psychopathol Behav Assess 2006 ; 28 : 186 - 91 .
29 Dudal P , Bracke P . Absolute and relative educational inequalities in depression in Europe . Int J Public Health [Internet]. 24. Mai 2016 ; Verfu?gbar unter: http://link. springer. com/10.1007/s00038-016-0837-5 ( 13 August 2016 , date last accessed)
30 Noelke C , Beckfield J. Recessions, job loss, and mortality among older US adults . Am J Public Health 2014 ; 104 : e126 - 34 .
31 Brown TA . Confirmatory Factor Analysis for Applied Research . New York/London: The Guilford Press, 2006 .
32 Karanikolos M , Mladovsky P , Cylus J , et al. Financial crisis, austerity, and health in Europe . Lancet 2013 ; 381 : 1323 - 31 .