The European Journal of Health Economics

, Volume 16, Issue 2, pp 175–184 | Cite as

The influence of the economic crisis on the association between unemployment and health: an empirical analysis for Spain

  • Rosa M. Urbanos-Garrido
  • Beatriz G. Lopez-Valcarcel
Original Paper



To estimate the impact of (particularly long-term) unemployment on the overall and mental health of the Spanish working-age population and to check whether the effects of unemployment on health have increased or been tempered as a consequence of the economic crisis.


We apply a matching technique to cross-sectional microdata from the Spanish Health Survey for the years 2006 and 2011–2012 to estimate the average treatment effect of unemployment on self-assessed health (SAH) in the last year, mental problems in the last year and on the mental health risk in the short term. We also use a differences-in-differences estimation method between the two periods to check if the impact of unemployment on health depends on the economic context.


Unemployment has a significant negative impact on both SAH and mental health. This impact is particularly high for the long-term unemployed. With respect to the impact on mental health, negative effects significantly worsen with the economic crisis. For the full model, the changes in effects of long-term unemployment on mental problems and mental health risk are, respectively, 0.35 (CI 0.19–0.50) and 0.20 (CI 0.07–0.34).


Anxiety and stress about the future associated with unemployment could have a large impact on individuals’ health. It may be necessary to prevent health deterioration in vulnerable groups such as the unemployed, and also to monitor specific health risks that arise in recessions, such as psychological problems.


Economic crisis Unemployment Self-assessed health Mental health Matching techniques Spain 

JEL Classification

J64 I12 I18 


The impact of economic recessions on health has been previously addressed. Researchers mainly focused on the role played by unemployment as a mediator agent [1, 2, 3], because unemployment and working conditions constitute major social determinants of health [4]. Beyond the influence of the institutional context of the labour market and social protection, most attention has been paid to the study of the risk factors linking labour status and health. Several health economics papers conclude that economic downturns have a counter-cyclical role in terms of health, and that short-term unemployment improves population health and reduces mortality in developed countries [5, 6, 7, 8, 9]. Moreover, public health literature provides evidence that being employed protects and promotes health [10, 11, 12, 13].

Previous studies show that unemployment and fall in income may lead to obesogenic diets [14] or be associated with health risk behaviours such as excessive alcohol consumption [15], more smoking [16] or decreased physical activity [17]. Furthermore, a reduction in the level of income may discourage seeking medical attention to avoid treatment costs [18]. This effect is particularly strong in those countries where health coverage is linked to labour status and/or the amount of copayments is significant [19]. Unemployment can also impair mental health by various psychological mechanisms, loss of self-esteem, pessimism about the future, etc. [20, 21, 22, 23, 24, 25]. At the same time, however, as the unemployed have a lower opportunity cost of their time, they may choose to invest in health through healthy lifestyles [26], or improve their mental health by doing volunteer work, although the psychological benefits of volunteering depend on factors such as reciprocity and the time devoted to volunteer [27]. It is also expected that work-related diseases will be reduced when unemployment increases [28, 29].

In this paper we provide new evidence about the impact of unemployment (particularly long-term unemployment) on overall health and on mental health by using microdata for the Spanish adult population. But, beyond the mere effect of unemployment on health, our main interest is to analyse the differential association between both variables in both pre-crisis and current-crisis periods. A priori, we cannot expect an unambiguous effect. On the one hand, with the crisis the situation of the unemployed becomes ‘normal’, so the stigma that could harm mental health disappears, whereas if unemployment is rare, the perception of low self-esteem and isolation may be amplified [30]. This effect is supported by some studies showing unemployment as a stronger risk factor when it is rare, for all-cause mortality [31], hospital-treated non-fatal suicidal behaviour [32, 33] or, more recently, suicides [30]. However, the higher the unemployment, the worse the perspectives of getting a job and the more precarious is the future as a worker. In this sense, we would expect further deterioration of jobless people’s health during the economic recession compared to previous times of economic upturn. This effect was confirmed by Preti and Miotto for Italy. They find that a rise in suicide rates is accompanied by a concurrent rise in unemployment rate percentage [34]. Finally, we can also find in previous literature that the level of unemployment seems to have no major influence on the mortality risk [35].

Spain is experiencing a lasting and severe economic crisis. The fall of GDP, the rise of public debt and the high public deficit highlight the gravity of the Spanish economic situation. But the most significant feature in the Spanish crisis is the increasing unemployment rate, which rose from 8.5 % in early 2007 to 27.2 % in the first quarter of 2013 [36], exceeding the rate of any other country in the European Union. In the same period, the percentage of unemployed who have been looking for a job for over a year (long-term unemployed) rose from 21.2 to 56.3 % [36]. These data widely justify addressing the relationship between unemployment, particularly long-term joblessness, and health. Besides, there is little relevant literature on this subject for the Spanish case, singularly from the beginning of the current economic crisis. In the 1990s, Benavides et al. [37], by using data from health surveys, showed a positive association between unemployment, ill health and more use of health services, but found that this association is less clear where high unemployment rates can be considered a long-standing phenomenon. Furthermore, Tapia [38] found a positive relationship between unemployment and mortality and showed how mortality rates increase when unemployment decreases during economic expansion periods. More recently, Pascual and Rodríguez [39] showed that being unemployed during the crisis tends to improve the self-assessed health (SAH) for people living in Catalonia, whereas the effect of unemployment on mental health seems to be unrelated to the economic crisis.

Materials and methods

We use microdata from the Spanish national health survey (SNHS) for two periods: 2006 (before the start of the crisis) and 2011–2012 (during the crisis) [40, 41]. Both surveys, which are comparable to other European health databases, include very similar questions. National health surveys employ a multistage, stratified-random design to identify samples of adults. We have restricted the selected sample to the working-age population (16–65 years old). As our main interest focuses on the relationship between long-term unemployment (over 1 year) and health, we use a restricted subsample only composed of employed and long-term unemployed (n = 13,663 for 2006, and n = 9,495 for 2011–2012). However, as a complementary analysis we also analyse the impact of unemployment on health from a wider perspective, thus using the sample including employed and all unemployed workers (n = 15,324 for 2006 and n = 10,855 for 2011–2012).

With the aim of checking the impact of unemployment on overall and mental health, we employ matching methods. Once this issue is addressed, we will test if there is an incremental effect of unemployment on health as derived from the economic recession. In order to test this effect, we will use difference-in-difference (DiD) techniques. These methods have been previously used to disentangle effects of unemployment on health [42].

Estimation of the impact of unemployment on health in 2006 and 2011–2012: matching methods

We use matching methods based on propensity score [43] separately for 2006 and 2011–2012. Probit regressions are used to estimate the probability of being unemployed for more than 12 months (‘treated’) as a function of the observable covariates vector X associated with unemployment for each year. The parameter of interest to estimate is the average treatment effect (ATT) of unemployment on unemployed. It is defined as
$${\text{ATT}} = E(Y_{1} - Y_{0} |D = 1) = E(Y_{1} |D = 1) - E(Y_{0} |D = 1),$$
where Y represents health, subscripts 1 and 0 mean unemployed and employed, respectively, and D = 1 means unemployed. The second term on the right side of Eq. (1) is the counterfactual: what the health level of an unemployed person would be if he/she had a job. Several assumptions need to be made in order to identify average unobserved counterfactuals. It is assumed that all the relevant differences between treated and non-treated are captured in the X vector. A common support condition is also imposed on the treated units. Treated units whose probability of being treated is larger than the largest p in the non-treated pool would be left unmatched. We use different matching methods (k-nearest neighbours, with k from 1 to 4—approximately the sample ratio between non-treated and treated—within calipers equal to 0.05, and a kernel with a normal distribution) to check for robustness. We report the results for the kernel with a normal distribution. As a complementary analysis we also estimate the matching models for the full sample of employed and unemployed (both short- and long-term).

Estimates of the incremental crisis effect: DiD

An estimate of the effect of the economic crisis on the health impact of long-term unemployment may be obtained by using a DiD technique. We estimate a regression model with the pooled data of both health surveys. Controlling by X covariates, the model includes two main fixed effects, one for the crisis (λ) and another for the employment status (δ), as well as the interaction between them (γ):
$$Y_{idt} = \alpha + \delta {\text{Unemp}}_{it} + \lambda t + \gamma ({\text{Unemp}}_{it} *t) + X_{it}^{{\prime }} \beta_{t} + \varepsilon_{idt} ,$$
where t = 0 means 2006, t = 1 means 2011–2012, and subscript d stands for the employment status. The effect of X variables is assumed to be different in both years. The unbiasedness of the structural estimators depends on the parallel paths assumption. In order to make that assumption as plausible as possible, we included in X all the covariates that could have an influence on health and could be related to the employment status before the crisis and during the crisis. Under the usual hypothesis on the stochastic ε term (mean zero, independent of the regressors), the parameters δ and γ provide information on the effects of unemployment on health before (δ) and during (δ + γ) the economic crisis.
Alternatively, we estimate the model including all unemployed, assuming that the impact on health of different lengths of unemployment may be different:
$$Y_{idt} = \alpha + \sum\limits_{k = 1}^{4} {\delta_{k} } {\text{Unemp}}_{kit} + \lambda t + \sum\limits_{k = 1}^{4} {\gamma_{k} } ({\text{Unemp}}_{kit} *t) + X_{it}^{{\prime }} \beta_{t} + \varepsilon_{idt}$$
where k = 1, 2, 3 and 4 stand, respectively, for unemployed who never worked, those who have been unemployed for less than 6 months, those who have been unemployed for a period between 6 and 12 months and, finally, for long-term unemployed.

Definition of variables

Overall health is proxied by self-assessed health (SAH). The SAH question is formulated as follows in the SNHSs: ‘During the last 12 months, would you say that your health status has been very good, good, fair, poor, very poor?’. Our variable which will take the value one if the individual declares his/her health as fair, poor or very poor, and zero if health is perceived as good or very good. This categorization has been used in previous studies [44, 45].

We consider that mental health risks linked to unemployment may operate in both the short and long term. As shown by Lucas et al. [46], the effect of unemployment on life satisfaction lasts for some time, but the unemployed quickly seem to be mentally adapted to their new status. Furthermore, as was mentioned above, in the context of economic crisis the social stigma of unemployment that could harm mental health tends to fade away. Thus, mental health is represented by two variables: first, a dummy variable which indicates the presence of chronic depression, anxiety or other mental problems during the previous year, which is used as a proxy of permanent mental health (Pmhealth); second, we use the Goldberg index [47], which represents short-term mental health risk and is frequently used in clinical medicine. This variable (Rmhealth) is computed by using the answers to a 12-item set of questions (see Table A1, Supplementary Material). Each question has four possible answers, which are recoded as 0 = ‘no problem’ or 1 = ‘with problems’. The final dummy takes the value 1 if the person has three or more positive answers to the Goldberg 12-item scale questionnaire (which is shown in Table A1, Supplementary Material). This categorization has been previously used in related literature [21].

In both models (matching and DiD regression) the X vector of covariates includes age, sex, education and region. The variable Female, taking the value one for women and zero for men, represents sex. Educational level is categorized by means of five dummies: primary education or below (Ed1, reference category), compulsory secondary education (Ed2), non-compulsory and pre-university secondary education (Ed3), specific labour training requiring non-compulsory secondary education (Ed4) and university graduate (Ed5). Household income could not be considered as a regressor because it is not available for the SNHS 2011–2012. We also include a set of dummies representing the region of residence, with Andalusia acting as the reference category. Regional dummies may act as a proxy of the availability of re-employment opportunities in the geographical area [48]. Furthermore, the regional factor may be relevant as regional public authorities can implement social policies aimed at moderating adverse effects of unemployment and precarious work on health [49].

Finally, in the matching probit equation for propensity of unemployment we control for ‘permanent’ health-related conditions. As has been discussed in previous studies [42, 50, 51], the causal relationship between unemployment and health is, a priori, bidirectional, as remaining jobless may increase the risk of illness, but also some conditions may affect the probability of being unemployed. Therefore, we include a dummy equal to one if the person declares that he/she suffers at least one of twelve chronic diseases: osteoarthritis, arthritis or rheumatism, chronic allergy, asthma, thyroid problems, heart disease, cervical hernia, lower back hernia, stomach ulcer, skin diseases, constipation, headache and haemorrhoids. As a robustness check, we also estimate the models excluding chronic conditions.

Definitions of all the variables are shown in Table 1. The results are detailed in the following section. All calculations were made with Stata12 software [52].
Table 1

Definition of variables and descriptive statistics




2006 (n = 15,324)

2011–2012 (n = 10,855)



Labour status


1 if the person declares to be employed

88.4 %

75.9 %


1 if the person declares to be unemployed and he/she has never worked

0.6 %

1.0 %


1 if the person declares to be unemployed for 6 months or less

5.1 %

7.0 %


1 if the person declares to be unemployed for a period between 6 months and 1 year

1.7 %

3.8 %


1 if the person declares to have been unemployed for 1 year or more

4.2 %

12.1 %

Overall health


Self-assessed health: 1 if fair, poor or very poor; 0 if good or very good

24.9 %

19.2 %

Mental health


1 if the person declares chronic depression, anxiety or other mental problems during the previous year

11.4 %

7.6 %

Short-term mental health risk (Goldberg index)


1 if the person has three or more positive answers to the Goldberg 12-item scale questionnaire

18.3 %

19.3 %


Age in years

40.49 (10.91)

42.02 (10.78)


1 if female

52.3 %

45.5 %



Primary education or below (reference category)

29.0 %

10.0 %


Compulsory secondary education

22.0 %

45.0 %


Non-compulsory and pre-university secondary education

16.1 %

13.7 %


Specific labour training

9.3 %

9.0 %


University graduate

23.0 %

22.3 %

Region (autonomous communities)


Andalucía (reference category)

7.9 %

12.5 %



9.0 %

4.0 %



2.8 %

3.5 %



6.9 %

3.9 %



4.2 %

5.5 %



5.5 %

3.1 %


Castilla y León

3.8 %

5.6 %


Castilla-La Mancha

3.3 %

3.8 %



9.2 %

10.7 %


Comunidad Valenciana

6.3 %

8.8 %



2.8 %

4.4 %



10.2 %

5.4 %



8.0 %

10.2 %



6.3 %

4.1 %



5.9 %

3.0 %


País Vasco

4.0 %

5.9 %


La Rioja

2.5 %

3.4 %


Ceuta and Melilla

1.7 %

2.2 %

Chronic conditions


1 if the person suffers from any chronic illness from a list of 12 conditions

56.4 %

48.2 %

Sample of employed and unemployed people aged 16–65

aFor the categorical variables, data are % in the category; for continuous variables, data include standard deviation in brackets

bThe categories do not add up to one because there are some persons with missing education level


Table 1 shows descriptive statistics for all the variables for the whole sample of people aged 16–65, employed and unemployed, for 2006 and 2011–2012. It is worth noting that the composition of both samples by education level and geographical location differs. As the samples are truncated at 65 years of age, a substantially higher proportion had attained the compulsory educational level in 2011–2012 than in 2006. Besides that, a number of people with very low education, who had been working in unskilled jobs in the building sector during the economic boom, might have left the labour market during the crisis. These changes in the composition of the active population after the crisis aftermath would induce some changes in the regional composition of the sample, too.

Table 2 shows basic descriptive statistics of health indicators by labour status before and during the crisis. The unemployed are classified in five categories according to the duration of unemployment. As may be observed, SAH of Spaniards participating in the labour market, employed or unemployed, is better in 2011–2012 than in 2006, despite the severity of the Spanish economic recession. The percentage of people declaring bad health drops from 24.9 to 19.2 %. Also the percentage of people declaring to have suffered depression, anxiety or mental problems in the last 12 months is lower in 2011–2012 than in 2006. These results, which may seem paradoxical, have also been observed with survey data from Catalonia [39]. They may reflect that health is being assessed in relative terms. Thus, in the context of economic problems and high unemployment, health would rank lower among individuals’ concerns. For the short-term mental health risk variable, however, the total percentage of individuals at risk in 2011–2012 is slightly higher than in 2006, mostly due to deterioration of this health proxy for workers who have recently lost their job (unemployment duration shorter than 6 months), and for unemployed persons who have been looking for a job for more than 1 year, as suggested by Table 2.
Table 2

Health indicators by labour status before and during the economic crisis

Health outcome

Bad health (%)a

Mental problems in last 12 months (%)b

Mental health risk in the short-term (%)c


















































All the reductions of the percentages between 2006 and 2011–2012 are statistically significant

aPercentage declaring that their self-assessed health in the last 12 months is fair, bad or very bad

bPercentage declaring that they have had chronic depression, anxiety or other mental problems in the last 12 months

cPercentage declaring three or more positive answers to the Goldberg items

According to Table 2, SAH and mental health seem to be worse among unemployed people (except for those who have never worked) than among employed people. The longer the unemployment period, the wider the gap. However, descriptive results shown in Table 2 could be biased estimators of group differences because of compositional effects. The groups of employed and unemployed by duration differ significantly by sex, age, educational level, region of residence and health conditions.

Table 3 reports the ATT estimates for 2006 and 2011–2012 by using kernel estimates with a Gaussian kernel, for the subsample of employed people (‘untreated’) and unemployed people (‘treated’). No individuals are excluded because of common support requirements in 2006 and only one is excluded in 2011–2012. We firstly show the results for the impact of long-term unemployment on health. The estimated probit equations for propensity score are displayed in Table A2 (Supplementary Material). The results obtained from alternative matching approaches are shown in Table A3 (Supplementary Material). All results are very robust to the matching method. The second column of Table 3 contains the sample data corresponding to the unemployed. The third column shows the estimated health of the unemployed if they had been working (counterfactual). The fourth column is the difference between the two previous columns and it estimates the impact of unemployment on health. This estimate is called the average treatment effect (ATT) as it measures the loss in health that may be attributable to unemployment. Finally, the fifth column shows the statistical significance of the ATT estimates. The left and right sides of the table show, respectively, results for 2006 and 2011–2012.
Table 3

Impact estimates of unemployment on health 2006 and 2011–2012

Long-term unemployment

Dependent variablea

Pre-crisis (2006)

Crisis (2011–2012)


E(Y1|D = 1) (%)b

Counterfactual E(Y0|D = 1) (%)c

Impact (ATT) (pp)d



E(Y1|D = 1) (%)

Counterfactual E(Y0|D = 1) (%)

Impact (ATT) (pp)


(SAH) % bad health









(Pmhealth) % mental health problems









(Rmhealth) % mental health risk in the short term (Goldberg)









Total unemployment (short- and long-term)

Dependent variable

Pre-crisis (2006)

Crisis (2011–2012)


E(Y1|D = 1) (%)

Counterfactual E(Y0|D = 1) (%)

Impact (ATT) (pp)



E(Y1|D = 1) (%)

Counterfactual E(Y0|D = 1) (%)

Impact (ATT)


(SAH) % bad health









(Pmhealth) % mental health problems









(Rmhealth) % mental health risk in the short term (Goldberg)









Matching methods. Propensity score with Gaussian kernel

*** p < 0.01. Control variables are age, sex, education, region and chronic conditions

aMatching models to estimate the effect of long-term unemployment on overall health (SAH), on mental health problems (Pmhealth) and on mental health risk in the short term (Rmhealth)

bSample data corresponding to unemployed

cEstimated data for unemployed if they had been working (counterfactual)

dAverage treatment effect (ATT) = column 2 − column 3. It measures the loss in health attributable to unemployment; as it is a difference between 2 %, it is expressed as percentage points

eRatio to determine statistical significance of the ATT estimates

The estimates show that one or more years of unemployment tend to significantly deteriorate the overall and mental health before the economic recession and also during the crisis. Once we account for the X covariates, long-term unemployment increases the probability of showing mental health risk by 10.4 percentage points (pp) before the crisis, and by 16 pp for the period 2011–2012. The ATT for mental health problems in the last year rises from 5.5 pp in 2006 to 9.7 pp in 2011–2012. The ATT when SAH is considered was 11.8 in 2006 and 10.7 in 2011–2012. Thus, the effects of long-term unemployment on mental health seem to be larger in times of economic downturn, whereas this association is not found in SAH. It is plausible that self-reported health is not capturing so much real changes in health but changes in the perceived level of health, which could be affected by the fact mentioned above that health is being assessed in relative terms.

As mentioned in the previous section, we initially consider in the X vector of covariates the presence of chronic conditions, as poor health may increase the risk of becoming unemployed. However, this variable could be endogenous as some chronic diseases could also worsen when a worker loses his/her job. To deal with this problem we did some robustness checks by excluding the dummy chronic from the model. The results, which are fully reported in Table A4 (Supplementary Material), which are similar to those shown in Table 3.

The bottom part of Table 3 collects the results from matching models estimated for the full sample of employees and unemployed, both short- and long-term (the corresponding probit estimates are reported in Table A5 of the Supplementary Material). It shows also the significant effects of unemployment on SAH and both dimensions of mental health, although these are much lower with regard to those linked to long-term joblessness. Furthermore, the results again suggest that the negative impact of unemployment on mental health may be higher during economic recessions.

The DiD estimates mostly confirm these results. Table 4 shows the estimates corresponding to the health impact of long-term unemployment. Therefore, the estimated effects correspond to Eq. (2) in the “Materials and methods” section. The parameter λ shows the change in each of the health measures that occurred during the crisis. The parameter δ accounts for the effects of long-term unemployment on health. Finally, the parameter of highest interest is γ, which measures the change in the effect of unemployment on health after the crisis compared to the effect in 2006. We estimate two alternative models in order to check how the X vector of covariates may alter the results.
Table 4

DiD estimates of the health impact of long-term unemployment before the economic crisis and changes during the economic crisis

Dependent variable


Coefficients (95 % CI)a

Pseudo (R2)b

Pseudo (R2)c


Model without controlsb

Full modelc

(SAH) % bad health

(λ) Change in SAH after the crisis

−0.28 (−0.45; −0.12)

−0.19 (−0.42; +0.04)




(δ) Effect of unemployment in the base year (2006)

0.40 (0.30; 0.50)***

0.32 (0.22; 0.42)***


(γ) Change in the effect of unemployment on SAH after the crisis aftermath

0.06 (−0.07; +0.18)

0.04 (−0.09; +0.17)


(Pmhealth) % mental health problems

(λ) Change in Pmhealth after the crisis

−0.20 (−0.42; +0.01)

−0.22 (−0.52; +0.08)




(δ) Effect of unemployment in the base year (2006)

0.25 (0.14; 0.37)***

0.21 (0.09; 0.32)***


(γ) Change in the effect of unemployment on Pmhealth after the crisis aftermath

0.34 (0.19; 0.49)***

0.35 (0.19; 0.50)***


(Rmhealth) % mental health risk in the short term (Goldberg)

(λ) Change in Rmhealth after the crisis

−0.02 (−0.18; −0.15)

0.0073 (−0.23; +0.25)




(δ) Effect of unemployment in 2006

0.36 (0.25; 0.47)***

0.3476 (0.24; 0.46)***


(γ) Change in the effect of unemployment on Rmhealth after the crisis aftermath

0.20 (0.07; 0.34)***

0.2027 (0.07; 0.34)***


Differences-in-differences model to estimate the effect of long-term unemployment on overall health (SAH), on mental health (Pmhealth) problems and on mental health risk in the short term (Rmhealth). The three dependent variables are defined in the “Materials and methods” section

*** Significant at 1 % (p < 0.01)

aEstimated effects correspond to Eq. (2). The parameter of highest interest is γ. It measures the change in the effect of unemployment on health after the crisis compared to the effect in 2006. Point estimates and 95 % CI

bModel controlling only for age and sex

cFull model that adjusts for age, sex, education and region allowing different effects in each year (2006 and 2011)

As shown in Table 4, long-term unemployment has a significant impact on both overall and mental health. Moreover, the interaction term γ is positive and significant for mental health models, suggesting that negative effects of unemployment on people’s psychological health are intensified because of the economic crisis. That intensification is higher for mental problems—e.g. depression and anxiety—in the last year than for the short-term mental health risk (Goldberg index). However, SAH does not seem to worsen more with unemployment in times of economic crisis than before the crisis aftermath. It may also be verified that estimates barely depend on the vector of covariates. Our results are consistent with the hypothesis suggested by Karasek and Theorell [53], in the sense that economic recessions may encourage individuals to anticipate stressful situations, including job loss and difficulty in dealing with financial obligations.

We have also estimated alternative DiD models that include all unemployed and the corresponding dummies for different periods of unemployment (Table A7, Supplementary Material). The obtained results are similar to those shown in Table 4. Except for those who have never worked, unemployment negatively influences overall health and mental health. The impact on overall health increases with the length of unemployment. Like in our base model, which was restricted to long-term unemployment, the impact on overall health does not seem to change in times of crisis. Moreover, the impact on mental conditions is larger after the crisis, as in the base model, only for the long-term unemployed. The effects on the Goldberg index become more serious after the crisis for those who are unemployed for less than 6 months and also for those who are unemployed for more than 12 months.


Our results are in line with previous work showing a positive relationship between unemployment rates and mental health risks [34], and are also consistent with those found by Gili et al. [54], who show how the economic crisis has significantly increased the frequency of mental health disorders among primary care users in Spain, particularly among families experiencing unemployment. However, the results here reported differ from those found by previous Spanish studies that check how the impact of unemployment on health varies depending on the economic context [37, 38, 39], which could be partially explained by differences in the definition of health variables.

Our study has a number of limitations. First, cross-sectional data do not allow for exploration of causal relationships between unemployment and health as longitudinal databases do. Previous research with panel data from the Spanish sample of the EU-SILC did not confirm the significant effect of unemployment on SAH for the period 2007–2010 [55]. A similar result is found by Böckerman and Ilmakunnas [42], who use panel data from the European Community Household Panel for Finland. They show that the event of unemployment does not matter as such for SAH and conclude that the cross-sectional negative relationship between unemployment and SAH is related to the fact that persons who have poor SAH are being selected for the pool of the unemployed. Nevertheless, the EU-SILC waves do not include specific information about mental health—although it may be assumed that SAH also includes the individuals’ rating of their mental health—so the impact of unemployment on Spaniards’ psychological health with longitudinal microdata cannot be verified.

Second, it has to be noted that some relevant determinants of unemployment may be excluded from the X vector of covariates, such as the occupational sector or the eligibility for public subsidies, and thus our estimates may be biased. The omission of other relevant variables may also bias the estimates. This is the case of household income, which is not available in the SNHS for 2001–2012. The effects of unemployment on health could in fact be reflecting the impact of the lack of income. However, this problem will be mitigated as long as omitted variables operate similarly in both periods. In addition, the dummy for chronic conditions included in the X vector could be endogenous, and the results consequently would be biased. To deal with this problem we did some robustness checks, with satisfactory results.

Third, the proxies for overall health and permanent mental health are constructed from survey questions, which refer to the last 12 months. Therefore, when we use the full sample of unemployed (including short- and long-term jobless people), we are searching for associations between variables which are defined for different reference periods. However, this problem disappears when the analysis focuses on the impact of long-term unemployment on health.

Fourth, the self-reported definition of unemployed could bias the estimation results owing to self-selection, if those who have been unemployed for a long time tend to classify themselves as inactive.

Fifth, although in the DiD estimation we adjusted for all the measured covariates that might be correlated to labour status before the crisis and/or during the crisis, we cannot ensure that the parallel paths assumption is satisfied. Finally, we cannot reach conclusions about the overall impact of the economic crisis on Spaniards’ health, as the recession has not yet finished.


We provide new and robust evidence about the significant impact of (particularly long-term) unemployment on overall health and mental health with individual-level data for Spain. We also investigate whether the effects of unemployment on health have increased or been tempered as a consequence of the economic crisis, confirming that psychological effects of unemployment are more serious in times of recession. Our results may lead one to conclude that anxiety and stress about the future associated with unemployment could have greater impact on individuals’ health than the palliative effects of social protection provided during the economic recession. Although economic effects of job loss may be softened by the safety net of the welfare state, the maximum duration of unemployment benefits is 2 years, far less than the duration of the economic recession. After the maximum period of unemployment benefits, many households are forced to take part into the minimum income programs offered by the regional administrations. In this sense, recent research shows how physical and mental health problems were better for those individuals benefiting from those programs who had taken part in work-related activities, thus suggesting that welfare-to-work policies may have positive unintended health effects [56].

It also has to be noted that Spain has adopted strict austerity measures in recent years, which include significant cuts in health spending and some reductions of the unemployment benefits. Furthermore, it is likely that additional cuts will occur in the near future. Therefore, the incremental effect on health shown here could be amplified when the recession comes to an end.

The results could also point to the need for preventing health deterioration in vulnerable groups such as the unemployed, and also for monitoring specific health risks that arise in recessions, such as psychological problems.

Supplementary material

10198_2014_563_MOESM1_ESM.doc (174 kb)
Supplementary material 1 (DOC 174 kb)


  1. 1.
    Brenner, M.H., Mooney, A.: Unemployment and Health in the context of economic change. Soc. Sci. Med. 17(16), 1125–1138 (1983)CrossRefPubMedGoogle Scholar
  2. 2.
    Ruhm, C.J.: Macroeconomic Conditions, Health and Mortality. NBER Working Paper No. 11007 (2004)Google Scholar
  3. 3.
    Stuckler, D., Basu, S., Suhrcke, M., Coutts, A., McKee, M.: The public health effect of economic crises and alternative policy responses in Europe: an empirical analysis. Lancet 374, 315–323 (2009)CrossRefPubMedGoogle Scholar
  4. 4.
    CSDH: Closing the Gap in a Generation: Health Equity Through Action on the Social Determinants of Health. Final report of the Commission on Social Determinants of Health. World Health Organization, Geneva (2008)Google Scholar
  5. 5.
    Ruhm, C.J.: Are recessions good for your health? Q. J. Econ. 115(2), 617–650 (2000)CrossRefGoogle Scholar
  6. 6.
    Gerdtham, U.G., Ruhm, C.J.: Deaths rise in good economic times: evidence from the OECD. Econ. Hum. Biol. 4(3), 298–316 (2006)CrossRefPubMedGoogle Scholar
  7. 7.
    Egan, M.L., Mulligan, C.B., Philipson, T.J.: Adjusting Measures of Economic Output for Health: Is the Business Cycle Countercyclical? NBER Working Paper No. 19058 (2013)Google Scholar
  8. 8.
    Neumayer, E.: Recessions lower (some) mortality rates: evidence from Germany. Soc. Sci. Med. 58(6), 1037–1047 (2004)CrossRefPubMedGoogle Scholar
  9. 9.
    Buchmueller, T., Grignon, M., Jusot, F.: Unemployment and Mortality in France, 1982–2002. CHEPA Working Paper 07-04 (2007)Google Scholar
  10. 10.
    Ross, C., Mirowsky, J.: Does employment affect health? J. Health Soc. Behav. 36, 230–243 (1995)CrossRefPubMedGoogle Scholar
  11. 11.
    Dooley, D., Fielding, J., Levi, L.: Health and unemployment. Annu. Rev. Public Health 17, 949–965 (1996)CrossRefGoogle Scholar
  12. 12.
    Platt, S., Pavis, S., Akram, G.: Changing labour market conditions and health: a systematic literature review (1993–98). In: Research Unit in Health and Behavioural Change. University of Edinburgh Medical School, Edinburgh (1999)Google Scholar
  13. 13.
    Kasl, S., Jones, B.: The impact of job loss and retirement on health. In: Berkman, L., Kawachi, I. (eds.) Social Epidemiology, pp. 118–136. Oxford University Press, Oxford (2000)Google Scholar
  14. 14.
    Darmon, N., Drewnowski, A.: Does social class predict diet quality? Am. J. Clin. Nutr. 87(5), 1107–1117 (2008)PubMedGoogle Scholar
  15. 15.
    Mossakowski, K.N.: Is the duration of poverty and unemployment a risk factor for heavy drinking? Soc. Sci. Med. 67, 947–955 (2008)CrossRefPubMedGoogle Scholar
  16. 16.
    Fagan, P., Shavers, V., Lawrence, D., Gibson, J.T., Ponder, P.: Cigarette smoking and quitting behaviors among unemployed adults in the United States. Nicotine Tob. Res. 9(2), 241–248 (2007)CrossRefPubMedGoogle Scholar
  17. 17.
    Grayson, J.P.: Health, physical activity level, and employment status in Canada. Int. J. Health Serv. 23(4), 743–761 (1993)CrossRefPubMedGoogle Scholar
  18. 18.
    World Bank: Averting A Human Crisis During the Global Downturn. World Bank, Washington (2009)Google Scholar
  19. 19.
    Dávila, C.D., González, B.: Crisis económica y salud. Gac. Sanit. 23, 261–265 (2009)CrossRefGoogle Scholar
  20. 20.
    Catalano, R., Dooley, D., Rook, K.: A test of reciprocal risk between undesirable economic and noneconomic life events. Am. J. Community Psychol. 15, 633–651 (1987)CrossRefPubMedGoogle Scholar
  21. 21.
    Artazcoz, L., Benach, J., Borrell, C., Cortès, I.: Unemployment and mental health: understanding the interactions among gender, family roles, and social class. Am. J. Public Health 94(1), 82–88 (2004)CrossRefPubMedCentralPubMedGoogle Scholar
  22. 22.
    Vinokur, A.D., Price, R.H., Caplan, R.D.: Hard times and hurtful partners: how financial strain affects depression and relationship satisfaction of unemployed persons and their spouses. J. Pers. Soc. Psychol. 71, 166–179 (1996)CrossRefPubMedGoogle Scholar
  23. 23.
    Paul, K.I., Moser, K.: Unemployment impairs mental health: meta-analyses. J. Vocat. Behav. 74, 264–282 (2009)CrossRefGoogle Scholar
  24. 24.
    Goldman-Mellor, S.J., Saxton, K.B., Catalano, R.C.: Economic contraction and mental health. A review of the evidence, 1990–2009. Int. J. Ment. Health 39(2), 6–31 (2010)CrossRefGoogle Scholar
  25. 25.
    Pharr, J.R., Moonie, S., Bungum, T.J.: The impact of unemployment on mental and physical health, access to health care and health risk behaviors. ISRN Public Health (2012). doi:10.5402/2012/483432 Google Scholar
  26. 26.
    Laufey, T., Corman, H., Noonan, K. et al.: Are Recessions Good for Your Health Behaviors? Impacts of the Economic Crisis in Iceland. NBER Working Paper 18233 (2012)Google Scholar
  27. 27.
    Jenkinson, C.E., Dickens, A.P., Jones, K., Thompson-Coon, J., Taylor, R.S., Rogers, M., et al.: Is volunteering a public health intervention? A systematic review and meta-analysis of the health and survival of volunteers. BMC Public Health 13, 773 (2013)CrossRefPubMedCentralPubMedGoogle Scholar
  28. 28.
    Boone, J., van Ours, J.C.: Are recessions good for workplace safety? J. Health Econ. 25, 1069–1093 (2006)CrossRefPubMedGoogle Scholar
  29. 29.
    Davies, R., Jones, P., Núñez, I.: The impact of the business cycle on occupational injuries in the UK. Soc. Sci. Med. 69, 178–182 (2009)CrossRefPubMedGoogle Scholar
  30. 30.
    Corcoran, P., Arensman, E.: Suicide and employment status during Ireland’s Celtic Tiger economy. Eur. J. Public Health 21(2), 209–214 (2010)CrossRefPubMedGoogle Scholar
  31. 31.
    Martikainen, P.T., Valkonen, T.: Excess mortality of unemployed men and women during a period of rapidly increasing unemployment. Lancet 348, 909–912 (1996)CrossRefPubMedGoogle Scholar
  32. 32.
    Ostamo, A., Lahelma, E., Lonnqvist, J.: Transitions of employment status among suicide attempters during a severe economic recession. Soc. Sci. Med. 52, 1741–1750 (2001)CrossRefPubMedGoogle Scholar
  33. 33.
    Platt, S., Kreitman, N.: Long term trends in parasuicide and unemployment in Edinburgh, 1968–87. Soc. Psychiatry Psychiatr. Epidemiol. 25, 56–61 (1990)PubMedGoogle Scholar
  34. 34.
    Preti, A., Miotto, P.: Suicide and unemployment in Italy, 1982–1994. J. Epidemiol. Community Health 53, 694–701 (1999)CrossRefPubMedCentralPubMedGoogle Scholar
  35. 35.
    Ahs, A.M., Westerling, R.: Mortality in relation to employment status during different levels of unemployment. Scand. J. Public Health 34, 159–167 (2006)CrossRefPubMedGoogle Scholar
  36. 36.
  37. 37.
    Benavides, F.G., García, A.M., Sáez-Lloret, I., Librero, J.: Unemployment and health in Spain. The influence of socio-economic environment. Eur. J. Public Health 4(2), 103–107 (1994)CrossRefGoogle Scholar
  38. 38.
    Tapia-Granados, J.A.: Recessions and mortality in Spain: 1980–1997. Eur. J. Popul. 21, 393–422 (2005)CrossRefGoogle Scholar
  39. 39.
    Pascual, N., Rodríguez, I.: Effects of unemployment on self-assessed health and mental health: empirical evidence for Catalonia (2006–2012). Masters Project. Masters in Health Economics and Policy, Barcelona Graduate School of Economics (2013)Google Scholar
  40. 40.
    Spanish Ministry of Health. National health survey 2006 (2007) Google Scholar
  41. 41.
    Spanish Ministry of Health. National health survey 2011–2012 (2013)Google Scholar
  42. 42.
    Böckerman, P., Ilmakunnas, P.: Unemployment and self-assessed health: evidence from panel data. Health Econ. 18, 161–179 (2009)CrossRefPubMedGoogle Scholar
  43. 43.
    Rosenbaum, P.R., Rubin, D.B.: The central role of the propensity score in observational studies for causal effects. Biometrika 70(1), 41–55 (1983)CrossRefGoogle Scholar
  44. 44.
    Kunst, A.E., Bos, V., Lahelma, E., Bartley, M., Lissau, I., Regidor, E., Mielck, A., Cardano, M., Dalstra, J.A., Geurts, J., Helmert, U., Lennartsson, C., Ramm, J., Spadea, T., Stronegger, W.J., Mackenbach, J.P.: Trends in socioeconomic inequalities in self-assessed health in 10 European countries. Int. J. Epidemiol. 34(2), 295–305 (2005)CrossRefPubMedGoogle Scholar
  45. 45.
    Hernández-Quevedo, C., Masseria, C., Mossialos, EA.: Analysing the Socioeconomic Determinants of Health in Europe: New Evidence from EU-SILC. Eurostat Working Paper (2010)Google Scholar
  46. 46.
    Lucas, R.E., Clark, A.E., Georgellis, Y., Diener, E.: Unemployment alters the set point for life satisfaction. Psychol. Sci. 15, 8–13 (2004)CrossRefPubMedGoogle Scholar
  47. 47.
    Goldberg, D.: Mental health priorities in a primary care setting. Ann. N. Y. Acad. Sci. 310, 65–68 (1978)CrossRefPubMedGoogle Scholar
  48. 48.
    Turner, J.B.: Economic context and the health effects of unemployment. J. Health Soc. Behav. 36, 213–229 (1995)CrossRefPubMedGoogle Scholar
  49. 49.
    Karlsdotter, K.: Análisis de la Influencia de las Desigualdades Socioeconómicas Regionales Sobre la Salud de Los Andaluces y Españoles en el Período 2004–2007. PhD Dissertation, University of Granada (2012)Google Scholar
  50. 50.
    Novo, M.: Young and unemployed: does the trade cycle matter for the health. In: A Study of Young Men and Women During Times of Prosperity and Times of Recession. Umeå: Umeå Universitet (2000)Google Scholar
  51. 51.
    Stewart, J.M.: The impact of health status on the duration of unemployment spells and the implications for studies of the impact of unemployment on health status. J. Health Econ. 20, 781–796 (2001)CrossRefPubMedGoogle Scholar
  52. 52.
    STATA®. Stata 12 Software (2012)Google Scholar
  53. 53.
    Karasek, R., Theorell, T.: Healthy work: stress, productivity, and the reconstruction of working life. Basic Books, New York (1990)Google Scholar
  54. 54.
    Gili, M., Roca, M., Basu, S., McKee, M., Stuckler, D.: The mental health risks of economic crisis in Spain: evidence from primary care centres, 2006 and 2010. Eur. J. Public Health 23, 103–108 (2013)CrossRefPubMedGoogle Scholar
  55. 55.
    Urbanos, R., González, B.: Desempleo y salud: un análisis de la repercusión de la crisis económica sobre la salud de los españoles. Estudios de Economía Aplicada 31(2), 303–326 (2013)Google Scholar
  56. 56.
    Ayala, L., Rodríguez, M.: Health-related effects of welfare-to-work policies. Soc. Sci. Med. 93, 103–112 (2013)CrossRefPubMedGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Rosa M. Urbanos-Garrido
    • 1
  • Beatriz G. Lopez-Valcarcel
    • 2
  1. 1.Complutense University of MadridMadridSpain
  2. 2.Departamento de Métodos Cuantitativos en Economía y GestiónUniversity of Las Palmas de Gran CanariaLas Palmas de Gran CanariaSpain

Personalised recommendations