The relationship between unemployment and increased risk of morbidity and mortality is well established.1^–8 However, what is less clear from the existing literature is whether the relationship between unemployment and health varies by welfare state and, if so, the extent to which this can be explained through reference to the different approaches to social protection (particularly wage replacement rates) taken by different welfare state regimes. In this paper, we examine the extent to which relative health inequalities between unemployed and employed people vary across 23 European countries and by the different approaches to social protection taken by the five European welfare state regimes (Scandinavian, Anglo-Saxon, Bismarckian, Southern and Eastern).

Unemployment and health

At the individual level, studies have particularly shown that unemployment is associated with worse mental health, including parasuicide.9 10 It has also been linked to higher rates of all-cause mortality6 7 as well as limiting long-term illness,8 and, in some studies, a higher prevalence of risky health behaviours (among young men), including problematic alcohol use and smoking.11 At the area level, rates of unemployment, especially when used as indicators of deprivation, correlate with poorer neighbourhood health,12 and at the country level increases in the unemployment rate have been associated with increased mortality.13 Research has also drawn attention to the contributory role of ill health itself as a factor behind unemployment (direct health selection),5 7 14 and the importance of ill health-related worklessness in terms of socioeconomic health inequalities.15 Studies from various countries have identified poverty as an important intermediary factor in the relationship between unemployment and health.7

Welfare state regimes

Welfare provision, in the form of cash benefits and welfare services, is acknowledged as an important mediatory factor in terms of the relationship between labour market position and health.16^–19 A crucial aspect of welfare provision, and one which most differentiates welfare states, is income maintenance (to prevent poverty),18 particularly during adverse events such as unemployment, old age or long-term sickness absence.

Welfare state regimes place those welfare states that are the most similar (in terms of political tradition, principles, levels of provision, etc) together, emphasising within-regime coherence and between-regime differences.20 There are various competing welfare state regime typologies which emphasise diverse aspects of welfare states such as social expenditure levels, decommodification or political traditions (for an overview see Bambra 22).21^–24 Ferrera’s fourfold typology,25 which focuses on different dimensions of how social benefits are granted and organised, has been highlighted as one of the most empirically accurate welfare state regime typologies.26^–29 Ferrera 25 makes a distinction between the Scandinavian, Anglo-Saxon, Bismarckian and Southern countries (box 1). More recently, the Eastern European countries have begun to be considered as a distinctive regime type.20 22 26 29 30

Welfare state regimes and unemployment protection

Social protection during unemployment varies by welfare state regime. To a large degree this reflects the historical influence of differing political traditions, with those countries experiencing more post-war years of Social Democratic rule providing more generous systems of support.24 Table 1 breaks down the various characteristics of social protection during unemployment in the five different welfare state regimes. In essence, there are three inter-relating principles underpinning provision: universalism, social insurance and means-testing.18 Systems based on universal provision do not make reference to previous contributions or means-testing and are offered to all citizens as long as specific demographic, social or health criteria are fulfilled. Often flat-rate benefits are paid. Under social insurance systems, entitlement to benefits is dependent on previous contributions and in most cases subsequent benefit levels reflect previous earned income. Under means-testing, entitlement is restricted on the basis of income and the (often minimal) financial support is targeted at those in most need, usually after they have exhausted all other means (eg, personal savings or social insurance).31

The welfare provision of different regimes is governed by these three principles in varying ways. For example, to differing degrees of generosity, universalism is more prominent within the Scandinavian welfare states (high population coverage) and the Anglo-Saxon regime (fixed benefit rates for all), whereas social insurance is the key component of provision within the Bismarckian, Southern and Eastern European welfare states. Means-testing is more commonly a characteristic of the Anglo-Saxon welfare states; however, it is also used for social assistance payments in other welfare state regimes. For example, in the UK (Anglo-Saxon) unemployment benefit (contribution-based Job Seekers Allowance) is only payable (for a maximum of 6 months) to those who fulfil the minimum National Insurance contribution requirement within the 2 years before claiming (table 1). Most claimants do not meet this criterion and are therefore reliant on means-tested social assistance benefits, particularly income-based Job Seeker’s Allowance and Income Support.32 However, this mixed approach is also evident in Sweden, where there is a social insurance-based benefit (Unemployment Insurance Benefit) based on past contributions and which pays a benefit as a proportion of previous wages, as well as a means-tested social assistance scheme (Unemployment Assistance Benefit), which pays a (lower) flat rate.32 Similarly, a three-tier system is operated in Germany (Bismarckian): those with a full contribution record receive the full unemployment insurance benefit (Arbeitslosengeld), those with a smaller contribution criteria, receive a means-tested insurance benefit (Arbeitslosenhilfe) whereas those who do not have a sufficient contribution record must rely upon the Sozialhilfe social assistance scheme.32

Unemployment protection in each welfare state regime therefore represents a complex mix of these differing principles. However, there are clear differences by welfare state regime, due to the influence of differing political traditions, in terms of how these principles are operationalised, particularly in terms of the generosity of benefits paid to the unemployed (replacement rates), the qualifying period and conditions, duration of benefit payments and the waiting period before entitlement is activated. In each of these regards, the Scandinavian welfare states are generally more generous than the other welfare state regimes (table 1), particularly compared with the Anglo-Saxon and Eastern European regimes.

Welfare state regimes, unemployment protection and health

Differences in the social protection offered to the unemployed could therefore be an important mediatory factor in the relationship between poverty, unemployment and health.7 This could be very important in terms of helping to develop policy interventions, particularly in terms of income maintenance provision, to improve the health of the unemployed, reduce inequalities between those in and out of work, and thereby potentially reduce the influence of labour market status on health. Indeed, a study comparing means-tested and non-means-tested unemployment benefits in three countries (UK, Germany, USA) found that among the unemployed, those in receipt of non-means-tested benefits had better health than those in receipt of means-tested benefits.3

However, as previous studies of the relationship between unemployment and health have tended to focus either on associations between unemployment and health,7 or changes to the employment status of people and their subsequent health, within one country,5 8 or in a very limited number of similar countries,3 a full examination of the possibly health protective role of different approaches to social protection has not yet been undertaken. Similarly, although there is an emerging comparative social epidemiology literature that examines differences in health by welfare state regime,17 19 22 26 28 29 35^–39 there has to date been little analysis by population sub-group.22 Therefore, in this study we examine the relationship between unemployment and self-reported health in 23 countries through reference to the different approaches to social protection taken by five different welfare state regimes. Specifically, given the differences in social provision by welfare state regime (as described in box 1 and table 1), we test the following two inter-related hypotheses: (1) that the self-reported health of the unemployed will be worse than the employed in all welfare state regimes, and (2) that the unemployed in those welfare state regimes with higher levels of social protection (the Scandinavian and Bismarckian regimes) will have comparatively better self-reported health than those in the other welfare state regimes (Anglo-Saxon, Southern and Eastern).

METHODS

Data

The data source is made up of two independent waves of the European Social Survey (ESS) (merged files from 2002 and 2004), from which we analysed 37 499 individuals (aged 25–60 years) from 21 countries (table 2). The two health outcome variables were self-reported limiting long-standing illness and fair/poor general health. The main objective of the ESS is to provide high-quality data over time about changing social attitudes and values in Europe. The data and extensive documentation are freely available for downloading at the Norwegian Social Science Data Services (NSD) web site (www.nsd.uib.no).

We used two indicators of morbidity available in the ESS: self-reported general health and limiting long-standing illness. Self-reported general health was constructed from a variable asking “How is your (physical and mental) health in general?” Eligible responses were “very good”, “good”, “fair”, “bad” and “very bad”. We dichotomised the variable into “very good or good” health versus “less than good” health (“fair”, “bad” and “very bad”). As for limiting long-standing illness, people were asked if they were hampered in daily activities in any way by any long-standing illness or disability, infirmity or mental health problem. Eligible responses were “yes a lot”, “yes to some extent” and “no”. We dichotomised this variable into “yes” (regardless of whether to some extent or a lot) and “no”. Unemployment was measured by comparing unemployed (including both those currently looking for a job and those who are not) with people in paid work. The question asked in the survey was “what is your main activity, the last 7 days?” Correlation tests between the reporting of employment status in the ESS largely correspond with the Organisation for Economic Co-operation and Development (OECD) rates from 2003 (table 2, last column). People who were currently under education, permanently sick or disabled, retired, doing community or military service were excluded from the analysis along with those doing housework/looking after children. A weight was applied in all analyses to correct for design effects due to sampling designs in countries where not all individuals in the population have an identical selection probability. All analyses were done for men and women separately.

Analysis

Relative health inequalities were calculated applying a series of logistic regression analyses, in which unemployment was introduced as an independent variable, controlled for age, with health outcomes as the dependent variables. Prevalence rates and rate differences were calculated additionally, using direct age standardisation. In addition, to test the robustness of the main findings, three sensitivity analyses were performed. First, the between-regime differences in the relationship between unemployment and health were tested separately for men and women using the interaction “employment status*regime” within a multi-level design. Second, one-way analysis of variance was used to examine whether the between-regime differences in health outcomes (overall prevalence, prevalence among unemployed, rate difference and relative inequalities) were greater than the within-regime differences. Finally, additional adjustments were made for between-regime differences in the prevalence of unemployment (by sex and country) and differences between regimes in terms of the socioeconomic status (education and occupational class) of the unemployed were also examined. These analyses are detailed further in the online appendix.

RESULTS

ORs (along with prevalence rates and rate differences) of ill health are presented in table 3 for men and women within each welfare regime separately (country-specific data are presented in table 1 in the Web-only appendix). All results in this table indicate that unemployed people feel unhealthier than those who report to be employed. This association is significant for all outcomes, with the single exception of men with limiting long-standing illness (OR 1.67) in the Anglo-Saxon welfare regime.

There are also clear differences by welfare state regime. Relative inequalities between employed and unemployed were largest in the Anglo-Saxon (men OR_PH 2.97, 1.92 to 4.60; women OR_LI 2.73, 1.50 to 4.95 and OR_PH 2.78, 1.63 to 4.73) Bismarckian (men only OR_LI 2.21, 1.74 to 2.79 and OR_PH 2.72, 2.21 to 3.35) and Scandinavian (women only OR_LI 2.28, 1.71 to 3.03 and OR_PH 2.99, 2.34 to 4.00) regimes, and smallest in the Southern (men OR_PH 1.82, 1.35 to 2.46; women OR_LI 1.52, 1.03 to 2.25 and OR_PH 1.66, 1.31 to 2.11) and Eastern (women only OR_LI 1.65, 1.24 to 2.10 and OR_PH 1.76, 1.38 to 2.25) welfare state regimes.

According to the size of rate differences and ORs, it appears that the negative health experiences of being unemployed are particularly strong for women within the Anglo-Saxon (OR_LI 2.73 and OR_PH 2.78) and Scandinavian (OR_LI 2.28 and OR_PH 2.99) welfare regime. Although the ORs of men’s reporting of limiting long-standing illness do not show a distinct pattern (except from the non-significant results in the Anglo-Saxon regime), the reporting of poor general health within the Anglo-Saxon regime again demonstrates the largest ORs.

The sensitivity analyses (presented in the online appendix) show that welfare state regimes are strongly related to the association of unemployment and women’s health and, in terms of health outcomes, that within-welfare state regime variance is significantly smaller than between-welfare state regime variance for measures of prevalence (but not with regard to rate differences and relative inequalities). The association between rate differences and ORs was more evident for women than for men. The additional adjustments made for the prevalence of unemployment confirmed the main findings, that the association between unemployment and health varies by welfare state regime, and in addition, a high correlation as found between the original ORs and the ORs adjusted for the prevalence of unemployment (r = 0.85 or higher). This suggests that it is not the higher prevalence of unemployment in some welfare states which has driven the observed differences in the health of the unemployed by welfare state regime. The sensitivity analyses also found that unemployed men and women were more likely to be from the lower socioeconomic groups than employed people in all welfare state regimes.

DISCUSSION

Our study has found that the relationship between unemployment and health is consistent across all 23 European countries with the unemployed in each country reporting worse self-reported health than the employed (either LI, PH or both). This is in keeping with our first hypothesis and in line with the majority of the existing research literature.5 7 8 40 For example, a longitudinal Swedish study found that self-reported physical health decreased with the advent of unemployment and that poorer self-reported physical health increased the likelihood of future unemployment.5 Similarly, a longitudinal study of UK men found an increased risk of limiting long-standing illness among the unemployed.8 It seems, therefore, that even though the levels of social protection offered to the unemployed vary by welfare state (and welfare state regime), in all countries, a relationship exists between unemployment and poorer self-rated health. This suggests that current wage replacement rates, even in the more generous welfare states, are not sufficient to overcome the financial effects of unemployment on health. On the other hand, it may indicate the importance for health of the non-financial losses associated with unemployment (eg, social isolation), as demonstrated in Rudas et al’s40 study of unemployed Italian workers who despite receiving a 100% replacement rate still reported elevated levels of physical and mental morbidity.7

Although we have found a consistent cross-European relationship between unemployment and poorer self-reported health, we have also identified differences in the magnitude of the relationship by welfare state regime. Specifically, we have found that relative inequalities are largest in the Anglo-Saxon, Bismarckian (men only), and Scandinavian (women only) regimes, and smallest in the Southern and Eastern (women only). The findings for the Anglo-Saxon welfare state regime are perhaps unsurprising given that wage replacement rates for the unemployed are the lowest in these welfare states, and that benefits are means-tested and subject to strict entitlement rules. The unemployed in the Anglo-Saxon welfare states are therefore at a great financial disadvantage in comparison to those in employment and this may well explain the magnitude of inequality as financial strain has been found to be an important factor in the relationship between unemployment and ill health.7 41 42 Furthermore, means-tested benefits are associated with stigma18 and so the non-financial problems of unemployment may be greater in the Anglo-Saxon welfare states. Our findings are in keeping with broader based studies of welfare state regimes and health indicators which have found that overall population health tends to be worse in the welfare states of the Anglo-Saxon regime.35 36 38 39 43 44

It is harder to explain the findings for the Bismarckian (men only), and Scandinavian (women only) regimes and certainly these are in contradiction to the expectations outlined in our second research hypothesis. Interestingly, unlike the Anglo-Saxon welfare states they apply only to one or other gender. It is possible that the status differentiating Bismarckian welfare states may tend to exasperate the relationship between unemployment and poor health by restricting access to the higher level social insurance benefits. The length of entitlement to social insurance is also comparatively low in the Bismarckian welfare states (table 1). That relative inequalities are greater between men than women, may also be in part due to stigma as the familial approach of the Bismarckian welfare states emphasises the male breadwinner role.28 45 46 In terms of the Scandinavian welfare state regime, the relatively large inequalities between employed and unemployed women may well reflect the fact that women are less likely to meet the qualification criteria for social insurance payments (for example due to higher rates of part-time working)47 and are therefore dependent on social assistance benefits which have a lower overall replacement rate.48

It is of interest that the smallest relative inequalities between employed and unemployed were found in the Southern and Eastern welfare states. For example, the health differences between unemployed and employed people in the East European welfare regime were never larger than OR = 2.15 throughout the study (table 3). This is somewhat counter to the wider inequalities in health literature, which suggests that relative inequalities in health by socioeconomic status should be larger in these countries.49 50 This finding is therefore very surprising and clearly requires further analysis (perhaps looking at individual countries in these regimes in more depth), not least as the replacement rates and eligibility criteria for the Southern and Eastern welfare state regimes are not particularly generous, holding a fairly moderate position in relation to other regimes (table 1). One possible explanation for the finding is that the more traditional family model in these countries means that additional material, and non-material, support is provided by the family to unemployed members thus buffering the impact of unemployment on health.

Our main results and the sensitivity analyses also suggest that there is an important gender dimension to the relationship between unemployment and poorer self-reported health. Health inequalities between the unemployed and employed were larger among women, most strikingly in the Anglo-Saxon and Scandinavian welfare state regimes. First, this is in contrast to most single country, longitudinal studies, in which the relationship between unemployment and poor health has generally found to be more noticeable among men. Caution should therefore be applied to our findings until they are replicated. However, from a social protection perspective it is less surprising that women experience a more adverse impact on health of unemployment. Women are often not entitled to the higher value social insurance benefits, due to a less coherent employment history, for example, part-time work, periods out of work due to caring etc, and therefore have to rely on lower level social assistance, which provides much lower replacement rates, even in the more generous Scandinavian welfare states. It is also possible that the selection effect is stronger for women than men, that is that unhealthy women are more likely to become unemployed than unhealthy men. Future research clearly needs to explore further the relationship between women, unemployment and health, and the role that the welfare state can play in supporting the needs of an increasingly feminised European workforce.

Limitations

Although the ESS presents an outstanding opportunity to investigate cross-national patterns of health inequality, as the survey asks the same questions at the same time in all countries, we acknowledge that there are many issues which may affect the comparability of multi-country studies, such as variations in response rate (table 2), modes of data collection, translations, cultural interpretation and conduct.26 29 Our study is further limited by utilising only self-reported health measures which may vary by country, socioeconomic or employment status and/or culture. For example, an unemployed immigrant living in Spain may use different criteria to define his or her health than an unemployed Finn living in Finland. However, studies have found a strong relationship, which does not vary by socioeconomic status,51 between self-reported health and mortality.52 Similarly, the measure of unemployment (unemployed in the last 7 days) may obscure important between country differences in the composition of the unemployed population (online appendix). Further, the complex nature of the relationship between unemployment and health means that, despite conducting a number of sensitivity analyses, we have not covered all the possible factors influencing between country differences. Another possible limitation is our choice of welfare state regime typology. As noted in the introduction, there is a multitude of competing welfare state regime typologies and no categorisation has yet been generally accepted as the standard typology (although Ferrera’s is one of the most accurate in terms of how social benefits are granted and organised). We also carried out a number of sensitivity analyses. However, it must be acknowledged that if the typologies of other authors were used it may have resulted in different results. Finally, as the ESS data is cross-sectional, we cannot rule out selection effects.