This study relates to three strands of research: studies that associate social policy to population health and health inequalities, studies on the impact of gender equality on population health, and studies that investigate trends and country differences in the male/female mortality gap.
Comparative policy studies with a health focus
As welfare policies have clear effects on income distribution and poverty outcomes [1–3], it would seem likely that they also have effects on the health of the population, as well as health disparities. Some researchers have investigated differences between country clusters, either defined by their political tradition [4, 5] or their welfare regime type [6–8]a. Studies on either cluster or country differences tend to find few significant differences favouring social democratic welfare regimes in terms of health inequalities defined according to education, social class or income [9–13] although there are exceptions [5, 8]. Instead, often the corporatist countries have the smallest (relative) differences. Some studies have however found cluster differences in absolute levels of health favouring the Nordic countries [6, 7, 14] and there are indications that the Nordic welfare state may buffer against detrimental effects of economic recession . Instead of clustering countries, Navarro has included accumulated years of political incumbency as a predictor, and found that reduced infant mortality is clearly related to years of government of redistributive parties . Also Muntaner  found consistent associations between indicators of working class strength and measures of birth and infant survival.
A different analytic strategy is to investigate specific policies and their impact on population health. Factors identified as strongly associated with health improvements are total social spending, universal access to social insurance , parental leave and the generosity of basic security pensions . One study found that social welfare spending led to a reduction of cause-specific mortality, whereas healthcare spending did not  but other researchers have found medical coverage and primary care to be important for health outcomes [16, 20, 21]. The idea behind this strategy, as argued by Lundberg , is that it is the existence and level of certain welfare state institutions that explain cluster differences and that crude categorisations fail to identify the specific welfare state characteristics that matter for health. He also criticises studies that investigate political party incumbency rather than what parties do. This critique receives some support from Chung & Muntaner , who found that the percentage of left vote lost its explanatory power when welfare state variables were entered in the model. Others have argued that the effect of politics goes beyond policy e.g. through social processes such as grass root movements and NGOs, and that studying policies one by one conceals possible inter-sectoral effects . A recent review of 73 comparative studies has shown that the factors most consistently related to good health was the strength of democracy and egalitarian political traditions, whereas studies using a welfare regime framework more often found mixed results .
Research on the health effects of gender policy, or the effect of policy on gender differences in health, is rare. A group of researchers, however, have examined the association between US state-level policies and women’s health. They found that access to health insurance and services, gun control, and policies on violence against women were related to female mental health and cause-specific mortality , and that access to health care, policies on violence against women and antidiscrimination policies were associated with blood pressure, smoking and obesity . Bambra et al.  have examined the relationship between gender and self-assessed health in 13 countries categorised according to an expanded welfare state framework . The study showed that while women in the social democratic and Southern welfare states were more likely to report worse health than men, there were no gender differences in the corporatist countries. Possible causes of the poor performance of the social democratic countries were, according to the authors, women’s dual roles in countries with high female labour force participation, combined with a sex segregated labour market offering worse jobs for women.
Gender equality as a health determinant
Aggregate and multilevel studies from the USA have examined gender equality on state level as a determinant of health. These have used indices of women’s political participation, economic autonomy, employment and earnings, and reproductive rights as indicators of gender equality. Investigated outcomes were mortality and reported days of activity limitations , women’s self-rated health , depressive symptoms , and child well-being . These studies have found that states that perform poorly on the gender equality indicators also have worse health outcomes, for men, women and children. A previous Swedish study, however, found negative associations between gender equality (measured as political participation, division of labour and economic resources) at municipal level and health for both men and women, while results regarding gender inequalities in health were inconclusive . These discordant results (cp to previous studies) were primarily attributed to an unbalanced mix of high gender equality regarding political participation and income, but with large remaining inequalities in the division of labour, both paid (i.e. sex-segregation) and unpaid. An international study of adolescent’s health showed that health complaints in both boys and girls were lower in countries with a high Gender Empowerment Measure (GEM)b, and that the gender gap in complaints was larger in countries with a low Gender-related Development Indexc. Another study focusing on male mortality used a sample of 51 countries from four continents and found a strong association between female homicide rates, seen as an extreme expression of patriarchy, and mortality . Thus, gender equality overall is positively associated with health and may also contribute to smaller gender gaps in health. However, as the studies by Bambra et al.  and Backhans et al.  have shown, the relationship may sometimes be reversed.
The male/female mortality gap
In countries where very long time-series data is available it has been shown that from the early 17th to the early 20th century male and female mortality differed only slightly, with absolute differences varying from 0 to 2 years of life expectancy at birth [36, 37]. In some age groups, there was a male advantage, due to a gender unequal resource allocation and high maternal mortality . During the 20th century, life expectancy for both men and women has been steadily rising . From 1950 the male/female mortality gap has increased and then decreased, with the peak year varying from the early 1970s to the 1990s and with some countries still seeing no decline . The largest decline is found among the middle-aged (55–75 years) and the causes of death that have contributed most to the decline in the mortality gap are heart disease, accidents and violence (excluding suicide), lung cancer and breast cancer .
There is a scientific controversy regarding the causes of the changing gender gap in mortality. Some scholars argue that changes are primarily driven by the stage of diffusion of cigarette smoking , unrelated to changes in women’s roles and relative status [43–46]. Others point out the high association between the GEM and the male/female smoking ratio , and it has been suggested that female emancipation is the underlying factor behind widespread take-up of smoking . Smoking has been found to be more common among highly educated women 60 years and older, but with a reverse pattern in women 25–39 years , reflecting the take-up of and abandonment of smoking through hierarchical diffusion, where the lifestyles of dominant groups are gradually adopted by the whole population, while the former continue to change and refine their consumption style [50, 51] One factor halting diffusion could be a high degree of gender inequality, making it less likely that behaviours are seen as either affordable or appropriate . Groups remote in social space are unlikely to influence each other’s habits directly . Therefore, female emancipation can be seen as a prerequisite for the adoption of (formerly) masculine behavioural patterns.
Some researchers, especially those with a biological/evolutionary outlook, have chosen to focus on risk-taking among (young) males as a primary explanation of gender differences in mortality [37, 54]. Data regarding the mortality gap in external causes shows that since ca 1940 there has been a faster mortality decline among women than men, followed by greater improvements among young men [37, 41]. Waldron  examined trends in gender differences in accident mortality in five large OECD countries 1950–1998 and concluded that a combination of convergence of gender roles and the differential impact (due to existing gender differences) of other societal trends (e.g. regarding drug use and improvements in medical care or public health measures) could account for most trends.
To summarise, few studies have examined the impact of policy on gender differences in health, or of gender policy on the absolute levels of health or disease. Bambra et al.  utilised a standard (not gender-focused) welfare policy framework, and this may arguably not be as relevant when gender gaps rather than social inequalities are studied. As far as we are aware, no previous study has examined the association between gender policy and the gender mortality gap, thus linking these two research strands together.
Aims and hypotheses
The aim of this study is to investigate the link between gender policy and the gender gap in external cause and circulatory disease mortality. These outcomes were chosen as they both contribute strongly to the overall gender gap in mortality and to its decline, while their association with smoking is different.
Our main hypothesis is that earner-carer countries  should have smaller and/or decreasing gender gaps in mortality, and that this difference is primarily due to the particular policies that distinguish them - policies which are employment-supporting for women/mothers, lessen the caring burden of families, support fathering, and decrease the effect of previous employment on economic conditions in old age. The association should to a large part be mediated through achieved gender equality, e.g. through convergence of women’s and men’s status, gender roles and health behaviours .
In the case of external cause mortality, women’s role expansion is likely to lead to increased risk exposure, and probably also to more risk-prone behaviour. For men, gender equality-friendly countries may be characterised by masculinities that are less ‘extreme’ and thus less risk-prone, than in more traditional countries .
In the case of circulatory disease mortality, women’s role expansion could be both health enhancing; due to increased status and economic resources at different life stages, and health endangering; due to increased stress and a move towards masculine eating, drinking and smoking habits. For men, increased competition with women may be a stressor, while increased female employment also leads to economic prosperity for society at large as well as for the individual family unit, alleviating the burden of the male breadwinner. This means that the net effect may go in either direction.
An alternative hypothesis is that cluster differences are due to other societal factors such as economic development or income inequality, or health behaviours unrelated to achieved gender equality.
Specific policy indicators were chosen not for their purported health effects, but to measure aspects of Sainsbury’s concept gender policy regimes, with gender policy being defined as policy “associated with a certain gender ideology, that describe actual or preferred relations between women and men, principles of entitlement, and policy construction” . Therefore it is difficult to make predictions regarding the health impact of specific policies. Also, single policies don’t appear in a vacuum but are part of a policy package, making it difficult to distinguish the influence of a specific policy from the influence of a gender policy regime in general. Nonetheless, we have included specific indicators in order to investigate whether cluster differences found are reflected also in estimates for policy indicators. This also links back to the ongoing discussion whether investigating country clusters or indicators is the best approach.