Discussion

Our total inequality method-derived results suggest that the Nordic welfare state countries tended to achieve both the highest life expectancy and the smallest inequality for men but not women. For men and women Nordic countries generally had low infant and working age mortality. In general, though older age mortality in Nordic countries is not the lowest and so reducing this (particularly for women) would increase life expectancy, and for ‘younger’ older age deaths, reduce inequality. In contrast, the most prominent comparative work using the socioeconomic group difference approach (using education as the main measure of socioeconomic status) to mortality by Mackenbach et al10 “found no evidence for systematically smaller inequalities in health in countries in northern Europe.” In that study, it was the Southern European countries that were found to have the smallest overall inequalities in mortality for men and women. We found the same result for women but not men. There are a number of differences between that study and ours including the age range (we covered the whole age range) and the time period (2006 compared to 1990s mortality). Most obviously we used a measure of mortality inequality, total inequality, that was not based on socioeconomic group comparisons.

Our research also found clear variation among the Nordic welfare state regime countries with Sweden, Iceland and Norway having higher life expectancy and smaller inequality than Denmark and Finland (men only). A recent study that explored both overall and within group variation in the age of death (ages 35 and above) also found that the Nordic countries for men (Sweden, Norway) and women (Sweden, Norway and Finland) had small overall variance in the age of death although it included fewer countries (10) and no Southern European countries.28 Another recent study using total inequality (but with a slightly restricted age distribution that excluded under 10 mortality, fewer countries and no gender split) also found that on average inequalities were smallest in the Nordic welfare state regime but with significant within welfare state variation.29

This highlights another crucial methodological problem in comparative epidemiological research—the use of ‘welfare state regimes’, a concept which assumes a homogenous approach to welfare provision within and between the countries of any particular regime type. Kasza has asserted that instead of an internal policy homogeneity or cohesion, welfare states and welfare regimes exhibit significant variation across different areas of social provision.30 Further, welfare state provision has changed over time (with decreasing decommodification especially in countries such as Sweden and Finland since the recession of the early 1990s) yet there is a lag between our typology (with its roots in the 1990s) and our mortality data (2006). For example, there is evidence from recent comparative analysis of a degree of convergence between Nordic countries and some other western European countries due to Nordic states becoming more like other European countries (slightly less equity focused as a result of welfare state retrenchment and marketisation since the 1990s).31 This has led to the development of other ways of analysing cross-national differences in social policy, most notably the ‘varieties of capitalism’ approach which acknowledges the commonality in economic organisation between the Nordic countries and the Bismarckian countries.32 Future comparative research into health inequalities would benefit from examining this approach.32 In this paper we used the regimes approach as we are building on existing work but we have also presented individual country results so that the limitations of the welfare regimes typology do not restrict our analyses. Others have advocated for the analysis of specific social policies as an alternative way of conducting comparative research on inequalities and the social determinants of health.33

The main criticism of our approach is that it is does not include a measure of socioeconomic inequality34 with a long running debate about whether the total inequality or between group approach is most appropriate measure of inequality. We, and others17 ,35 see these as complementary rather than rival methodological approaches with both being important to understanding and tackling inequality. In cross-country comparative work there are concerns about how comparable similarly classified socioeconomic groups actually are in different countries.36 There are also concerns that using different socioeconomic classifications gives differing results (eg, income-based studies of morbidity by welfare state have produced different country configurations than education based ones11 ,13). There is also the issue that categorisation into relatively large socioeconomic groups may well obscure health differences between people similarly grouped as the concept of the social gradient in health indicates an often continuous relationship between socioeconomic position and health.1 Relatedly life course research indicates that socioeconomic position across the life course may have an incremental impact on health that may not be fully captured by contemporaneous or single measures of stratification.2 Using a total inequality measure overcomes these difficulties and captures inequality related to all forms of socially determined stratification.36 A potential drawback is that it may capture non-socially determined inequality with recent estimates suggesting that between socioeconomic group differences capture only a minority of the overall total inequality.35 While it is likely that there is a large amount of essentially random variation in the age of death within populations, due to chance and individual genetic susceptibility, it has been argued that when comparing groups (in our case countries) this random variation may well be averaged out meaning that differences in the level of inequality between groups are likely driven by causal factors.37 What is a key but still open question is explaining why total inequality is different between countries (and for specific age groups) and how much of this difference is socially determined. It is prominently argued that differences in (premature) mortality between countries are mostly socially determined with those lower down the social scale within the country most vulnerable.1 For example, recent work found that within education group variance in the age of death was greatest for those in the lowest groups in all countries and that countries with the lowest level of overall variation in the age of death achieved this by having the least variation among those lowest educated.28

It has been argued that equity focused social policy alone may not be enough if other important policies for good health are not in place.10 This is evident in how our research highlights that Denmark had higher death rates and inequalities than we might expect and this is probably linked to high levels of smoking.38 National tobacco control policy in the mid-2000s was noticeably weaker in Denmark compared to other Nordic countries and this may have had a negative impact on the quit rates of smokers from across the social spectrum.39 National tobacco control policy is a political issue and so should be regarded as an important part of a social determinant-based public health policy.1 In line with Karim et al40 then, we would argue that future research on welfare state regimes and population health needs to incorporate wider public health indicators, not just social policy ones, resulting in a ‘public health regime’ approach that examines the legislative, social, political and economic structures that have an impact on both public health and public health interventions.

Of course simply describing these differences is not sufficient and further work is needed to fully understand the reasons for differences in inequality between countries. Health inequalities exist in all countries but they may do so for different reasons. For example, Southern European countries are at an earlier stage of the smoking epidemic than Northern European countries. So comparing Southern and Northern European countries without accounting for this difference is not unproblematic. As others have argued future research should explore in more detail the temporality of policy effects as there may well be significant delays in effects on mortality.2 Moreover, and including this study, most comparative research on mortality inequalities has been descriptive and observational and there is a need for more quasi experimental work to properly assess the impact of different policies independent of possible confounders. Another limitation is that we used 1 year's worth of data—2006. We chose to use the most up-to-date figures (based on the most recent year covering the most countries for which data was available at the start of our study) and future studies should explore longitudinal changes in inequality broken down by age and cause of death to gain a fuller understanding of how welfare state and other aspects of the public health regime of countries influence death rates. A final limitation is that we created a theoretical lowest mortality country in order to conduct the age decomposition. This may be an artificial comparison as such a combination of death rates is not observed in a single country but it does have the advantage of not being a comparison to an arbitrarily chosen comparator country.