In contexts characterised by changing population structures and growing imbalances across and within generations and social groups, ensuring a high quality of life and a healthy life course development for everyone becomes a paramount welfare objective. Despite the abundant literature on health and wellbeing, how these indicators vary with age remains the subject of theoretical and empirical debates. While there is consistent evidence that physical health tends to worsen with age, patterns of mental health and subjective wellbeing are less clear, with a majority of studies showing that both might actually be lower among mid-age groups (Lang et al. 2011; Blanchflower and Oswald 2016). However, other studies have reported that wellbeing increases with age (Walker 2005; Frijters and Beatton 2012), even though physical health worsens, or that levels of wellbeing do not change at all across age groups (De Neve et al. 2012). The few studies that have considered domain-specific satisfaction have tended to show that age trajectories of satisfaction diverge considerably across domains (McAdams et al. 2012; Easterlin 2006).

In parallel, a growing number of studies in the fields of economics, psychology, sociology and gerontology have focused on how wellbeing is (unequally) distributed across sociodemographic groups in contemporary societies, and have shown that inequalities are increasingly being observed in a number of countries (Mackenbach 2012; Townsend and Davidson 1982; Elo 2009). A rich literature on health inequalities using different measures of both self-reported and objective health has reported that individuals with higher incomes and educational levels tend to experience better health over their life course (e.g., Mackenbach et al.1997; Marmot 2005), albeit with important differences across countries (Kunst et al. 1995). While there are indications that gender differences in overall health might be smaller than was previously thought (Arber and Cooper 1999; Oksuzyan et al. 2019), women generally report having lower levels of physical and mental health than men (Crimmins and Saito 2001; Dahlin and Härkönen 2013; Lee et al. 2016). It also appears that subjective wellbeing is heterogeneously distributed across social groups. There is, for example, extensive evidence that higher education tends to be associated with greater happiness, as well as with better health. However, findings on gender differences in life satisfaction have been more contradictory, with some studies showing that women tend to be happier than men (Blanchflower and Oswald 2004; Easterlin 2001), and others reaching the opposite conclusion (Pinquart and Sörensen 2001).

Moreover, even less is known about the divergence of age trajectories of health and wellbeing across those social groups. According to the cumulative disadvantage theory, early inequalities may be exacerbated by the ageing process, causing wellbeing trajectories to diverge over the life course (Dannefer 2003). Well-educated individuals are less likely to face risky situations and are less vulnerable when they experience adverse life course events such as job loss, illness, or financial strain. Educational achievement equips individuals with positive attitudes that are useful for the maintenance of health across the life course. Furthermore, being employed favours social integration, and can increase an individual’s social support and opportunities (Loscocco and Spitze 1990; Cerci and Dumludag 2019). By contrast, the age-as-leveller hypothesis argues that the ageing process could smooth initial differences across groups if the resources that generated the inequality in the first place become less useful over the life course (Lynch 2003). In addition, the survival selection process means that the happier and healthier individuals in each group live longer, and thus participate in surveys longer (Ulloa et al. 2013; Kratz and Patzina 2020). This implies that the group of survivors at older ages are more homogenous. The empirical findings reflect this theoretical duality: some studies have found that gender and educational gaps widen with age (Pinquart and Sörensen 2001), while others have reported constant or shrinking inequalities (Yang 2008).

Finally, a number of studies have highlighted the importance of broadly interpreting the concept of wellbeing as being composed of a number of factors (Pollard and Lee 2003; Cronin de Chavez et al. 2005; Huppert 2013; Infurna and Luthar 2017). A satisfactory ageing process entails not just having high levels of subjective, psychological, relational, and financial wellbeing; but also remaining in good health. A number of investigations have compared multidimensional measures with standard satisfaction questions, and most found only small or moderate correlations (e.g., Huppert and So 2013; Ryff and Keyes 1995). Thus, measurements of wellbeing may not be reducible to a simple, unidimensional notion such as life satisfaction without losing a great deal of potentially valuable information (Keyes 2007). However, most of the work on wellbeing has been undertaken within single disciplines, and has tended to focus on one aspect of the concept, rather than on drawing together the physical, psychological and socioeconomic aspects of wellbeing (Cronin de Chavez et al. 2005).

In this study, we contribute to the literature on unequal age trends in health and wellbeing across gender and socioeconomic status by illustrating how these trends unfold in two different contexts using data from France and Switzerland. While both of these European countries are considered conservative welfare states (Esping-Andersen 1990), they differ in a number of aspects that shape inequalities in sociodemographic health and wellbeing. First, compared to the French welfare state, the Swiss welfare state shares many more features with the liberal welfare system, especially in terms of its privatised health care system. Second, gender norms and the related family and labour market institutions tend to be much more traditional and oriented towards the male breadwinner model (and towards greater financial insecurity among women) in Switzerland than in France. Interesting differences in the health and wellbeing profiles of the two countries can be observed, which provide us with important insights into the cultural and institutional differences in their determinants, and again highlight the importance of considering the multiple dimensions of quality of life (Huppert et al. 2009). Because the questions on health and wellbeing in the two surveys used here were formulated in different ways, we cannot make direct cross-country comparisons. However, exploring the unfolding of health and wellbeing inequalities across age groups in two different contexts may enable us to identify the circumstances that are associated with growing or shrinking inequalities within countries (Kunst et al. 1995).

We use data from the 2006 waves of the Santé et intineraire professionnel (SIP) for France and the Swiss Household Panel (SHP) for Switzerland to study individuals of working ages. The two surveys provide a variety of indicators of health and wellbeing, as well as demographic and socioeconomic variables. We focus on four dimensions of physical and mental health and two domain-specific wellbeing measures: relational and professional wellbeing. Our study is descriptive in nature. Our main aim is to explore the gender and educational inequalities in the age trends in two different contexts, while focusing on less investigated indicators of health and wellbeing. We contribute to – and hope to partially reconcile – the streams of literature on the age-wellbeing nexus, the unfolding of wellbeing inequalities over the life course, and the contextual determinants of those age trends and heterogeneities.

In our analysis, we use data from the Santé et Itinéraire Professionnel (SIP) and the Swiss Household Panel (SHP). The SIP survey was conducted among individuals aged 20 to 74 in 2006 living in ordinary households in mainland France. Two waves (2006 and 2010) are available. The Swiss Household Panel is a random sample of private households in Switzerland in which all members of the family aged 14+ are interviewed annually. The SHP has 20 waves (1999–2018), but to ensure comparability, we used the 2006 wave for both surveys. We decided to use the 2006 wave instead of the 2010 wave for two main reasons. First, compared to the data from the 2010 wave, the French and Swiss panel data from the 2006 wave either are not or are less affected by attrition, and the samples are larger.³ Second, while there is no reason to expect to observe significant differences between the two waves since they are only few years apart, cross-sectional findings based on the 2010 wave could be affected by the financial crisis of 2008–2009. The recession had heterogenous effects on health and wellbeing across contexts and social groups that are not the focus of this study (Burgard and Kalousova 2015). We considered individuals of working ages (aged 25–69) in order to further limit the impact of attrition in the Swiss data, and because one of the dimensions of wellbeing we are interested in is professional satisfaction (N=2136 for men and N=2753 for women for Switzerland and N=4367 for men and N=5300 for women for France).

We focused on four outcomes of physical and mental health: self-reported health; daily activity limitations; depressive symptoms and sleep disorders; and two relational wellbeing indicators: satisfaction with social relationships (reliability of the social network in the SIP) and satisfaction with professional life. These measures were chosen because they provide information about similar health and wellbeing domains across the two surveys. However, as we mentioned, some important differences in terms of how the specific questions were phrased and the responses were coded persist, which limit their direct cross-country comparability. The first difference between the two surveys concerns the time frame the questions refer to. Table 2 reports the specific questionnaire formulations in the two surveys. In the French SIP, most of the health questions refer to a specific time frame, albeit one that varies across items: six weeks for depressive symptoms, six months for daily activity limitations and 12 months for sleeping problems. The Swiss SHP, by contrast, only uses a specific time frame (of four weeks) for sleeping disorders, while all of the other questions are posed in more general terms.

The second difference between the two surveys relates to the operationalisation of the responses. Table 3 reports the summary statistics of the main variables in the two surveys. For health status, the responses vary in the two surveys on a scale from one (very well) to five (not well at all), and are recoded in the opposite direction for both countries so that higher values mean better health. For work satisfaction, the responses vary on a scale of 0-10. All of the other indicators in the SIP data are binary (0/1). However, while sleep disorders were originally measured on a scale of 1–4, they are also used as a binary variable indicating whether the individual has had sleeping problems for at least several days a week. By contrast, in the SHP, the variables for satisfaction with social relationships, feelings of depression, and daily activity limitations are measured on a scale of 0–10, while sleep disorders are measured on a scale of one (not at all) to three (very much).

These phrasing and measurement differences limit the comparability of the levels of the indicators both between the two countries (e.g.: asking whether French women of different ages suffer from more or fewer sleeping problems than their Swiss counterparts) and within the countries. However, our primary research interest lies not in performing cross-country comparisons, but rather in illustrating the age trends and providing within-country comparisons across gender and educational groups. Therefore, we do not expect the different formulations of the questions to invalidate our analyses.

We studied separately for the two countries the association between age and the multiple measures of health and wellbeing by gender and socioeconomic status. Ages from 25 to 69 were grouped into five-year categories to allow some degree of flexibility in the association between age and health and wellbeing, but without over-specifying the model with year-dummies. We focused on respondents aged 25–69 to ensure that the respondents’ education had been completed, and to exclude the oldest old, whose participation in the survey would have been even more selected based on health reasons than was the case for the other age groups. Education is a widely applied measure of socioeconomic position that reflects people’s material and non-material resources, and that precedes their labour market and income prospects. We recoded education into three categories: primary or lower secondary, upper secondary and tertiary education. However, for the sake of clarity, we present in the figures only the findings for the low and the high educated.

The statistical analyses conducted in this study are based on the Ordinary Least Squares (OLS) method. We use linear regression models for continuous dependent variables and linear probability models for the binary health indicators in the SIP. Logistic regressions were also performed and provided similar results. However, we decided to show the results of the linear probability models because they are easier to interpret. We use cross-sectional weights corresponding to the countries’ populations. We do not control for socioeconomic variables, such as marital or employment status, because we are interested in the total association between age and health and wellbeing, and part of this association is indirectly convened by the expedience of such life course events at given ages (Easterlin 2006). We report our findings here by graphically showing predicted health and wellbeing by age groups in the two countries, while the complete tables are included the appendix (Tables A.2–A.5).

Results from models with age-gender (Figures 1–2) and age-education (Figures 3–4) interactions (although many interaction terms do not reach statistical significance) show that in both France and in Switzerland and in all groups, people’s overall health status and physical health clearly deteriorate at older ages, while the trends in mental health vary more depending on gender and education. In France, the two measures of relational wellbeing display opposite age trends: i.e., the reliability of people’s social networks steadily drop with age, while people’s satisfaction with their career increases with tenure. In Switzerland, levels of relationship satisfaction are quite constant across age groups, while job satisfaction increases significantly at very old ages only.

Ensuring a satisfying quality of life throughout the life course means not only remaining in good physical and mental health and having sufficient financial means, but also maintaining high levels of overall and relational satisfaction. All of these dimensions contribute to the overall construct of wellbeing. There is increasing evidence that there are growing inequalities in wellbeing in contemporary societies – albeit with large contextual differences – which strongly suggests that researchers and policy-makers should focus on reducing demographic and socioeconomic gaps in quality of life.

The main contribution of our study was to show how unequally wellbeing is distributed across social groups, and how such inequalities develop over age in two different contexts. We were particularly interested in examining how different dimensions of physical and mental health and relational wellbeing heterogeneously develop with age. We exploited two rich household surveys, SIP and SHP, to examine the development of health and wellbeing across age by gender and education in France and Switzerland. Our assumption was that analysing these trends in different contexts could provide insights into the social and institutional differences in the determinants of health and wellbeing trajectories. Despite both being rich European societies with conservative welfare regimes, France and Switzerland differ in two crucial domains: namely, in their prevailing gendered family and labour market institutions and in the nature and accessibility of their health care systems (Cullati 2015).

Before examining our findings, it is important to consider that our study had a few limitations. First, the questions on health and wellbeing were phrased differently in the questionnaires of the two surveys. Nevertheless, we believe that these measures provided reliable information for examining age trends in health and wellbeing across gender and education within countries. Interestingly, some wording differences between the two surveys even allowed us to explore various aspects of the same indicator. For instance, questions about work satisfaction in the SIP asked specifically about career, while the SHP asked about overall job satisfaction. Our finding of a large gender gap in France suggests (among other plausible explanations outlined below) that it is the specific aspect of career progression that causes women to feel much less satisfied than men, even in a context like France, which ranks relatively high in terms of gender equality.

Another important difference in how the questions on health and wellbeing were phrased in the two surveys was in the time frames they referred to. While most of the questions in the French survey referred to a specific time frame, most of the questions in the Swiss survey were phrased in general terms. This difference might have had some implications for the observed inequalities. Overall, the differences across age groups might have been smaller and the estimates might have been more prone to measurement error in Switzerland, where the questions were more general. It is also important to keep in mind that all of our measures of health and wellbeing were self-reported. This means that differences in reporting behaviour by age, gender or education might have played a role in the observed differences. While recent studies found no systematic gender or educational differences in reporting either good or bad health, they found a declining concordance between self-reported and actual health with age among older respondents (Spitzer and Weber 2019; Oksuzyan et al. 2019), although much less so among younger adults (Miilunpalo et al. 1997; Pursey et al. 2014).

The second limitation of our study was its cross-sectional nature. This means that our findings reflect differences between individuals of different ages in 2006, and not the effect of the ageing process within the individual. Healthier and happier individuals tend to live longer and participate longer in surveys; therefore, our older respondents were selected in terms of higher wellbeing. This might have led us to underestimate the age differences, especially in the Swiss data (2006 is the first wave of the SIP), as well as the education gap, given that the health selection mechanism tended to be stronger among the lowest socioeconomic strata of the population. Relatedly, because of the cross-sectional character of our analysis and the well-known age-period-cohort problem, we could not distinguish between the age and the cohort effects on health and wellbeing (Bell and Jones 2014).

The final limitation of our analyses was the low number of observations – again, especially in the SHP, and especially in the group of primary educated respondents – which prevented us from providing statistically sound estimates for these groups, and from further exploring the gender and educational inequalities.

Despite these limitations, we can draw important conclusions from our analysis. We found crucial heterogeneities across indicators, and depending on the types of inequalities we looked at. Our results indicated that in both countries, overall and physical health deteriorated across age groups. We found gender inequalities in physical health only in Switzerland, but a rather large socioeconomic divide in both countries. The educational gap was shown to be largest in mid-life, which might be explained by the different degrees of physical effort required in the types of work the lower and the highly educated individuals did. The low skilled workers tended to have more physically demanding jobs, and this might have affected their health, even though no cumulative effects were observed (although, as we noted above, there was a selection process). The age trends in mental health were less clear, and varied more among the social groups. Large gender differences that widened with age were observed in both France and Switzerland. This finding suggests that there were consistent gender inequalities in both contexts, and that those inequalities tended to accumulate over time. In contrast, no large educational differences in mental health emerged.

The results further showed that the reliability of social networks decreased over age in France, while relationship satisfaction in general was more stable across age groups in Switzerland. Gender and socioeconomic status differences in relational wellbeing were observed in France, but not in Switzerland. The strong reference to the support received from significant others in the SIP questionnaire may explain the decline in relational wellbeing found for older age groups in France. Significant gender differences that increased with age were observed in France for measures of career satisfaction, but not for measures of general job satisfaction in Switzerland. This finding suggests that women tended to have less satisfying professional career opportunities than men, but may not have been less satisfied than men with other aspects of their work. Our failure to find a gender gap in job satisfaction in Switzerland may have also been driven by the selected group of Swiss women who were in full-time work (especially during their childrearing years), as they may have been more career-oriented. Additionally, the low levels of satisfaction observed among French women might be due to a mismatch between the women’s professional careers and their expectations, which might have been higher in a country where gender norms are more egalitarian.

To conclude, our results suggest that both gender and socioeconomic inequalities in health and wellbeing are present in France and Switzerland. In particular, we observed significant gender inequalities on a number of wellbeing dimensions in Switzerland, as well as pronounced socioeconomic inequalities in France. Our results on gender in the Swiss context are in line with the traditionalist character of Swiss institutions, which often cause women to be dependent on men. In contrast, the finding of pronounced education inequalities in health in France might be surprising in light of the inclusive public health French system. The crucial factor here might be the much lower prevalence of primary education in Switzerland (6%, Table A.1) than in France (26%, Table A.1). On the one hand, due to the low number of observations in the lower educated group, our results might be underestimating the true inequalities in Switzerland because of the large confidence intervals of the estimates. On the other hand, the low prevalence of low education in Switzerland, which is a very rich society, shows that socioeconomic inequalities (at least those based on education) are relatively low in this context. However, this observation also implies that the very few individuals who are disadvantaged in Switzerland are much more deprived than they would be in other contexts, which makes it all the more crucial that studies are conducted to identify the sources and the outcomes of inequalities.

Finally, our findings show that while gender inequalities, when present, tended to accumulate with age, the educational gaps did not seem to diverge over time, but were largest in mid-life. This observation suggests that gender inequalities tend to be based on structural disadvantages that grow and accumulate with life events and the ageing process. For instance, the difficulties women face in reconciling family and work and managing career interruptions might have cumulative negative effects that make them more vulnerable than men at older ages. In contrast, socioeconomic inequalities, at least in the two contexts and given the indicators analysed here, seem to be linked to characteristics that matter mostly in mid-life, while they are less relevant for reducing inequalities at older ages.