Taking time trajectories into account is fundamental for understanding levels of SWB. There is extensive evidence indicating that age affects SWB. In an analysis of British panel data from 14 data collection waves starting in 1991 on respondents aged 16–64 years, Clark (2007) found a U-shaped relationship between age and SWB (i.e., overall mental well-being, overall life satisfaction). A U-shaped age-SWB relationship has also been confirmed for other countries: Blanchflower and Oswald (2008) found evidence of the existence of such a pattern for the US and Western Europe between 1976 and 2006 based on data from the US General Social Survey (1972–2006), the Eurobarometer survey (1976–2002) and the World Values Survey (1981–2004) covering respondents aged 20 and older in 80 countries. A U-shaped pattern was also observed for Eastern European, Latin American and Asian nations, but not for some developing countries. More recently, Blanchflower (2021) re-examined this U-shaped happiness-age curve globally based on recent data for 145 nations, including 109 developing countries. After controlling for socio-demographic variables and analysing samples covering respondents under age 70, a U-shaped pattern was confirmed globally.

The existence of such a relationship may seem counterintuitive, as the elderly are more likely than younger people to experience negative life events, such as reduced income, deteriorating health or the death of a partner. This pattern has been explained through socioemotional selectivity theory (Carstensen et al. 2003), which argues that compared to younger people, the elderly pay more attention to emotional goals and experiences, and are more likely to remember emotional and positive information.

It appears, however, that this U-shaped curve is not universal. Frijters and Beatton (2012) rejected the claim that there is a U-shaped relationship between age and life satisfaction (i.e., individual happiness) based on an analysis of panel survey data for respondents aged 18 and older in Germany, the United Kingdom and Australia. When using conventional regressions of fixed-effects models, the U-shaped relation disappears due to reverse causality. They argued that ‘happiness-increasing variables, like getting a job, a high income, and getting married, appear to happen mostly to middle-aged individuals who were already happy’ (ibid., 540). There is other evidence that the U-shaped curve does not exist across all countries. Based on the results of a Gallup poll among people aged 15 in 160 countries, Steptoe et al. (2015) examined the relationship between age and three aspects of SWB (i.e., evaluative, hedonic and eudemonic). They found a U-shaped relationship in high-income and English-speaking countries, but a negative association in other regions, including former Soviet states and Eastern Europe.

Levels of SWB also appear to be determined in part by people’s birth cohort. Empirical evidence from the US generally shows that levels of SWB are lower among baby boomers than among members of earlier birth cohorts, keeping other factors constant; even though selection effects exist (Yang 2008). Based on data from the English Longitudinal Study of Ageing (ELSA), Jivraj et al. (2014) found that, after controlling for age, older cohorts (aged 70+ in 2002–2011; i.e., pre-baby boomers) had the same or higher levels of SWB (i.e., quality of life, depressive symptomatology, life satisfaction) than younger cohorts (aged 50–59 in 2002–2011; i.e., late baby boomers).

There are many reasons why baby boomers might be distinct from members of other cohorts. The share of elderly in the population is currently growing primarily because the baby boom cohorts (i.e., cohorts born in 1945–1964) are so large. Baby boomers have a socio-economic and cultural identity related to their educational and occupational careers that is distinct from the identities of the previous and the following generations. Compared to the preceding generation, baby boomers have tended to report lower levels of well-being, which can likely be attributed to their early and formative life conditions and experiences, including being faced with rising economic inequality in the post-war era resulting from a larger cohort size that created greater competition for education and jobs (Easterlin et al. 1993; Fukuda 2013). Moreover, at least in Western countries, baby boomers can be seen as the forerunners in the dramatic changes in family life that characterised the Second Demographic Transition (SDT) regime (Lesthaeghe 2014). Based on data from three US censuses conducted between 1980 and 2000, Lin and Brown (2012) found that one out of every three baby boomers was unmarried, largely due to divorce or having never married, with a very small proportion having been widowed. Baby boomers continue to experience unstable marriages as rates of divorce increase in later life, as part of the so-called ‘grey divorce’ trend (Brown and Wright 2017; Stepler 2017).

While baby boomers as a group differ from other cohorts in terms of their socio-economic, cultural and demographic characteristics, considerable heterogeneity can also be found within the group of the baby boomers (i.e., early boomers were born in 1945–1954 and late boomers were born in 1955–1964). Based on the ‘Midlife in the U.S.’ surveys of 1995–1996 and 2004–2006, Barrett and Toothman (2014) examined the subjective experiences of ageing and their health consequences (i.e., individual views of the life course, such as the structure, timing and advancement of careers) during the midlife period. The early boomers reported having significantly more youthful identities and later boundaries for middle age than the late boomers. The late baby boomers also had lower socio-economic positions than the early baby boomers, keeping age constant. These findings imply that the early and the late boomers have experienced different economic and social conditions that have produced different life course trajectories (ibid.).

While age and cohort effects are rather well-established, evidence on the impact of historical events or trends on levels of SWB has been rather mixed. While economic prosperity has increased over time, happiness has not risen in a linear manner (Easterlin 2010). This non-linear relationship between income and happiness over time has also been documented by Yang (2008). An analysis of General Social Survey (GSS) data showed that happiness levels decreased in the 1970s–1980s, and gradually rebounded thereafter. However, Ortiz-Ospina and Roser (2017) provided another perspective. Based on an analysis of data from the Gallup World Poll and the World Values Survey, they found evidence that happiness has increased globally over the past three decades (1984–2014). In the same study, and based on data from the Eurobarometer survey on life satisfaction, the authors also reported observing fluctuations across historical periods (1973–2016). However, they generally found decade-long positive trends for most European countries except Greece, where there was a substantial decrease in life satisfaction during the financial crisis (2007–2012).

Levels of SWB are highly dependent on people’s socio-economic and demographic characteristics. Multi-dimensional concepts and theoretical frameworks can provide valuable insights into this relationship by highlighting a number of determinants of SWB, such as income, health, social contacts, employment status, personality and culture (religion) (OECD 2013, 33, Figure 1.1); as well as marital status, education and living arrangements (Thompson and Marks 2008). These social differentials of SWB levels are also strongly associated with the individual’s place of origin (OECD 2017).

Clear evidence exists on the effects of socio-economic position on levels of SWB. According to a systematic review of literature from the 1995–2013 period, among the population aged 60+, having a poor socio-economic position (as reflected in, for example, levels of education, income, wealth, financial assets, social class and neighbourhood characteristics) is associated strongly with poor SWB (i.e., subjective health, life satisfaction, quality of life, well-being), whereas this association weakens when mediated by social support and a positive self-perception of health (Read et al. 2016). A significant positive relationship has been found between SWB (i.e., happiness, life satisfaction, morale) and levels of education in particular (Witter et al. 1984). However, while there is strong evidence that higher education leads to greater happiness, either directly or indirectly (e.g., via income), this relationship is complex due to the difficulties that can arise in the operationalisation of education (Michalos 2008).

Empirical research has identified gaps in levels of SWB between native and non-native people, even after controlling for country and individual characteristics (Hadjar and Backes 2013). Migrant populations account for around 10% of Europe’s total population, although the proportions and socio-economic backgrounds of these migrant groups differ substantially across countries. Based on an analysis of data from the Survey of Health, Aging and Retirement in Europe (SHARE), Sand and Gruber (2018) presented empirical evidence indicating that non-European migrants and migrants from Southern and Eastern Europe have substantially lower levels of SWB than natives. They also found, however, that migrants from Northern and Central European countries have levels of SWB comparable to those of non-migrants. In general, migrants have been shown to have lower socio-economic status and SWB levels than the native population. However more empirical evidence of this general finding is required because of the significant methodological issues in this research related to non-response and measurement errors (OECD 2017).

An earlier study has provided evidence of gaps in levels of SWB by gender, albeit with large differences by country (Inglehart 2002). The analysis of data from the World Values Survey in 65 countries between 1980 and 1999 using a life satisfaction measure found that women had higher levels of SWB than men in the more developed countries. However, it appears that gender can have mixed effects. On the one hand, based on a meta-analysis of developmental and gerontological literature obtained through electronic databases (i.e., PSYCINFO, MEDLINE, PSYINDEX), Pinquart and Sörensen (2000) found that associations between socio-economic status (e.g., education, income and social class) and SWB levels (i.e., life satisfaction, happiness, self-esteem) were generally stronger for men than for women. The authors attributed this finding to men having a traditional occupational path focused on personal career development. Another view has been provided by Witter and others (1984). Based on a meta-analysis of US literature prior to 1980, they found a significant positive relationship between education and SWB levels that was stronger for women than for men. The authors noted, however, that the strength of the relationship was reduced when occupational status was controlled for in the model.

What is less clear is how these differences in levels of SWB by socio-economic characteristics described above vary over the life course. On the one hand, cumulative advantage and disadvantage theory argues that socio-economic gradients in SWB levels can vary over the life course (Dannefer 2003). This implies that relatively small differences in education or income during adulthood can widen during the life course. It has, for example, been shown that there is a strong negative relationship between financial distress and well-being (i.e., control autonomy, self-realisation, pleasure) that widens in old age (Niedzwiedz et al. 2015). Interestingly, this study found that this association is weaker for Northern countries.

On the other hand, a number of studies have shown that SWB levels tend to converge over the life course. For example, a study for the US found that having advantageous socio-economic characteristics (e.g., being male, white and highly educated) does not continue to confer advantages at older ages (Yang 2008). Sand and Gruber (2018) also found evidence that the gap in levels of SWB between natives and migrants diminishes at older ages. Another study by Inglehart (2002) found that in rich Western societies – but not in many non-Western (Asian, Latin American, African) societies – the relationship between gender and well-being reverses, from women having an advantage at younger ages to women having a disadvantage at older ages. Gender differences in SWB (i.e., life satisfaction, happiness) have been attributed to women generally reporting higher frequencies of depression because of their disadvantaged socio-economic position (Van de Velde et al. 2010), and experiencing and expressing emotion more than men (Plant et al. 2004).

Research has shown that demographic factors play an important role in SWB levels (and the reverse). Lower levels of SWB, measured in a multi-dimensional manner (i.e., life satisfaction, emotional well-being, vitality), were found for divorced than for married people in Northern Belgium (Jenkinson et al. 2020). The relationship between fertility and SWB has been studied extensively, mostly in North America and Europe. An analysis of data from the British Household Panel Survey found that having children has large positive effects on life satisfaction (Angeles 2010), and that these effects become stronger as the number of children increases. Recent cross-national research has shown that high levels of life satisfaction instigate fertility, even in low-fertility countries (Mencarini et al. 2018), although this relationship can vary depending on individual factors and across institutional contexts. Based on data from a large number of OECD countries, Glass and others (2016) found large variations in the happiness of parents, depending on the generosity of family policies. There is evidence that the association between parenthood and SWB can be weakened or strengthened depending on cultural practices and institutional arrangements, in particular when gender inequality is present (Aassve et al. 2015). Thus, the generally positive relationship between the number of children and happiness can be influenced by contextual factors that lead to happiness levels being higher at younger ages (i.e., around the birth of first child), but decreasing thereafter (Myrskylä and Margolis 2014). Moreover, it has been shown that life-cycle conditions can affect the relationship between happiness and fertility due to the financial and emotional costs of childrearing, but that the high costs for parents when children are young can be compensated when the children are older and provide their parents with social protection (Margolis and Myrskylä 2011).

In an analysis of the World Database of Happiness that controlled for individual factors, including income, unemployment, inequality, social capital and life satisfaction, Spruk and Kešeljević (2016) found that countries with better economic institutions and higher levels of economic freedom (e.g., property rights, open markets, limited governance) have higher individual happiness levels. Based on data from the World Values Surveys in OECD countries, Welsch and Kühling (2016) found evidence of a positive relationship between macro-economic performance and life satisfaction levels, with trade openness and institutional quality playing a crucial role. Rather than mediating effects, Pacek and Radcliff (2008) found that welfare programmes have direct effects on individual happiness and life satisfaction. These findings imply that a country’s social welfare characteristics also have an effect on individual well-being levels.

In order to examine SWB empirically, this paper focuses from a life course perspective (Giele and Elder 1998;Shanahan et al. 2016) on one aspect of SWB, happiness. Specifically, we examine the time, socio-economic, demographic and behavioural determinants of happiness. The framework comprises four life course components ‘individual agency’, ‘linked lives’, ‘historical period and time’ and ‘timing’ – that correspond broadly to age, period and cohort. Our study focuses in particular on age and cohort. Cohort is an important concept for understanding social change because each cohort has distinct experiences and characteristics, driven by the changing contexts of formal education, socialisation and historical events (Ryder 1965). Compared to other generations, the baby boomers experienced higher levels of competition in schools and jobs, which resulted in socio-economic and demographic differentials in education, income, family and health status.

Over the life course, gaps in levels of SWB by socio-economic characteristics can either increase or decrease. On the one hand, cumulative advantage and disadvantage theory argues that relatively small differences that emerge during adulthood can widen during the remainder of the life course. These relative advantages and/or disadvantages can result in widening differences in later life (Dannefer 2003). For instance, even though in the post-war period overall educational levels increased and the welfare state expanded, having low educational attainment early in life continued to affect people’s subsequent occupational trajectories, resulting in economic hardship even after retirement (Oris et al. 2017) that could influence their SWB levels. On the other hand, there may be age effects of happiness, which suggests that socially advantaged groups are not necessarily better off than the advantaged groups later in life. This implies that crossovers or smaller gaps in SWB levels by socio-economic gradients can occur during the life course, resulting in converging SWB levels.

Two research questions, and corresponding hypotheses, are formulated as follows.

1. What are the effects of different dimensions of time (age, period, cohort) on people’s levels of happiness? Following previous empirical research on the impact of age, we expect (H1a) to observe a general U-shaped pattern. (H1b) We also assume that the cohort effects (i.e., effects resulting from changes that occur and influence people in their earlier years) are more substantial than the period effects (i.e., effects resulting from changes that influence all age groups equally). This means that we anticipate finding some negative period effects for the years of the economic-financial crisis (2008 and later), and that we expect to observe more substantial negative effects for baby boomers, and particularly for late baby boomers (born in 1955–1964) relative to pre-baby boomers (born before 1945). (H1c) These time-specific trends are, however, assumed to differ cross-nationally: we expect to find U-shaped age effects in Northern and Western European countries, but smaller effects in Eastern European countries; and lower levels of happiness in Eastern European countries than in Northern and Western European countries.

2. Do we observe differences in happiness levels depending on people’s socio-economic, demographic and behavioural characteristics, controlled for time effects? And, if so, do we observe widening or decreasing differences in people’s happiness levels over time? (H2a) We assume that people with weaker socio-economic profiles have lower levels of happiness in general, but that these patterns are diverse during the life course, and to some extent across cohorts: i.e., we expect to find that some socio-economic profiles (e.g., low education) exhibit growing disparities; and that others (e.g., female, non-native) show crossover trends whereby socio-economic advantages turn into disadvantages, and vice versa. (H2b) We also expect to observe lower happiness levels among people with lower socio-economic and non-traditional family status; albeit with substantial cross-national differences, as discussed above. We furthermore expect socio-economic gaps due to gender and educational level to decrease moderately in Northern and Western European countries because of the effects of general social welfare programmes, but to remain large in Eastern and Southern European countries, where social welfare programmes are rather limited. While non-native status is largely related to the socio-economic characteristics and country origin of the migrants in each country, decreasing native-non-native gaps are generally anticipated.

We operationalised the hedonic aspect of SWB through measures of happiness, and created an ordinal response variable (1=veryhappy, 2=happy, 3=unhappy) based on an originally 11-point scale variable (i.e., how happy are you?; 0: extremely unhappy to 10: extremely happy). The ordinal response variable ranges from 1=veryhappy (original points scale of 8, 9, 10), over 2=happy (original points scale of 5, 6, 7) to 3=nottoohappy(original points scale of less than 4). We chose to use an ordinal response variable because we observed that there were few responses for some of the response choices for the continuous variable. This was considered a major disadvantage in the application of cross-national analysis.

Key socio-economic individual items – age, birth year, gender, level of educational attainment and native/non-native status – were included in the fixed part of the initial model. These variables were mostly transformed into dummies, except for age and birth year, which were grand mean centred and squared. In addition, the interaction effects between these socio-economic variables and time (i.e., age and birth year) were included in the extended models (models 2, 3). The interaction effects between age and birth year that test for life course trends were incorporated in order to examine varying cohort effects (Bell 2014; Bell and Jones 2015). This item made it possible to assess whether and how life course trajectories change across cohorts. In addition to the above, the respondent’s assessment of income, employment status, partnership status, child status, health condition and social interaction level were all included in the full model.

Period (survey rounds) and cohort groups (10-year groups) were included in the random part of the hierarchical model. In each survey round, data collection took place in the same year in all countries (European Social Survey 2016).

Profiles of high happiness levels are shown first, distinguished by periods and birth cohorts in each country. Each survey round (eight rounds corresponding to 2002–2016) and each cohort group (comprising 10 years, corresponding to birth years between 1905 and 1985) are shown.

Analytical models

Hierarchical age-period-cohort (HAPC) models applying Bell and Jones’s modifications are used (Snijders and Bosker 1999; Yang and Land 2013; Bell 2014; Bell and Jones 2013, 2015) to examine the dynamic changes in SWB levels through happiness measures by time trends. Cohort effects, which result from events and experiences that occur and influence people in the earlier years of their life course, are considered more substantial than period effects, which result from general historical events that influence all age groups equally. This assumption is justified because it enables us to take into account the upbringing and experiences of particular cohorts by examining pre- and post-baby boomers’ levels of happiness.

In order to address the linearly dependent issue when modelling age, period and cohort, our model constrains the period linear trends to zero by including the birth year as a linear fixed effect (Bell 2014; Bell and Jones 2015). Basic level-1 consists of (Yang and Land 2013; Bell and Jones 2015, 203):

The random intercept model is:

The dependent variable is the ordinal response outcome of happiness of the ith respondent in the j1 th period (survey rounds) (j1=1–8) and the j2 th cohort (j2=1–9) (ibid. 203). The addition of a single birth year (cohort) in the fixed level is a modification by Bell (2014) and Bell and Jones (2015) of the model proposed by Yang and Land (2013). Period and cohort residuals remain in the model to illustrate contextual effects. Based on the combined 2-level models, the likelihood of ‘being very happy’ is predicted. The model includes gender (female/not), migrant status (native origin/not) and levels (low/high) of educational attainment; the interaction effects between those variables and age/birth year; and socio-economic, demographic and behavioural items (e.g., partnership status, child status, employment status, health status, income assessment, social interaction levels).

Happiness levels are estimated with three models for each country using the SAS programme and implementing proc glimmix. All models consist of individual fixed and random parts. The basic model consists of time and individual fixed core socio-economic determinants (model 1), to which interaction effects are added (model 2), followed by additional socio-economic, demographic and behavioural determinants (model 3). Time trends are examined by studying the statistical significance of individual estimated coefficients for birth cohorts and periods that identify patterns in graphs, as proposed by Frenk and others (2013). Models are estimated with the hierarchical ordinal logit models of happiness using residual pseudo-likelihood estimation. Fit statistics (i.e., -2 Res Long Pseudo-likelihood) are documented and examined across models. All model results incorporate design weights, which are publicly available in datasets.