Generational variations in wellbeing: suicide rates, cohort characteristics, and national socio-political context over seven decades

Abstract

Over recent decades, the relative wellbeing of younger birth cohorts declined in many western countries, indicating growing generational inequality. Building on Durkheimian theory, this paper examines explanations for these changes, hypothesizing that differences in cohort wellbeing are related to variations in social integration associated with birth cohorts and national socio-political contexts. Age–period-specific suicide rates of men and women from 1950 to 2020 in 19 highly developed western nations, including 26 birth cohorts, born from 1875 to 2004, are examined using estimable function analysis and age–period–cohort characteristic (APCC) models. Cohort variations in wellbeing are significantly greater in English-speaking nations, which have traditionally provided less institutionalized support and social integration than continental European nations. Age-specific suicide rates are larger for cohorts with childhood demographic characteristics associated with less social integration (relative cohort size and family structure). Major historical events associated with social integration in formative years of late adolescence and young adulthood also influence cohort wellbeing, with higher age-specific rates for cohorts experiencing the Great Depression of the 1930s and health pandemics of the early 20th and 21st centuries and lower rates for those experiencing periods of war and national conflict. However, the magnitude of these associations is strongly influenced by socio-political context. Negative effects of cohort characteristics are muted and positive effects are enhanced in the continental nations. In addition, patterns of associations vary by age and gender. Results remain with strong controls for the pace of change, additional measures of national context, and sensitivity analyses.

Generational variations in wellbeing: suicide rates, cohort characteristics and national socio-political context over seven decades

Over the last few decades, in many western countries, the wellbeing of young people, relative to those who are older, has declined. Change has occurred in areas as wide-ranging as chronic disease and physical disabilities (Gimeno, et al. 2024); depression, anxiety, and general mental health (Beller 2022; Twenge 2011, 2015); economic wellbeing and stability (Chavel and Schroeder 2014, Freedman 2017; Green, 2017); criminal offending (O’Brien 2019); and deaths from homicide and suicide (O’Brien and Stockard 2006; Pampel 1996, Pampel and Williamson 2001  Stockard and O’Brien 2002a, b, 2006). These patterns indicate an increase in generational, or cohort, inequality, with more recent birth cohorts having lower levels of physical, mental, and social wellbeing (Chauvel and Schroeder 2014, Pampel 1998a, Preston 1984). The changes have clear, lifelong implications for individuals. But they also have serious implications for societies, as more recent cohorts may have greater needs, but diminished ability to contribute to the general good.

Echoing the classic 19th century writings of Durkheim (1897/2006), decades of research have documented how social ties and social integration increase physical, mental, and social wellbeing (Berkman, et al. 2000). Social integration and regulation can be conceptualized and measured on multiple levels of analysis, from individual experiences to cultural norms and patterns. This paper examines the influence of social integration and regulation associated with birth cohorts on cohort variations in wellbeing and how the impact of cohort-related characteristics is influenced by social integration associated with socio-political contexts. Suicide rates, long considered an important indicator of wellbeing (Durkheim, 1897/2006; Land, 1971; Land, Lamb and Mustillo 2001) are the focus of the analysis. Because nations throughout the world gather age and gender-specific data on suicide deaths in standard formats, it is possible to compare patterns across broad spans of time, age, and national context.

Background and related literature

Four specific research questions are examined.

National socio-political context and the magnitude of cohort effects

The concept “families of nations” describes differences between countries in structures and cultural traditions that promote social integration and regulation. The literature distinguishes nations that share geographic, linguistic, cultural, and/or historical characteristics as well as legal traditions and social policies (Buhr and Stoy, 2015; Castles, 1993; Esping-Andersen, 1990; Inglehart, 2006; Inglehart and Baker, 2000; Obinger and Wagschal, 2001; Therborn, 1993).

Among western industrialized countries, four families of nations are often distinguished. The English-speaking, liberal welfare states, which adhere to the legal tradition of common law, have historically provided the least institutional support for families and children. Continental European countries adhere to European civil law traditions but have had variable political approaches and cultural traditions. The Nordic family of nations, representing Scandinavian social democratic welfare regimes, have traditionally provided the most extensive legal support for children born outside of marriage as well as generous and strong institutionalized systems of support. Nations within the Romanist and Germanic families have traditionally provided less generous institutionalized support than Nordic nations, but more than English-speaking countries. Previous studies have documented the association of these varying levels of institutional support with cross-national variations in cohort-related inequality in economic wellbeing (Chavel and Schroder, 2014) and suicide (Pampel, 1993, 1996, 1998a, 1998b; Stockard and O’Brien, 2002a, 2006). This paper expands upon those analyses, focusing on cohort differences in suicide through the first two decades of the 21st century.

Age, period, and cohort are statistically dependent; knowing two of these factors determines the third. It is, however, possible to determine the extent to which cohort effects are distinct from age and period (O’Brien, 2000, 2014, 2020). Several scholars have concluded that changes in the age distribution of suicide in the latter part of the 20^th century reflect cohort effects, as distinct from those associated with age or period (e.g. Chauvel, et al., 2016; Phillips, 2014; Thibodeau, 2015). The first research question examines the extent to which trends in age-specific suicide rates represent cohort, rather than age or period effects and whether the magnitude of cohort effects varies across national contexts. Building on Durkheim’s insights and the literature regarding families of nations, it was expected that larger cohort effects, indicating greater cohort inequality, would appear in nations with lower levels of institutionalized integration and regulation.

Social integration and regulation in childhood

Previous explanations of cohort variations in suicide have focused on two characteristics associated with social integration and regulation in childhood: the size of a cohort relative to others and family structure. Theorists emphasize that these variables are structural in nature, reflecting opportunity structures in which cohorts are born and grow to maturity. They influence cohort-related integration and regulation in a variety of ways including financial strains resulting from more children within a cohort or fewer adults within a household. This results in adult resources spread more thinly, less attention and supervision for children, and a stronger influence of peers. In part, cohort effects reflect the aggregation of individual effects as more children grow up in larger or single parent families. But, all members of a birth cohort are affected by these characteristics, no matter what the size or composition of their family (Carlson 1979, Easterlin 1987, McCall and Land 1994, Moller 1968, Pampel 1996, Ryder 1974). A number of studies, with varying methodological approaches, have examined the association of these demographic characteristics with suicide rates. Some have used time series designs, others age–period–cohort characteristic (APCC) models. They have also employed varying age ranges and operationalized demographic characteristics in different ways.

Despite methodological variations, results have been consistent. Larger birth cohorts have higher suicide rates; and, in studies including both gender groups, associations are stronger for men than women (Messner, et al. 2006; Pampel 1996; Pampel and Williamson 2001; Stockard and O’Brien 2002a, b, 2006). Some studies have also noted stronger associations of cohort size and suicide rates at younger ages and even reversed patterns, with lower suicide rates appearing at the oldest ages (65 and older), hypothesized as resulting from the greater economic and political power that can accrue to larger cohorts (McCall and Land 1994; Pampel 1996).

Studies also consistently indicate the importance of family structure, with higher suicide rates in cohorts with a greater incidence of divorced or single parents. These results appear with aggregate data (Freeman 1998; Messner et al. 2006; Pampel 1996; Pampel and Williamson 2001; Stockard and O’Brien 2002a, b, 2006) and individual level data examining suicidality or completed suicide attempts (Björkenstam, et al. 2017; Denney 2010; Denney et al. 2009; Maimon and Kuhl 2008; Maimon, et al. 2010). Differences by gender do not consistently appear in analyses of family structure, but some authors report stronger associations at younger ages (Stockard and O’Brien 2006), when the impact of family of orientation may be strongest.

Importantly, the magnitude of these effects varies by socio-political context. Associations of suicide rates with cohort-related demographic measures of social integration and regulation are substantially weaker in Continental European than English-speaking nations (Pampel 1996; Pampel and Williamson 2001; Stockard and O’Brien 2002a, 2006). While the literature is relatively extensive, the studies do not incorporate data from the current century. Nor are the samples large enough to explore possible interactive effects of age, social-political contexts, and cohort-related indicators of regulation and integration. The second research question addresses this area, examining the extent to which variations in age-specific suicide rates are associated with cohort-related indicators of social regulation and integration in childhood and the extent to which these associations vary across national context, gender, and age. Based on previous literature, it was expected that cohorts with childhood demographic characteristics associated with less social integration and regulation would have higher age-specific suicide rates but that this impact would be smaller in national contexts with greater social integration and regulation. It was also expected that the impact of childhood-related demographic characteristics involving family structure would be stronger at younger ages.

Historical events at formative ages and social integration and regulation

Cohort theorists interested in attitudes and life outcomes have described the important role of historical experiences in the formative ages of late adolescence and young adulthood, such as the impact, for men, of entering the labor force at the height of the Great Depression (Elder 1974), or, for women, of coming of age before women were allowed to vote (Firebaugh and Chen 1995). People who were in late adolescence or early adulthood at the time of a given war or conflict are more likely to cite this experience as especially memorable, and these experiences are related to cohort differences in attitudes and actions (Corning and Schuman 2015; Elder, et al. 2002; Mannheim, 1928/1952; Schuman and Corning 2006, 2012; Schuman and Scott 1989). Much of Durkheim’s analysis involved the ways in which historical events altered social integration and regulation and impacted suicide rates. While his analysis involved periodic changes for total populations, one could hypothesize that changes in social integration associated with these events would have a greater impact on those in their formative years.

Durkheim described how periods of war could promote social integration and regulation, resulting in lower suicide rates. Subsequent studies have generally supported his findings, documenting decreased rates during conflicts such as the world wars of the 20th century (Lester 1994, Snowdon and Hunt 2002) and the Sri Lankan civil war (Aida 2020). Although Henderson and associates (2006) noted that changes in the suicide rate in Scotland during World War II varied by age, there do not appear to be direct tests of the association of experiencing times of war in formative years with cohort variations in suicide.

Durkheim also contended that economic changes affect integration and regulation and thus suicide rates. While he suggested that periods of increased prosperity “act on suicide in the same way as economic disasters” (Durkheim, 1897/2006, p. 264), contemporary analysts have found different patterns for varying types of economic change. In general, suicide rates appear to be lower in periods of economic growth and prosperity, but higher in times of economic stress (e.g. Fountoulakis, et al. 2014; Maag 2008; Milner, et al. 2012; Phillips and Nugent 2014; Pridemore, et al. 2007; Richardson et al. 2020; Solano, et al., 2011; Stuckler, et al. 2009a; Stuckler, et al. 2009b; Thomas and Gunnell 2010). While not directly examining cohort effects, studies of youth suicide in the United States have documented significantly higher rates in locales with greater levels of disadvantage (Kubrin and Wadsworth 2009; Kubrin, et al. 2006; Wadsworth, et al. 2014), and a study of trends in youth suicide in New Zealand (Curtis and Curtis 2011; Curtis, et al. 2013) suggests that economic changes were as, if not more, important than demographic factors in explaining variations in suicide rates.

Health pandemics certainly occur less frequently than wars or economic change. While Durkheim did not address the potential association of pandemics with social integration and regulation, other scholars, building on his insights, have suggested that health crises, like wars, could promote greater integration and regulation and thus lower suicide rates. Studies of changes in suicide rates for the total population during the 1918 flu pandemic (Bastiampillai, et al. 2021; Gaddy, 2021, Hughes, 2022. Wasserman, 1992) and the more recent COVID-19 pandemic (Leske, et al. 2021, Pirkis, et al. 2021) have found either no periodic change in suicide rates or, as Durkheimian theory would predict, slight declines. However, studies focused on younger populations during the recent pandemic documented increased rates (Bridge, et al. 2023; Goto, et al. 2022). This suggests that the impact of a health pandemic could be especially influential during formative years.

The third research question builds on this research, examining the possibility that cohorts that experienced historical events related to war, economic depression, or health pandemics in their formative years have suicide rates that differ from other cohorts and that these associations might vary across national context, gender, and age. Based on the literature regarding period effects it was expected that cohorts in their formative years in times of war, experiencing greater social integration and regulation, would have lower suicide rates, while those in their formative years during severe economic crises or health pandemics would have higher rates. Building on the literature regarding associations with cohort characteristics from childhood, it was also expected that negative effects associated with these formative experiences would be muted and positive effects would be enhanced within national contexts that embody greater levels of social integration and regulation.

Social change and social integration and regulation

Decades of literature suggest that rapid social change is associated with suicide rates. As noted above, economic downturns are thought to weaken social integration and regulation, leading to higher suicide rates, while rapid economic growth is associated with lower rates. Studies also document the association of family-related change and suicide. Durkheim hypothesized that higher rates of divorce and separations led to declining integration and regulation and documented higher suicide rates in locales and periods with these conditions (1987/2004, p. 284). Empirical studies, using a variety of samples, measures, and methodological approaches, have supported his conclusion. They consistently report higher suicide rates in times of greater family change (e.g. Denney 2010; Gibbs 1969; Maimon, Browning, and Brooks-Gunn 2010; Messner et al. 2006; Pampel and Williamson 2001; Stockard and O’Brien 2002a).

A few studies directly examine the association of social change and cohort variations in suicide rates. However, most use either cross-sectional or time series designs, and those that address cohort effects do not examine the extent to which the magnitude of such effects might alter with controls for the pace of social change. Some research suggests that interactions could occur. For instance, Messner et al.'s time series analysis of suicide rates of 15–19-year-olds found that, for men, cohorts with a higher level of non-marital births had higher suicide rates during periods of higher divorce. Using data from the United States, Marshall (1981) found that the association of periods of war with lower suicide rates could be partially explained by improved economic conditions. In short, previous research appears not to have fully examined the possibility that the magnitude of cohort effects on suicide might be influenced by the pace of social change nor the extent to which these period level impacts are independent of cohort-related variables and the extent to which they are modified by socio-political context. The fourth research question tests this possibility, examining the extent to which associations of cohort characteristics and suicide rates remain when period-related indicators of social change are considered. Previous literature suggests that age-specific rates would be higher during periods of rapid social change, but it is unclear if the pace of change would impact associations of these rates with cohort characteristics.

Methodology

Age-specific suicide rates for men and women from 1950 to 2020 in 19 highly industrialized, western nations are examined. The sample includes six English-speaking countries (Australia, Canada, Ireland, New Zealand, the United Kingdom, and the United States) and thirteen continental European nations, representing the Nordic (Denmark, Finland, Norway, and Sweden), Germanic (Austria and Switzerland), and Romance (Belgium, France, Greece, Italy, Netherlands, Portugal, and Spain) families of nations. Twenty-six birth cohorts are included. The oldest, cohort 1, born in 1875 to 1879, was 15–19 years of age in 1890–1894, and 70–74 in 1950. The youngest, cohort 26, was born 125 years later (2000–2004) and was 15–19 in 2020.

Table 1 summarizes this pattern, showing cohorts in each age group across selected years. Following the standard methodology in age–period–cohort analyses, one can trace the path of individual cohorts through time. For instance, cohort 13, which was born in 1935–1939, was 15–19 in 1955, 20–24 in 1960, 25–29 in 1965, and 70–74 in 2010. Suicide rates for all age groups are available for cohorts 12 to 15; and for three-fourths of the age groups and periods for cohorts 9 to 18. Data on cohort characteristics are not available for all cohorts in some countries, so the number of observations ranges from 97 to 180 per country (mean = 153). In total the analyses include 3092 age-period-country specific suicide rates for each gender group, a sample 24 percent larger than the most extensive previous analyses of cohort characteristics and suicide rates (Stockard and O’Brien, 2006).

Measures

The World Health Organization mortality database reports the annual number of deaths, by cause and gender, and population by gender in 5-year age categories beginning in 1950 (WHO, n.d.). These data were used to calculate age-period-country specific suicide rates within each gender group from 1950 through 2020 for ages 15–19 through 70–74. For a few nations, data for 2020 were obtained from national statistical bureaus.

Following earlier studies, two demographic indicators of cohort-related integration and regulation in childhood are used, both expressed as percentages. Relative cohort size is measured as the number of youth relative to the total population of working age adults (e.g. the population aged 15–19 relative to those aged 15–64). Family structure is measured as the percentage of births within a cohort that were outside of marriage.

Dummy variables indicate if a cohort experienced historical events that could be expected to influence social integration and regulation. Cohorts 7, 8, and 9 were in late adolescence or early adulthood (ages 15–14) during the Great Depression of the 1930s,¹ and cohorts 5, 6, 25 (age 20–24 only), and 26 were in these formative ages during the global pandemics of 1918–1919 or 2020–2022. Table 2 lists the cohorts and nations that experienced wars during their formative years. Seventeen conflicts were distinguished including world wars of the 20^th century, civil wars, and colonial or neo-colonial conflicts.

Two measures indicate the pace of social change. The measure of family change is based on a scale developed by Pampel and Williamson (2001) and includes indicators of women’s labor force participation, marriage, and fertility. A period-specific measure was created for each nation by calculating the difference of a score in one period from that in the previous period. The annual rate of inflation measures economic change and was obtained from the OECD (n.d.).² Higher values on both variables indicate a greater pace of change.

Three indicators of national socio-political context are used. Two are factor scores derived from a principal component analysis of six country level indicators: income inequality (the GINI index), obtained from the Standardized World Income Inequality Database (SWIID, Solt 2019); gross domestic product (GDP) per capita, adjusted for purchasing power parity (PPP); a human capital index, based on the average years of education and returns to education; (Feenstra, et al., 2015); Pampel and associates’ country level measures of collectivism and women friendly institutions (Pampel 1993, Pampel and Williamson 2001); and Hofstede and associates’ measure of individualist versus collectivist values (Hofstede 2011, Hofstede, et al. 2010). Higher scores on one factor indicate stronger collectivist and women friendly institutions, less income inequality and a lower GDP. Higher scores on the second factor indicate greater adherence to individualist than collectivist values and higher average levels of education and GDP. Extensive preliminary analyses indicated that, paralleling results in earlier literature, the association of cohort characteristics and suicide rates differed markedly between the English-speaking and the continental European families of nations. Results, including interaction effects, associated with the continental families were very similar and contrasted sharply with those associated with the English-speaking family. Thus, to simplify interpretations, the third measure of national context is a dummy variable, distinguishing English-speaking countries from those in continental Europe. Results are presented separately for the English-speaking and continental nations.

Analysis

Descriptive statistics are used to examine characteristics of the sample. Estimable function analyses are used to address the first research question regarding the magnitude of age, period, and cohort effects that can be uniquely attributed to each of these factors (O’Brien, 2014). The proportions of explained variance in age-period-specific suicide rates are calculated for each nation and gender group. Average proportions in country groupings are compared using t-tests and effect sizes (Cohen’s d).

APCC models, modeled on the work of Stockard and O’Brien (2002a, 2006), supplement this analysis and address the remaining three questions.³ By using dummy variables for age and period and single variables for cohort characteristics, APCC models avoid the issue of non-independence of age, period and cohort (O’Brien 2000). Nation was treated as a random variable.⁴ The xtmixed procedure in STATA was used.

Analysis began with a baseline, intercept only model including random effects of nation. Then four increasingly more complex models were examined adding: (1) dummy variables of age and period, (2) measures of cohort characteristics, (3) interactions of younger (15–29) and older (60–74) age groups with cohort characteristics, and (4) measures of pace of change and national context. Change in model fit was assessed with −2 Log Likelihood (−2LL) and proportionate reduction of error (PRE) statistics. Changes in model fit from baseline to Model 1 were used to address research question one regarding the magnitude of cohort effects. Fixed effects were examined to address research questions regarding the impact of cohort characteristics on suicide rates (questions 2 and 3) and the extent to which the associations remained with additional controls (question 4). Analyses were conducted within each gender and nation group. To facilitate interpretation, the dependent variable and continuous measures of cohort characteristics were logged.

Sensitivity analyses examined if variations in data availability, coding decisions, and use of random, rather than fixed, effects alter the results. In addition, because the United States differs so markedly from other developed countries in its extraordinarily high rate of gun ownership and low scores on measures of national context, the impact of its inclusion was examined. The supplemental file contains additional details.

Results

Characteristics of the sample

Descriptive statistics indicate substantial variability on all measures. The first four columns in Table 3 give the mean and standard deviation of the age-specific suicide rates for men and women in each nation. As documented since the 19th century, the highest average values are in the northern European countries and the lowest in southern Europe.

The last four columns in Table 3 report descriptive statistics for the two measures of childhood related cohort characteristics: relative cohort size and non-marital births. There is substantially more inter-country variation in cohort size than in family structure. However, in all nations, for both variables, there is substantial inter-cohort variation. While the largest relative cohort sizes are associated with the oldest cohorts and the smallest with the most recent, changes over time were not linear, with decreases often coinciding with times of war or economic downturns and increases after those periods. Similarly, changes in the measure of family structure were not linear over time.

There is also substantial variability in the extent to which cohorts experienced one of the historical events hypothesized to influence social integration and regulation. Of the 442 cohorts in the analysis, almost two-thirds (63 percent) did not experience any of the events at formative ages, and about a third (32 percent) experienced one event. All countries had at least two cohorts that experienced a global pandemic and/or the Great Depression. However, there was substantial inter-country variation in the extent to which cohorts experienced warfare in their formative years. No conflicts were recorded for cohorts in Ireland, Sweden, and Switzerland,⁵ and cohorts in the English-speaking countries were more than twice as likely to have experienced times of war in their formative years (28 percent versus 13 percent, t = 3.97, df = 444, Cohen’s d = 0.39).

The first four columns of Table 4 report descriptive statistics on the measures of social change for each nation and also indicate substantial variability. Rates of inflation were slightly greater in earlier periods. Family change was greater in some periods than others, but there were no differences between countries in the average value of this measure nor any parallel pattern in the rate of change across nations. In other words, over the 70-year timespan in the analysis, all nations had similar amounts of family-related change, but the period in which such change occurred differed.

The final columns of Table 4 report values on the continuous measures of socio-political context. The English-speaking nations are the only countries with low scores on the factor associated with collectivist institutions and equality and high scores on the factor associated with individualist values and education. The United States has the most extreme scores on both measures. Paralleling theoretical descriptions above, the Nordic nations have the highest scores on the measure of collectivist institutions, but near the mean on the measure associated with individualist values. Overall, there is very little overlap of values for the English-speaking countries and those for the other nations. These data support the decision to examine results separately for the English-speaking family of nations and other countries.

Research question one: the magnitude of cohort effects

As expected, cohort effects are larger in the English-speaking countries which have traditionally provided less institutionalized support associated with social integration and regulation. Table 5 reports results of the estimable function analysis that compares the unique amount of variance accounted for by age, period and cohort for each nation and gender. The first column gives the total proportion of variance in age-specific suicide rates jointly explained by age, period, and cohort. The remaining three columns report the proportion of the total that is uniquely associated with each of these elements. Results for men are in the first panel and those for women in the second. Comparisons of the two nation groups are in the bottom two lines of each panel. Within each nation at least 70 percent of the total variance is associated with age, period, and cohort, and there are no significant differences between the two nation groups in this proportion. But the two groups differ substantially in the proportion of variance that can be uniquely attributed to each factor, with higher proportions within the English-speaking nations. All the differences exceed 0.60 of a standard deviation, and 5 of the 6 comparisons are statistically significant, even with small degrees of freedom (df = 17).

The model fit statistics for the mixed model analyses (Table 6) support this conclusion. When dummy variables for age and period are added to the intercept only model (Model 1) the proportionate reduction in residual variance (PRE) is substantially larger in the continental nations (0.64 for men and 0.57 for women) than in the English-speaking nations (0.37 and 0.36). Correspondingly, the PRE values associated with adding cohort characteristics (Models 2 and 3) are substantially larger in the English-speaking nations.

Research question two: childhood cohort-related variables and suicide

Model fit statistics in Table 6 indicate that, for each nation-gender group, Model 4, which includes dummy variables of age and period, measures of cohort characteristics, interactions of age with cohort characteristics, and the measures of pace of change and national context, provides the best fit. Table 7 reports the fixed effects associated with this model. A substantial proportion of the interactions of cohort characteristics and age are statistically significant, but the nature of the interactions varies across the four nation-gender groups. Thus, to aid interpretation, Table 8 reports fixed effects associated with each cohort characteristic, by age and gender, calculated from the results in Table 7.

Results related to the childhood-related measures of cohort integration and regulation are in the top two panels of Table 8. Because these measures and the dependent variable are logged, the associated fixed effects can be interpreted as a proportionate change in the age-specific suicide rate. As hypothesized, the results generally replicate those found in earlier studies. Suicide rates are usually higher for birth cohorts that are relatively larger and have higher proportions of non-marital births. But these associations are substantially stronger in the English-speaking nations with less institutionalized support. For instance, in the English-speaking nations, a one percent change in cohort size would be predicted, net of other variables in the models, to be associated with a 0.74 to 2.36 percent increase in the age-specific suicide rate, but 0.59 or less in the continental nations. Similar differences appear with the measure of family structure. For men, coefficients range from 0.25 to 0.56 across age groups in the English-speaking nations but are near zero (0.02 to 0.05) in the continental nations. For women at middle age or younger, coefficients range from 0.31 to 0.79 in the English-speaking nations, but equal 0.09 in the continental countries.

Within each set of nations, the estimated effects vary by age. Paralleling results from earlier years, effects associated with family structure are stronger at younger ages, particularly in the English-speaking countries. In contrast, and contradicting some earlier findings, effects associated with relative cohort size are largest at oldest ages. Differences by gender are relatively small, but more marked at both younger and older ages.

Research question 3: experiences in formative years

The results related to experiences at formative ages are in the bottom three panels of Table 8 and indicate that such experiences are significantly related to cohort variations in suicide. But, as with the cohort characteristics associated with childhood, the magnitude of association varies across nation groups and by age and gender.

Based on the literature regarding period effects, it was expected that cohorts in their formative years during times of war would experience greater social integration and regulation and thus have lower suicide rates. Yet, this association only appeared within the continental nations, and more strongly for men and at younger ages. In the English-speaking nations there was no association of suicide rates with such experiences at any age. In other words, a “protective” pattern of increased integration and regulation from experiencing war in the formative years was limited to men within the continental nations.

Results regarding experiencing the Great Depression during formative years parallel the literature regarding period effects. In both nation groups, cohorts in their formative years during this economic crisis had higher suicide rates. The associations were greater for men at younger and middle age and greater for women at older ages. However, in all but one comparison of coefficients within each age-gender group, the predicted fixed effects were larger in the English-speaking nations, paralleling the pattern with other cohort characteristics.

The results related to experiencing a health pandemic at formative ages contrast rather sharply to previous findings regarding period effects. While the literature suggests health pandemics can increase social regulation and integration thus resulting in lower suicide rates, the opposite pattern appears in the analysis of cohort effects, although the pattern varies by gender, age and national context. Cohorts that were in their formative years during the flu pandemic of the early 20th century (and at older age groups in the present sample) had significantly higher age-period specific suicide rates than would be expected given other model variables. The associated effects were larger for men in both sets of countries and markedly smaller for women at middle age in the continental nations. Women cohorts that were in their formative years during the recent COVID pandemic (in the younger age group) had higher suicide rates than would be expected given other model variables. Within this gender-age group, the effects were larger within the continental nations, the only reversal of the general pattern.

Research question 4: controlling for the pace of change

Fixed effects associated with the measures of social change parallel patterns reported in other studies. In all four country-gender groups, age-period specific suicide rates were significantly higher in times of greater family change and lower during rapid economic growth. While the coefficients associated with economic change were relatively similar across the analyses, those associated with family change were more than twice as large within the English-speaking nations. Thus, paralleling results with cohort characteristics, the impact of family change on suicide rates appears to be moderated within the continental countries with stronger institutional and cultural sources of social integration and regulation. Importantly, there were no statistically or substantively significant differences in the fixed effects associated with cohort characteristics when the measures of social change were added to the models.

Fixed effects associated with the nation-level measures of socio-political context provide additional controls and also help confirm the findings described above. Within the continental nations, women’s age-period specific suicide rates were significantly higher in nations whose residents more often express individualist values. Within the English-speaking nations, rates were higher for both men and women with these value orientations and also higher for women within countries with stronger collectivist and women-friendly institutions. Importantly, however, the associations of cohort characteristics and age-specific suicide rates were unchanged when nation-level variables were added to the models.⁶

Sensitivity analyses

Sensitivity analyses confirmed the findings reported above. Fixed effects associated with cohort characteristics were substantively identical when the analysis distinguished the four families of nations, when the sample was limited to nations with data on all periods in the analysis or restricted to only ages and periods in which a suicide was reported. Patterns remained with altering the measure of the oldest age group and examining Nordic countries separately from other continental countries. They also remained when the United States was omitted from the analysis or indicated by a dummy variable. Finally, the results were unchanged with a fixed, rather than random, effects model. Details in supplemental file and Tables S14 to S25.

Summary and discussion

The results support the general tenets of Durkheimian theory regarding the impact of social integration and regulation on wellbeing and the expectations outlined above regarding cohort inequality. As expected, cohort variations in wellbeing are significantly greater in the English-speaking nations, which have traditionally provided less institutionalized support (research question one). Age-specific suicide rates are larger for cohorts with childhood demographic characteristics as well as experiences in the formative years associated with less social integration and regulation. However, these associations are markedly greater in the English-speaking nations and vary by age and gender (research questions two and three). Importantly, the results remain with strong controls for the pace of social change and additional measures of national context (research question four), as well as extensive sensitivity analyses. The results suggest important directions for future work related to cohort inequality and wellbeing.

The importance of cross-cultural analyses

The strong differences between the English-speaking and continental nations in the magnitude and nature of cohort effects highlight the importance of cross-cultural analyses. If the work had been restricted to the United States or even several English-speaking nations, the conclusions would clearly have been different and far from accurate. Differences between nation groups occurred with all the cohort characteristics examined, indicating the potential power of the seemingly simplistic concept of “families of nations.” At the same time, the strongest differences occurred with childhood-related characteristics, reflecting the key role of institutionalized support for families and children in distinguishing the groups.

Perhaps surprisingly, the sensitivity analyses showed that results were similar in the United States and other English-speaking countries. The United States provides markedly less institutionalized support than other English-speaking nations. Some social policies in other English-speaking countries resemble those of the continental countries and several of these nations have also implemented policy changes focusing on youth suicide (e.g. AGNMHC, 2017; Ministry of Health, 2019). It will be important to examine long-term impacts of these changes on variations within the English-speaking family and the patterns found in this analysis.

The United States also differs markedly in the availability of lethal weapons and has an extraordinarily high homicide rate. In fact, the homicide rates are so low in other sampled countries that valid age-specific rates could not be examined. Previous research has documented the association of the childhood-related cohort characteristics used in this study with cohort variations in homicide in the United States (e.g. O’Brien and Stockard, 2006). Thus, while cohorts in the United States may have a similar risk of suicide compared to cohorts with similar characteristics in other English-speaking nations, they have a substantially larger risk of homicide. This “double toll” of cohort inequality in the United States should not be forgotten.

Experiences in formative years

The impact of historical events at formative ages on cohort variations in wellbeing has not been examined in earlier research. The results suggest several avenues for future research, both in developing more precise measures of events and in identifying other important historical experiences. For instance, while Durkheim suggested that periods of war could heighten integration resulting in lower suicide rates, this pattern only appeared, albeit quite strongly, for cohorts within the continental countries. While partly due to the higher levels of social integration within the continental nations, this pattern could also reflect differences in the extent to which national conflicts disrupted the social order. Recall that cohorts in the English-speaking countries were more than twice as likely as those in the continental nations to have experienced wartime in their formative years and a substantial number of the conflicts were colonial or neo-colonial in nature. More precise measures could potentially tap the extent to which national conflicts impacted social integration.⁷

The hypothesized negative effect of experiencing the Great Depression occurred in both groups of nations and was, as hypothesized, slightly stronger within the English-speaking countries. But, because there was no similar event that lasted long enough to match the age span within the defined cohorts, the measure involved only one economic downturn. Future research could potentially involve more precise measures of formative experiences related to economic change. This work could examine alternative age groupings and include both positive and negative economic changes.

The results related to experiencing the health pandemics of the early 20th and 21st centuries differed from expectations based on literature regarding period effects. While earlier literature suggested little to no impact of the events for the total population, the results indicated increased rates at older ages (involving those who experienced the early 20th century pandemic) in both sets of countries and at younger ages (those who experienced the 21st century event) for women, especially in the continental nations. Although the analysis involved only two events and relatively few cohorts, the contrast illustrates the way in which period effects can differ from cohort effects and could be seen as highlighting the importance of formative experiences. Future research should examine any continuing impact of the COVID-19 pandemic on suicide rates. Might the cohorts that were young during the recent pandemic have higher suicide rates through the life span? And to what extent will the greater risk for women, especially in the continental nations, persist at older ages?

Finally, future researchers could explore the association of other historical experiences in formative years on cohort variations in wellbeing. One could hypothesize that innovations associated with communication and transportation might enhance possibilities of social interaction and thus increase social integration. For example, within the present sample the oldest cohorts might have experienced the advent of the telephone and automobiles at formative ages, while more recent cohorts experienced the development of electronic communications and widely available air travel. One could also hypothesize that periods of political repression and intense social upheaval could impact social integration and affect cohort variations in wellbeing. While some events might be nation-specific, others such as climate change, could be global in scope much like the Great Depression.

The impact of age and gender

While cohort effects appear throughout the life cycle, the results revealed differences by age. One could hypothesize that some differences reflect variations over the life cycle in the extent to which specific cohort characteristics are associated with social integration. For instance, in the English-speaking countries, the fixed effects associated with the measures of family structure were substantially larger at younger ages, reflecting the importance of adult attention at that time. Fixed effects associated with cohort size were weaker in midlife, a time with potentially more sources of social support and integration. The protective effect of war experiences and the negative effects of having experienced the Great Depression were strongest at younger ages and midlife, suggesting that the impact of these events might lessen as cohorts age and events become more distant. Notably, however, the results associated with experiencing health pandemics did not conform with this pattern, for there were stronger effects at older ages. While this experience involved relatively few cohorts, it deserves further research and theoretical attention.

Gender differences were most striking with the measures associated with formative experiences related to war and economic crises, with substantially larger effects for men than women. It is reasonable to hypothesize that these differences reflect gender roles and expectations, with men more likely than women, especially through the lifespans of cohorts in the analysis, to have been more directly impacted by these events. Because gender roles have changed markedly over recent decades, future research could examine impacts of these historical trends. In addition, while the present analysis included changes in gender roles as part of the measure of family change, future research could consider developing contextual measures of changes in gender roles and examining the extent to which changing gender-related contexts are associated with the effects of cohort characteristics.

The importance of replication

This paper focused on western democracies to include the longest possible data series and nations with relatively similar political histories in the post-World War II era. Future studies could examine a much more diverse set of nations and data from periods and cohorts before those in this analysis. Several European countries have long series of vital statistics, beginning in the early years of the 19th century. The APCC approach is quite flexible and can easily handle interruptions in data patterns that might occur in very long data series. Such longer periods would facilitate analysis of variations in associations over time and greatly expand the possible range of substantive measures of social change.

By definition, studies of cohort effects involve social change. Thus, it is important to replicate this work as new cohorts enter the population, younger cohorts enter older ages, cohorts experience new events at formative ages, and nations introduce policies directed toward youth wellbeing. While suicide rates are especially useful in studying cohort variations over long timespans and multiple national contexts, researchers should also study variations in other aspects of wellbeing, including physical and mental health. The present paper illustrates the importance of conceptualizing social integration at multiple levels of analysis and examining their joint and interactive impact on wellbeing. Future research could incorporate contextual measures at levels such as the family and community and dependent variables at the individual or community level. While often limited to one national context, longitudinal data sets that allow such exploration and involve alternative measures of wellbeing are increasingly available (e.g. Harris 2013).

Understanding the multiple levels at which social integration can occur is also important for those concerned with mitigating cohort inequality. For instance, while the social challenges associated with global climate change will affect people of all ages, one could predict, based on the results described above, that the impact of these changes will be especially strong for cohorts at formative ages. However, the results also suggest that these negative impacts could be mitigated with strong social integration from other levels of analysis, potentially involving not just nation states, but communities, families, and other social groups. Understanding that social integration and social support can derive from multiple contexts could help address cohort inequality both now and in the future.

About the author

Jean Stockard is Professor Emerita at the University of Oregon and taught in Departments of Sociology and Planning, Public Policy, and Management. Her research has often examined factors associated with social wellbeing, incorporating work from sociology of education, gender, and human development. Recent publications include All Students Can Succeed and A Decade of Change and Continuity in Midlife. This paper builds on a series of work with Robert O’Brien regarding cohort variations in lethal violence.

Acknowledgements

Appreciation is expressed to Nicole Ngo, Robert O’Brien, Luis Sanchez, and the anonymous reviewers for helpful comments on earlier drafts and to Rachel Margolis for sharing data on family change in Canada.

Funding

None declared.

Conflicts of interest

None declared.

Data availability

The data underlying this article will be shared upon request.

Footnotes

References