The US reports the highest suicide rate (14 deaths per 100,000 people) among all high-income countries (Tikkanen & Abrams, Reference Tikkanen and Abrams2020). Experts attribute this trend to increased national despair that may approximate the bleak social and economic outlook experienced by a society on average (Case & Deaton, Reference Case and Deaton2020). Not surprisingly, suicides appear to track or correlate positively with some aggregate indicators of national despair (Agrrawal et al., Reference Agrrawal, Waggle and Sandweiss2017; Collins et al., Reference Collins, Cox, Kizys, Haynes, Machin and Sampson2021; Heilbron et al., Reference Heilbron, Franklin, Guerry, Prinstein and Nock2014; Lee et al., Reference Lee, Lee, Myung, Song, Lee, Kim, Carroll and Kim2018; Phillips & Nugent, Reference Phillips and Nugent2014; Won et al., Reference Won, Myung, Song, Lee, Kim, Carroll and Kim2013).

In his seminal treatise on social determinants of suicides, Durkheim (Reference Durkheim, Spaulding and Simpson1897/Reference Durkheim, Spaulding and Simpson1951) posited that suicide prevalence in a population is a consequence of the underlying social structure, beliefs and practices that he referred to as collective inclination (Durkheim, Reference Durkheim, Spaulding and Simpson1897/Reference Durkheim, Spaulding and Simpson1951). A population’s collective inclination towards suicide may exert coercive pressure on individuals towards suicide completion and may vary based on social despair during stressful periods (Durkheim, Reference Durkheim, Spaulding and Simpson1897/Reference Durkheim, Spaulding and Simpson1951). As an extreme consequence of increased despair, trends in suicides may reflect perturbations in aggregate sentiments of despair across place and time (Larsen et al., Reference Larsen, Boonstra, Batterham, O’Dea, Paris and Christensen2015; Lee et al., Reference Lee, Lee, Myung, Song, Lee, Kim, Carroll and Kim2018; Won et al., Reference Won, Myung, Song, Lee, Kim, Carroll and Kim2013). Ecologically, collective despair or (conversely) optimism may reflect a population’s shared outlook (Carver et al., Reference Carver, Scheier and Segerstrom2010) and manifest as contagious, collective expectations about the future (Bennett, Reference Bennett2011; Catalano et al., Reference Catalano, Goldman-Mellor, Karasek, Gemmill, Casey, Elser, Bruckner and Hartig2020; Herzberg et al., Reference Herzberg, Glaesmer and Hoyer2006; Hooker et al., Reference Hooker, Monahan, Shifren and Hutchinson1992). Suicides, by definition, reflect termination of the future (Baumeister, Reference Baumeister1990; Martin et al., Reference Martin, LaCroix, Novak and Ghahramanlou-Holloway2020; Shneidman, Reference Shneidman, White and Bruner2006). Whereas research on suicides and optimism largely focuses on individual-level mechanisms, national trends in select measures of optimism appear to explain (to an extent) spatial and temporal variation in suicides (Agrrawal et al., Reference Agrrawal, Waggle and Sandweiss2017; Collins et al., Reference Collins, Cox, Kizys, Haynes, Machin and Sampson2021; Heilbron et al., Reference Heilbron, Franklin, Guerry, Prinstein and Nock2014; Lee et al., Reference Lee, Lee, Myung, Song, Lee, Kim, Carroll and Kim2018; Phillips & Nugent, Reference Phillips and Nugent2014; Won et al., Reference Won, Myung, Song, Lee, Kim, Carroll and Kim2013). Collective optimism, however, remains a complex construct that may derive not just from aggregate economic expectations, but also social integration, cohesion and social capital (Anglin et al., Reference Anglin, McKenny and Short2018; Gómez et al., Reference Gómez, Kleinman, Pronk, Gordon, Ochiai, Blakey, Johnson and Brewer2021; Olson, Reference Olson2006; Stack, Reference Stack2000a, Reference Stack2000b).

Collective optimism refers to the shared belief or positive outlook that a group of individuals holds regarding the future of their collective endeavors (Anglin et al., Reference Anglin, McKenny and Short2018; Catalano et al., Reference Catalano, Goldman-Mellor, Karasek, Gemmill, Casey, Elser, Bruckner and Hartig2020). Other scholars interpret it as the collective faith that, as a society, community, or group, things will generally improve, progress, or work out for the better (Anglin et al., Reference Anglin, McKenny and Short2018). This optimism pertains not only to the expectations of individual members but also to a collective sentiment that the group shares (Anglin et al., Reference Anglin, McKenny and Short2018). Collective optimism may play a crucial role in motivating and uniting societies to work towards common goals and may underlie social change, resilience and cooperation, by potentially fostering a sense of hope and purpose (Anglin et al., Reference Anglin, McKenny and Short2018). Various factors, such as shared values, leadership, and the prevailing psychosocial climate may influence the temporal variation of collective optimism in a population (Anglin et al., Reference Anglin, McKenny and Short2018). Quantification of collective optimism in a manner that accounts for all facets of this construct remains difficult. However, as suggested by Durkheim, to the extent that suicides correspond with a society’s collective experiences of despair and future outlook, the patterning of suicides in a population may signal periods of acute decline in collective optimism (Durkheim, Reference Durkheim, Spaulding and Simpson1897/Reference Durkheim, Spaulding and Simpson1951).

Decline in collective optimism may also manifest as risk aversion or reduced willingness to invest in high-cost, uncertain ventures (Gómez et al., Reference Gómez, Kleinman, Pronk, Gordon, Ochiai, Blakey, Johnson and Brewer2021). Uncertainty and unfavorable expectations about the future correspond with reduction or postponement of expensive, long-term investments such as fertility, home purchase or new business creation (Anglin et al., Reference Anglin, McKenny and Short2018; Karasek et al., Reference Karasek, Goodman, Gemmill, Falconi, Hartig, Magganas and Catalano2015; Puri & Robinson, Reference Puri and Robinson2007). Childbearing remains one of the most fundamental, long-term investments over the life course (Hirshfield & Tinkle, Reference Hirshfield and Tinkle1975; Nurmi, Reference Nurmi, Strathman and Joireman2006; Vitzthum, Reference Vitzthum2008). Substantial research documents reproductive risk aversion during stressful ambient circumstances (Bruckner & Catalano, Reference Bruckner and Catalano2018) that, in turn, may also correspond with reduced collective optimism (Karasek et al., Reference Karasek, Goodman, Gemmill, Falconi, Hartig, Magganas and Catalano2015). Parental and social optimism appears to vary inversely with risk aversion, which spills over into conscious and non-conscious decisions regarding offspring conception, sex-specific fetal loss and fitness of resulting live births (Forbes & Mock, Reference Forbes and Mock1998; Trivers & Willard, Reference Trivers and Willard1973).

Selection in utero, defined as spontaneous abortion of gestations (also referred to as fetal loss), offers one mechanism through which populations may respond to heightened uncertainty, or reduced collective optimism (Bruckner & Catalano, Reference Bruckner and Catalano2018). Selection in utero against frail or risky gestations, by way of increased spontaneous abortion or fetal loss, may indicate a pregnant person’s non-conscious, biological risk-averse response to unfavorable external circumstances (Bruckner & Catalano, Reference Bruckner and Catalano2018). These unfavorable circumstances may diminish a frail infant’s survival or reproductive success should the pregnancy result in a live birth (Trivers & Willard, Reference Trivers and Willard1973). Relative to females, male gestations appear particularly vulnerable to selection in utero owing to their greater demand on maternal caregiving resources but lower likelihood of yielding grandchildren (i.e., reproductive success) if born in stressful conditions (Bruckner & Catalano, Reference Bruckner and Catalano2018; Trivers & Willard, Reference Trivers and Willard1973). Male twin gestations, on average, occupy the right tail of the gestational frailty distribution and fare poorer in terms of relative survival, longevity and (future) reproductive success relative to all other types of gestations (Bolund et al., Reference Bolund, Lummaa, Smith, Hanson and Maklakov2016; Lummaa et al., Reference Lummaa, Haukioja, Lemmetyinen and Pikkola1998). Hence, the incidence of male twin live births (or selection in utero against male twins) may indicate the extent to which a population is willing to invest in risky gestations that, in turn, may gauge a population’s sensitivity to variations in collective optimism (Catalano et al., Reference Catalano, Saxton, Gemmill and Hartig2016, Reference Catalano, Goldman-Mellor, Karasek, Gemmill, Casey, Elser, Bruckner and Hartig2020; Karasek et al., Reference Karasek, Goodman, Gemmill, Falconi, Hartig, Magganas and Catalano2015)

Collective optimism may serve as a shared precursor of both (1) increase in suicides and (2) decline in male twin live births (Catalano et al., Reference Catalano, Goldman-Mellor, Karasek, Gemmill, Casey, Elser, Bruckner and Hartig2020). If a population exhibits risk aversion following sudden perturbations in collective optimism, temporal changes in one of the most acute indicators of despair — suicides — should precede changes in population-level markers of selection in utero, such as the incidence of male twin live births. Two recent studies support this hypothesis (Catalano et al., Reference Catalano, Goldman-Mellor, Karasek, Gemmill, Casey, Elser, Bruckner and Hartig2020; Karasek et al., Reference Karasek, Goodman, Gemmill, Falconi, Hartig, Magganas and Catalano2015). Karasek and colleagues (Reference Karasek, Goodman, Gemmill, Falconi, Hartig, Magganas and Catalano2015) report an inverse association between a measure of consumer confidence index among reproductive aged women (indicative of macroeconomic climate) and male twin births in Sweden. This research suggests that women may ‘sense’ ambient macroeconomic disturbances and, as a risk-averse response, yield fewer-than-expected male twins in the short term (two months following extreme decline in consumer confidence). Catalano and colleagues (Reference Catalano, Goldman-Mellor, Karasek, Gemmill, Casey, Elser, Bruckner and Hartig2020) test this phenomenon through the lens of collective optimism and examine the relation between suicides among women aged 15−49 years and male twin births in Sweden. They find an inverse association between female suicides and male twinning in that greater-than-expected incidence of suicide among reproductive aged women precedes a decline in male twin live births by three months. To our knowledge, no other studies have examined the relation between collective optimism and male twin live births.

Replication and extension of this work to other national contexts may hold interest among evolutionary theorists, fertility scholars and epidemiologists. The external validity of findings from tests of collective optimism in Sweden to other populations remains unexplored. Sweden differs markedly from the US in terms of healthcare systems, social safety nets, racial/ethnic diversity and income inequality (Shea et al., Reference Shea, Davey, Femia, Zarit, Sundström, Berg and Smyer2003; Starfield, Reference Starfield2000). We examine whether and to what extent sudden variations in collective optimism, as indicated by monthly changes in overall suicides, precede changes in the ratio of male twin to male singleton live births in the US using publicly available, nationally aggregated monthly data from 2003 to 2019. Consistent with prior work, we also test this relation with suicides among women aged 15−49 years (Catalano et al., Reference Catalano, Goldman-Mellor, Karasek, Gemmill, Casey, Elser, Bruckner and Hartig2020; Karasek et al., Reference Karasek, Goodman, Gemmill, Falconi, Hartig, Magganas and Catalano2015).

Results

Table 1 presents the descriptive statistics of our analytic data. The 204 months in our time-series analysis yielded a total of 1,129,712 male twins and 33,850,336 male singleton births. Total number of overall suicides over our study period equaled 670,893, with 73,826 suicides among females aged 15−49 years. There were approximately 5538 male twin births per month (SD = 327.08) and male twin ratio per month averaged 0.0334 (SD = 0.001). Percent monthly change in overall suicides averaged 0.0045 (SD = 0.071) and among females aged 15−49 years, averaged 0.0075 (SD = 0.107).

Table 1. Counts and distribution of male twin live births, male singleton live births, male twin ratio, overall suicides and suicides among females aged 15 to 49 years in the US, from 2003 to 2019

Figure 1 plots the monthly male twin ratio from January 2003 to December 2019. We observe a slight increase in this ratio starting in 2009 that persists through 2019 (Figure 1). Figure 2 presents the percent monthly change in overall suicides. As expected from our transformation of monthly suicide counts into percent monthly change, this series centers around zero but shows seasonal peaks and troughs (Figure 2). The plot of percent monthly change in female suicides (aged 15−49 years) appears in Supplementary Figure 1.

Figure 1. Plot of monthly male twin ratio (male twin live births/male singleton live births) in the US, from January 2003 to December 2019.

Figure 2. Plot of percent monthly change in overall suicides in the US, from January 2003 to December 2019.

Box-Jenkins routines identified autoregression (i.e., long-term retention or ‘echoes’ of observations) in the monthly series of male twin ratios at lags 1 and 12 months (AR 1, 12). Figure 3a plots the residuals of monthly male twin ratios after removal of AR 1, 12 from the original series and represents the counterfactual. Application of Box-Jenkins ARIMA routines to percent monthly change in overall suicides detected two moving average parameters at lags 1 and 12, indicating high or low values of residual errors being remembered at 1 and 12 months later. We also detected an integration parameter that required differencing of the percent monthly change in overall suicides series over the previous 12 months (I 12). Figure 3b plots the de-trended series of percent monthly change in suicides after removal of detected autocorrelation parameters (MA 1, 12; I 12). This series exhibits no temporal pattern and serves as the exposure or independent variable of our test (Figure 3b).

Figure 3. Residual series (after removal of autocorrelation) of {1) monthly male twin ratios (Figure 3a) and (2) percent monthly change in overall suicides (Figure 3b), from January 2003 to December 2019, USA. Initial 12 observations lost to autocorrelation parameter modeling.

Table 2 shows the estimated coefficients for our time-series test model (corresponding to analytic step 3). A 1% increase in overall suicides corresponds with a 0.005 unit decline in male twin ratio 6 months later (p = .004). Put another way, a 1% increase in overall suicides in a month (relative to the previous month) precedes a decline in 5 male twin live births per 1000 male singleton live births 6 months later. This translates to a 15% decline in the average incidence of male twin births per 1000 male singleton live births (based on 33.4 average monthly male twin births per 1000 male singleton births in our sample; 5/33.4 = 15). Examination of residuals indicates absence of autocorrelation (Supplementary Figure 2). Sensitivity tests support our original inference in that log-transformed series of monthly male twin ratio exhibits similar relation to de-trended residuals of percent monthly change in overall suicides at exposure lag of 6 months (coefficient = −0.137, p = .004; Table 3). As a robustness check and to maintain comparability with prior work, we repeated our main test (from Table 2) with de-trended residuals of percent monthly change in female suicides (aged 15−49 years) as our exposure (Supplementary Figure 3). Results from this robustness check support our main results in terms of exposure lag and statistically detectable decline in male twin ratio (Supplementary Table 1). However, the relation between the outcome and exposure at lag 6 appears attenuated in magnitude (coefficient = −.0012, p = .014), suggesting a stronger relation between male twinning and changes in overall suicides, relative to male twinning and suicides among reproductive-aged females in the US (Supplementary Table 1).

Table 2. Time-Series results for monthly male twin ratios from January 2003 to December 2019, as a function of exposure to de-trended residuals of percent monthly change in overall suicides and autocorrelation parameters

Note: *p < .05; two-sided test, **p < .01; two-sided test, ***p < .001; two-sided test.

Table 3. Time-series results for log-transformed monthly male twin ratios from January 2003 to December 2019, as a function of exposure to de-trended residuals of percent monthly change in overall suicides and autocorrelation parameters

Note: *p < .05; two-sided test, **p < .01; two-sided test, ***p < .001; two-sided test.