Linking Widowhood and Later-Life Depressive Symptoms: Do Childhood Socioeconomic Circumstances Matter?

Abstract

Widowhood and adverse childhood socioeconomic circumstances (CSC) have both been linked to increased levels of depressive symptoms in old age. Beyond their independent impact on depressive symptoms, experiencing adverse CSC may also trigger a cascade of cumulative adversity and secondary stressors across the life course that shapes how individuals weather stressful life events later on. We examine whether exposure to adverse CSC moderates the relationship between later-life widowhood and depressive symptoms. Using data from the Survey of Health, Ageing and Retirement in Europe (2004–2017), results revealed that both widowhood and adverse CSC were associated with increased levels of depressive symptoms among men and women. Associations between widowhood and depressive symptoms, however, were not moderated by CSC for both genders. We conclude that persisting differences in the levels of mental health-related to later-life widowhood did not further widen in the presence of disparities experienced early in the life course. This may reflect the life-altering impact of this age-normative, yet stressful life event across the social strata.

A large body of research documented the impact of widowhood on depressive symptoms in old age (Bennett & Soulsby, 2012; Carr & Utz, 2020; Das, 2013; King et al., 2019; Manning & Brown, 2011). This is particularly relevant because depression is a leading cause of disability worldwide (Murray et al., 2010) and rates are peaking among older adults (Falkingham et al., 2019; Fiske et al., 2009). Older adults’ exposure to adverse childhood socioeconomic circumstances (CSC) has also shown to be a prominent factor contributing to social disparities in later-life mental health and related outcomes (Angelini et al.; 2019; Ferraro et al., 2016; Torres & Wong, 2013; von Arx et al., 2019) because stressors encountered in early life can leave irreversible effects that magnify over time through an increasing cascade of subsequent risks (e.g., Carr, 2019; Dannefer, 2020; O’Rand, 2016). Although individuals’ ability to weather critical transitions—age-normative or not—also depends on their experiences with prior stressors (Elder & Shanahan, 2006; Umberson & Thomeer, 2020), it is less whether the exposure to adverse CSC accentuates the impact of later-life marital loss on depressive symptoms.

To fill this research gap, we draw from different theoretical lenses (e.g., the life course framework, cumulative disadvantage model, and stress process frameworks; e.g., Dannefer, 2020; Elder & Shanahan, 2006; Pearlin, 2010) that consider childhood as a sensitive period during which significant economic adversity may launch chains of disadvantage and secondary stressors throughout the life course. This could, in turn, have cumulative effects on individuals’ mental health over time, which may further aggravate response to age-normative, yet stressful life events in later life. Thus, we expect that both the exposure to adverse CSC and experiencing widowhood are associated with increased levels of later-life depressive symptoms. Moreover, we test whether exposure to adverse CSC amplifies marital loss-depressive symptoms linkages in later life.

Widowhood and Later-Life Depression

Even though marital strain can also undermine mental health as well (Carr & Utz, 2020), the death of a spouse is one of the most stressful life events (Holmes & Rahe, 1967). It represents an impactful turning point in one’s biography that goes along with the loss of companionship and intimacy, as well as increased levels of intense grief and reactivity to daily stressors (Das, 2013; Hahn et al., 2013; Utz et al., 2012). In addition to the emotional stress associated with widowhood, it can trigger a cascade of secondary stressors (Pearlin, 2010), such as eroding social networks, declining instrumental support, or shrinking financial assets (e.g., Manning & Brown, 2011; Perrig-Chiello et al., 2016), that can increase older adults’ vulnerability to mental health problems. Thus, prior research linked widowhood to a range of negative outcomes, such as poorer physical health, lower life satisfaction, increased levels of loneliness (e.g., Bennett & Soulsby, 2012; Carr & Utz, 2020; Infurna et al., 2016; Zhang et al., 2016) and—particularly in old age—to a heightened risk of experiencing depressive symptoms (e.g., Hewitt et al., 2012; King et al., 2019; Lin et al., 2019).

This is a particularly relevant public health issue because the prognosis of recovering from later-life depression is poor and it may lead to severe declines in one’s physical, cognitive, and social functioning (Fiske et al., 2009; Murray et al., 2010). Women may further be more affected because they are especially prone to suffer from later-life depression (Fiske et al., 2009) and are also twice as likely to be widowed by age 65 compared to men based on data from the U.S. (Carr & Utz, 2020). In contrast, being married in old age is associated with a lower risk of depression for both men and women compared to widowed (e.g., King et al., 2019; Sasson & Umberson, 2014) and never-married individuals (Yan et al., 2011) because of social selection into marriage (Umberson & Thomeer, 2020) and because marriage can buffer risks of depression (e.g., through emotional and instrumental support from the spouse, a joint accumulation of financial assets, and social control of health behaviors; LaPierre, 2009).

In addition to later-life mental health disparities by marital status, the benefits of marriage and the risks associated with widowhood may further vary by factors related to social class (Umberson & Thomeer, 2020). Social selection along key sociodemographic characteristics, such as a lower socioeconomic status (SES) across the life course and experiencing poverty in old age, has shown to play a role in the likelihood of experiencing (Ervin et al., 2021; Kung, 2020) and subsequently coping with widowhood (e.g., Recksiedler & Stawski, 2019). For example, Kung (2020) showed that wealth was a protective factor in the adaptation to widowhood among men in Australia. Mental health gaps between those who experienced widowhood and those who did not could, in turn, increase over time due to cumulative inequalities in the distribution of and access to economic resources and social capital fueled by initial group differences (Dannefer, 2020).

Theoretical Perspectives on CSC

Prior research has documented the pivotal association between CSC and mental health in later life (Angelini et al., 2019; Falkingham et al., 2019; Kamiya et al., 2013; Tani et al., 2016; Torres & Wong, 2013). CSC can undermine later-life mental health by, in part, compromising one’s economic well-being, career and family trajectories, health behaviors, and financial security in adulthood (Angelini & Mierau, 2018; Carr, 2019; Fiske et al., 2009; Umberson & Thomeer, 2020; von Arx et al., 2019). The life course perspective, which highlights the link between socially-stratified exposures to stressors and their long-term ripple effects (e.g., Elder & Shanahan, 2006), cumulative disadvantage models (e.g., Dannefer, 2020), as well as stress process frameworks (e.g., Pearlin, 2010) are useful theoretical lenses to understand how these processes unfold over time. More specifically, CSC contribute to adverse later-life mental health through two major pathways.

First, the timing of one’s exposure to economic adversity is linked to the severity of its impact. Experiencing economic adversity during a formative and sensitive period, such as childhood, can in itself have direct, lasting, and potentially irreversible harmful effects on individuals’ subsequent life course (O’Rand, 2016) because they become deeply ingrained in children’s physical, psychological, and social development. For example, Shonkoff et al. (2012) showed that experiencing childhood poverty launches a lifelong path of increased psychological and physiological reactivity to stress, which can magnify one’s risk of (mental) illness later-on. Second, exposure to adverse CSC can trigger a lifelong chain of cumulative disadvantage or adversity (Dannefer, 2020; O’Rand, 2016). This means that the relationship between CSC and mental health becomes magnified over time because exposure to economic adversity early-on leads to a series of gradual risk factors and other secondary stressors that unfold subsequently (Pearlin, 2010). For example, individuals who experienced adverse CSC were more likely to forgo higher education among a sample of Mexican older adults (Torres & Wong, 2013). This can, in turn, constrain their socioeconomic upward mobility in mid-life (Ferraro et al., 2016; Tani et al., 2016), and subsequently affect later-life mental health (e.g., Angelini et al., 2019 using SHARE).

Because an early launch of cumulative economic adversity can magnify over time and contribute to later-life mental health disparities, it may also influence how older adults weather the adjustment to stressful life events in old age (Umberson & Thomeer, 2020). This could be the case because older adults with exposure to adverse CSC may have fewer economic or social resources to cope with the loss of their spouse as a result of accumulated adversity and stress proliferation compared to individuals growing up in more prosperous CSC. Relatedly, van den Berg et al. (2010) showed that a Dutch sample of older adults experienced a stronger cognitive decline following the death of a family member if they grew up in adverse early-life conditions. It could also be the case that older adults’ exposure to adverse CSC had a direct and lasting effect on their (mental) health later-on, which was subsequently linked to a higher risk to experience complicated grief and depression in a U.S. sample of older adults (Utz et al., 2012). Thus, the psychosocial adjustment to widowhood could be mitigated either because of direct effects or through the cumulative burden of CSC, or both.

Hypotheses

Against this backdrop, we anticipate that widowhood is related to increased levels of depressive symptoms in old age compared to men and women who were continuously married or single (Hypothesis 1). We further expect that the exposure to adverse CSC is linked to increased levels of depressive symptoms in old age compared to men and women who grew up in more prosperous CSC (Hypothesis 2). Lastly, we expect CSC to moderate the relationship between later-life widowhood and depressive symptoms in a way that the exposure to adverse CSC aggravates widowhood-depressive symptoms linkages for men and women (Hypothesis 3).

Method

Data

We used data from the Survey of Health, Ageing and Retirement in Europe (SHARE; Börsch-Supan et al., 2013), which is a cross-national panel study with biennially repeated measurements of older adults in 28 European countries and Israel. For our study, we pooled data from seven waves (2004–2017). SHARE further collects information on respondents’ childhood conditions, which were first assessed in the retrospective SHARELIFE module in Wave 3 and repeated in Wave 7 for respondents who did not complete it in, or were included after, Wave 3. We included participants aged between 50 and 96 years if they fulfilled two main criteria: they completed the SHARELIFE and had at least one measurement of depressive symptoms. The final analytic sample consisted of 45,135 respondents (136,158 observations; 57.9% women) from 21 countries (see online appendix for a flowchart of the sample construction).

Dependent Variable

Depressive Symptoms.

We used the validated Euro-Depression 12-item scale as an indicator of depressive symptoms (Prince et al., 1999). Participants are asked to rate whether they experienced a range of different emotions during the past month (e.g., pessimism or fatigue; 0 = no; 1 = yes). Answers were collapsed into a composite score, which served as a continuous measure ranging from 0 to 12.

Independent Variables

Marital Transitions.

Self-reported marital status at each wave was used to identify respondents who had experienced widowhood at some point during the observation period. More specifically, respondents were categorized as became widowed if they reported being married and living together with a spouse, or in a registered partnership, at the beginning of the study, and subsequently experienced marital loss through the death of the spouse at some point during follow-up. To contrast these respondents with those who reported stability in their marital status during the observation period, we distinguish three additional groups. First, we classified respondents as continuously widowed if they reported being widowed throughout the observation period. Second, respondents were categorized as continuously coupled if they reported being either married and living with a spouse, or in a registered partnership, throughout the observation period. Third, respondents were classified as continuously single if they reported having been either never married or divorced throughout the observation period. We also grouped individuals who reported being married and living separated from a spouse throughout the observation period into this category because health and well-being benefits associated with marriage are less likely to apply to estranged (or non-residential) spouses compared to co-residential couples sharing daily routines. Note that those who transitioned into a relationship, became divorced, or reported several transitions during follow-up were excluded from the analyses due to low case numbers (less than 1%).

CSC.

In line with prior research by Wahrendorf and Blane (2015), we created a multidimensional index of CSC that covered four different domains of socioeconomic circumstances at Age 10. Indicators for each domain have shown to be relevant for operationalizing the long-term effects of CSC on mental and physical health (see appendix for more information).

First, respondents reported the occupation of the main breadwinner of their parental household, which was grouped into 10 main occupational groups based on the International Standard Classification of Occupations (ISCO). These occupational groups were categorized into the four ISCO skill-levels representing the complexity and range of tasks and duties related to occupations and then dichotomized into either low (1) or high (0) occupational positions. Second, respondents were asked to report the number of books in their parental home. Having had ten books or less compared to more than ten books indicated socioeconomic disadvantage (0 = not disadvantaged; 1 = disadvantaged). Third, household overcrowding was computed by forming a ratio between the number of people and the number of rooms per household. We classified households as overcrowded if there was more than one person per room (0 = not overcrowded; 1 = overcrowded). Lastly, housing quality was assessed by the presence of the following facilities: fixed bathtub, cold and hot water supply, inside toilet, and central heating. If the household had none of them, it was coded as disadvantaged (1), and as not disadvantaged (0), if at least one facility was present. These binary responses were combined into a count variable ranging from 0 (most advantaged) to 4 (most disadvantaged).

Controls.

Adulthood SES is likely to be associated with both CSC and levels of later-life depressive symptoms. We therefore used three domains of adulthood SES as our main control variables to examine persisting effects of CSC in later-life above and beyond the more proximal impact of adulthood measures. First, we used the UNESCO’s International Standard Classification of Education to group participants into primary, secondary, or tertiary levels of education. Second, the skill level of participants’ main occupational position during mid-life formed in parallel to the occupational skill level of the main breadwinner during childhood (0 = low; 1 = high). Additionally, participants who were never gainfully employed were assigned into a separate category. Lastly, respondents were asked (each wave) if they are able to make ends meet based on their total monthly household income on a scale from 1 (with great difficulty) to 4 (easily). We used the mode over all available waves per individual to compute a time-invariant variable to preserve cases with some missing observations.

We also adjusted the models for attrition across waves (no drop-out, drop-out, or death during follow-up), birth cohort (1919–1928 to after 1945), and migrant status (0 = not native-born; 1 = native-born) to control for potential selection biases concerning the likelihood of remaining in the study throughout the observation period.

Time-varying and -invariant risk factors commonly associated with depressive symptoms and the adjustment to stressful life events in old age (Carr & Utz, 2020; Fiske et al., 2009) further served as control variables (see appendix for more information). These were respondents’ level of physical activity (0 = low, 1 = high), whether respondents had two or more chronic diseases (0 = no; 1 = yes) and one or more limitation with (instrumental) activities of daily living ([I]ADL; 0 = no; 1 = yes), as well as the number of living children (time-invariant; count) as a proxy for social support.

Analytic Approach

We first examined the sociodemographic composition of our sample, as well as the distribution of marital transitions, by gender and CSC descriptively. Then, we used mixed-effects models to account for the nested structure of the data. Here, the mixed models encompass two random factors: repeated observations nested within individuals and individuals nested within countries. Mixed-effects models do not require an equal number of observations from all participants and can include participants with missing observations (see Table 1 and 2 for the distribution of attrited cases). Logistic regression models predicting attrition further showed that only respondents from older birth cohorts were more likely to drop out or be deceased at some point during follow-up compared to those born after 1945 (results not shown). The latter was also true for male compared to female respondents.

Table 1.

Participants’ Characteristics by Gender at Baseline (N = 45,135).

	Women N = 26,114 (57.9%)		Men N = 19,021 (42.1%)
	M	SD	M	SD
Outcome
Depressive symptoms	2.79	2.34	1.85	1.92
Predictors
Age in years	63.08	9.49	62.99	9.02
Number of children alive	2.17	1.33	2.14	1.35
	N	%	N	%
Marital transitions
Continuously coupled	14,451	55.3	14,348	75.4
Continuously single	4,313	16.5	2,906	15.3
Continuously widowed	5,310	20.3	1,052	5.5
Became widowed	2,040	7.8	715	3.8
Childhood socioeconomic circumstances
Most disadvantaged	4,103	15.7	2,925	15.4
Disadvantaged	6,490	24.9	4,590	24.1
Medium	8,596	32.9	6,204	32.6
Advantaged	5,244	20.1	4,025	21.2
Most advantaged	1,681	6.4	1,277	6.7
Level of education
Primary	6,932	26.5	4,044	21.3
Secondary	14,128	54.1	10,274	54.0
Tertiary	5,054	19.4	4,703	24.7
Main occupation in mid-life
Low-skill job	16,086	61.6	12,193	64.1
High-skill job	6,978	26.7	6,609	34.7
Never worked	3,050	11.7	219	1.2
Making ends meet in later-life
With great difficulty	3,241	12.4	1,827	9.6
With some difficulty	5,835	22.3	3,645	19.2
Fairly easily	7,854	30.1	5,682	29.9
Easily	9,184	35.2	7,867	41.4
Birth cohorts
1919–1928	1,723	6.6	1,064	5.6
1929–1938	5,033	19.3	3,715	19.5
1939–1945	5,294	20.3	4,147	21.8
1945 and later	14,064	53.9	10,095	53.1
Physical activity
Low	8,228	31.5	5,087	26.7
High	17,886	68.5	13,934	73.3
Having disabilities
ADL	1,858	7.1	985	5.2
IADL	3,704	14.2	1,357	7.1
Number of chronic conditions
Less than 2	14,152	54.2	11,595	61.0
2 or more	11,962	45.8	7,426	39.0
Attrition
No	21,711	83.1	15,199	79.9
Dropout	3,149	12.1	2,562	13.5
Death	1,254	4.8	1,260	6.6
Migrant	2,390	9.2	1,659	8.7

Notes. (I)ADL = Limitations with (instrumental) daily activities.

Table 2.

Participants’ Characteristics by Childhood Socioeconomic Circumstances at Baseline.

	CSC (N = 45,135)
	Most advantaged N = 2,958 (6.6%)	Advantaged N = 9,269 (20.5%)	Middle N = 14,800 (32.8%)	Disadvantaged N = 11,080 (24.5%)	Most Disadvantaged N = 7,028 (15.6%)
	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Outcome
Depressive symptoms	2.03 (1.95)	2.09 (2.04)	2.23 (2.12)	2.61 (2.34)	2.93 (2.44)
Predictors
Age in years	60.98 (8.95)	61.03 (8.77)	61.66 (8.87)	64.34 (9.31)	67.4 (9.22)
Number of children alive	2.04 (1.29)	2.04 (1.23)	2.10 (1.24)	2.22 (1.40)	2.38 (1.54)
	N (%)	N (%)	N (%)	N (%)	N (%)
Marital transitions
Continuously coupled	1,947 (65.8)	6,172 (66.6)	9,698 (65.5)	6,848 (61.8)	4,134 (58.8)
Continuously single	625 (21.1)	1,766 (19.1)	2,566 (17.3)	1,572 (14.2)	690 (9.8)
Continuously widowed	270 (9.1)	896 (9.7)	1,755 (11.9)	1,888 (17.0)	1,553 (22.1)
Became widowed	116 (3.9)	435 (4.7)	781 (5.3)	772 (7.0)	651 (9.3)
Male	1,277 (43.2)	4,025 (43.4)	6,204 (41.9)	4,590 (41.4)	2,925 (41.6)
Level of education
Primary	81 (2.7)	816 (8.8)	2,432 (16.4)	3,833 (34.6)	3,814 (54.3)
Secondary	1,166 (39.4)	5,064 (54.6)	9,193 (62.1)	6,148 (55.5)	2,831 (40.3)
Tertiary	1,711 (57.8)	3,389 (36.6)	3,175 (21.5)	1,099 (9.9)	383 (5.5)
Main occupation in mid-life
Low-skill job	1,021 (34.5)	4,659 (50.3)	9,462 (63.9)	7,931 (71.6)	5,206 (74.1)
High-skill job	1,846 (62.4)	4,267 (46.0)	4,629 (31.3)	2,045 (18.5)	800 (11.4)
Never worked	91 (3.1)	343 (3.7)	709 (4.8)	1,104 (10.0)	1,022 (14.5)
Making ends meet in later-life
With great difficulty	78 (2.6)	537 (5.8)	1,256 (8.5)	1,729 (15.6)	1,468 (20.9)
With some difficulty	312 (10.6)	1,301 (14.0)	2,948 (19.9)	2,752 (24.8)	2,167 (30.8)
Fairly easily	755 (25.5)	2,656 (28.7)	4,544 (30.7)	3,499 (31.6)	2,082 (29.6)
Easily	1,813 (61.3)	4,775 (51.5)	6,052 (40.9)	3,100 (28.0)	1,311 (18.7)
Birth cohorts
1919–1928	141 (4.8)	375 (4.1)	665 (4.5)	802 (7.2)	804 (11.4)
1929–1938	398 (13.5)	1,249 (13.5)	2,337 (15.8)	2,501 (22.6)	2,263 (32.2)
1939–1945	519 (17.6)	1,770 (19.1)	2,908 (19.7)	2,553 (23.0)	1,691 (24.1)
1945 and later	1,900 (64.2)	5,875 (63.4)	8,890 (60.1)	5,224 (47.2)	2,270 (32.3)
Physical activity
Low	655 (22.1)	2,255 (24.3)	3,956 (26.7)	3,656 (33.0)	2,793 (39.7)
High	2,303 (77.9)	7,014 (75.7)	10,844 (73.3)	7,424 (67.0)	4,235 (60.3)
Having disabilities
ADL	129 (4.4)	388 (4.2)	773 (5.2)	803 (7.3)	750 (10.7)
IADL	216 (7.3)	689 (7.4)	1,309 (8.9)	1,467 (13.2)	1,380 (19.6)
Number of chronic conditions
Less than 2	1,980 (66.9)	5,855 (63.2)	8,895 (60.1)	5,850 (52.8)	3,167 (45.1)
2 or more	978 (33.1)	3,414 (36.8)	5,905 (39.9)	5,230 (47.2)	3,861 (54.9)
Attrition
No	2,485 (84.0)	7,738 (83.5)	12,103 (81.8)	8,996 (81.2)	5,588 (79.5)
Dropout	371 (12.5)	1,224 (13.2)	2,056 (13.9)	1,376 (12.4)	684 (9.7)
Death	102 (3.5)	307 (3.3)	641 (4.3)	708 (6.4)	756 (10.8)
Migrant	369 (12.5)	850 (9.2)	1,226 (8.3)	934 (8.4)	670 (9.5)

Notes. (I)ADL = Limitations with (instrumental) daily activities.

All models were stratified by gender because previous studies showed that women were at greater risk for both depression and marital loss compared to men (e.g., Carr & Utz, 2020; Fiske et al., 2009). Comparing exploratory models with and without interactions between gender and marital transitions further supported our decision that gender stratification in models is warranted (p < 0.0001 from an analysis of variance comparing models without and with interaction terms; results available upon request). Analyses were performed with R (using the lme4 and lmerTest packages) and were run in the following order: Model 1 tested the main associations between becoming widowed and levels of depressive symptoms (Hypothesis 1), as well as between CSC and levels of depressive symptoms (Hypothesis 2). In Model 2, interaction terms between CSC and marital transitions were added to examine whether CSC moderates the link between becoming widowed and depressive symptoms (Hypothesis 3).

To further probe whether the sociodemographic sample composition (i.e., attrition, birth cohort, and migration status), risk factors for later-life depressive symptoms, and adulthood SES may confound or suppress CSC-related effects, we entered these controls at different steps of the analyses. First, we estimated models without any controls (results available upon request), which largely yielded the same patterns of results as models controlling for the sociodemographic sample composition and risk factors for later-life depressive symptoms. Second, we further added adulthood SES to the models including the main effects of marital transitions and CSC on depressive symptoms (Model 3), as well as interactions between CSC and marital transitions (Model 4). We further tested and assessed these nested models based on the Bayesian information criterion (BIC) and likelihood-ratio tests (LRT).

Results

Descriptive Results

Table 1 shows summary statistics of participants’ characteristics at baseline by gender. The majority of the sample, and particularly men, were continuously coupled (about 75% of men vs. 55% of women). A sizable share of women was continuously widowed (about 20% vs. 6% of men) and only a smaller faction, but almost twice as many women, reported to have become widowed throughout the observation period (about 8% of women vs. 4% of men). Moreover, about 40% of men and women reported having grown up in either the most disadvantaged or disadvantaged CSC.

To further probe the CSC-related differences in the sociodemographic composition and the distribution of marital transitions of the sample descriptively, Table 2 shows participants’ characteristics at baseline by CSC. Respondents in the most disadvantaged CSC group (about 16% of the sample) tended to be older and to have a larger number of living children compared to respondents that grew up under more advantaged CSC. A larger share of respondents in the most disadvantaged and disadvantaged CSC categories reported lower levels of adulthood SES (e.g., primary levels of schooling only and great difficulties to make ends meet), as well as poorer health outcomes in later life (e.g., low physical activity and having two or more chronic conditions) compared to those in more advantaged CSC. Concerning marital transitions, the share of those who were continuously coupled was lowest, and highest for those who were either continuously widowed or became widowed, among respondents with the most disadvantaged CSC (e.g., about 9% became widowed vs. 4% among those with most advantaged CSC). Interestingly, only the share of those who were continuously single was lower for respondents with more disadvantaged and disadvantaged CSC compared to those with more advantaged CSC.

Multivariate Regression Results

For both men and women, models comparisons using the BIC (see Table 3 for women and Table 4 for men) and LRT showed that Model 4 fit the data best compared to Models 1 and 2 (p < 0.001), while no significant differences were observed between Models 3 and 4 (p = 0.929 for women and p = 0.767 for men).

Table 3.

Mixed-Effects Regression Results Predicting Depressive Symptoms among Women

	Model 1	Model 2	Model 3	Model 4
	Coef (95% CI)	Coef (95% CI)	Coef (95% CI)	Coef (95% CI)
Age, 10 years	0.30 (0.26–0.34)***	0.30 (0.26–0.34)***	0.31 (0.27–0.36)***	0.31 (0.27–0.36)***
Age, 10 years, squared	0.19 (0.17–0.21)***	0.19 (0.17–0.21)***	0.19 (0.17–0.21)***	0.19 (0.17–0.21)***
Marital transitions
Continuously single a	0.25 (0.19–0.31)***	0.19 (−0.01–0.38)	0.08 (0.02–0.14)**	−0.00 (−0.20–0.19)
Continuously widowed a	0.20 (0.14–0.26)***	0.22 (0.09–0.34)**	0.09 (0.03–0.15)**	0.10 (−0.03–0.22)
Became widowed a	0.51 (0.43–0.59)***	0.51 (0.35–0.68)***	0.43 (0.35–0.50)***	0.40 (0.24–0.56)***
Childhood socioeconomic circumstances (CSC)
Disadvantaged b	−0.15 (−0.21−−0.08)***	−0.16 (−0.25−−0.06)**	−0.08 (−0.15−−0.01)*	−0.10 (−0.19−−0.01)*
Medium b	−0.32 (−0.39−−0.25)***	−0.32 (−0.42−−0.23)***	−0.17 (−0.24−−0.10)***	−0.18 (−0.27−−0.08)***
Advantaged b	−0.41 (−0.49−−0.33)***	−0.40 (−0.50−−0.30)***	−0.21 (−0.29−−0.13)***	−0.21 (−0.32−−0.11)***
Most advantaged b	−0.45 (−0.55−−0.34)***	−0.42 (−0.56−−0.29)***	−0.18 (−0.28−−0.07)***	−0.17 (−0.31−−0.04)*
Educational attainment
Secondary c			−0.15 (−0.21−−0.10)***	−0.15 (−0.21−−0.10)***
Tertiary c			−0.22 (−0.30−−0.14)***	−0.22 (−0.30−−0.14)***
Main occupation in mid-life
High-skill job d			−0.06 (−0.11−−0.00)*	−0.06 (−0.11−−0.00)*
Never worked d			−0.03 (−0.10–0.04)	−0.03 (−0.10–0.04)
Making ends meet in later-life
Easily e			−1.22 (−1.30−−1.14)***	−1.22 (−1.29−−1.14)***
Fairly easily e			−1.05 (−1.12−−0.97)***	−1.05 (−1.12−−0.97)***
With some difficulty e			−0.69 (−0.76−−0.61)***	−0.68 (−0.76−−0.61)***
Marital transitions x CSC
Continuously single a x Disadvantaged b		0.15 (−0.08–0.38)		0.17 (−0.06–0.39)
Continuously widowed a x Disadvantaged b		−0.05 (−0.21–0.11)		−0.02 (−0.18–0.13)
Became widowed a x Disadvantaged b		0.06 (−0.15–0.28)		0.09 (−0.12–0.31)
Continuously single a x Medium b		0.04 (−0.18–0.25)		0.06 (−0.15–0.27)
Continuously widowed a x Medium b		0.03 (−0.13–0.19)		0.01 (−0.14–0.17)
Became widowed a x Medium b		−0.02 (−0.23–0.20)		0.02 (−0.19–0.23)
Continuously single a x Advantaged b		0.04 (−0.19–0.27)		0.06 (−0.16–0.28)
Continuously widowed a x Advantaged b		−0.02 (−0.21–0.16)		−0.02 (−0.20–0.16)
Became widowed a x Advantaged b		−0.07 (−0.31–0.18)		0.00 (−0.24–0.24)
Continuously single a x Most advantaged b		0.07 (−0.21–0.34)		0.12 (−0.16–0.39)
Continuously widowed a x Most advantaged b		−0.14 (−0.42–0.14)		−0.10 (−0.38–0.17)
Became widowed a x Most advantaged b		−0.16 (−0.56–0.24)		−0.10 (−0.49–0.29)
BIC	330,527	330,655	329,484	329,613

Notes. Reference categories are:

^acontinuously coupled.

^bmost disadvantaged.

^cprimary.

^dlow-skill job.

^ewith great difficulty.

BIC = Bayesian information criterion.

All models are adjusted for attrition, birth cohort, migrant status, physical activity, number of chronic conditions, disabilities ((I)ADL), and number of living children. Clustering on the country-level was taken into account in the random part of the model.

Age was entered centered at the midpoint of the sample’s age range (73 years) and then divided by 10. Age squared was also included in the model to better reflect non-linear distribution of depressive symptoms across respondents’ age.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Table 4.

Mixed-Effects Regression Results Predicting Depressive Symptoms among Men

	Model 1	Model 2	Model 3	Model 4
	Coef (95% CI)	Coef (95% CI)	Coef (95% CI)	Coef (95% CI)
Age, 10 years	0.48 (0.43–0.53)***	0.48 (0.43–0.53)***	0.49 (0.45–0.54)***	0.49 (0.45–0.54)***
Age, 10 years, squared	0.22 (0.20–0.24)***	0.22 (0.20–0.24)***	0.21 (0.19–0.23)***	0.21 (0.19–0.23)***
Marital transitions
Continuously single a	0.36 (0.30–0.42)***	0.25 (0.08–0.42)**	0.27 (0.21–0.33)***	0.19 (0.03–0.36)*
Continuously widowed a	0.26 (0.16–0.36)***	0.19 (−0.02–0.40)	0.26 (0.16–0.35)***	0.21 (−0.00–0.42)
Became widowed a	0.38 (0.27–0.48)***	0.40 (0.18–0.63)**	0.34 (0.24–0.45)***	0.37 (0.14–0.59)**
Childhood socioeconomic circumstances (CSC)
Disadvantaged b	−0.02 (−0.09–0.05)	−0.03 (−0.11–0.05)	0.02 (−0.05–0.09)	0.01 (−0.06–0.09)
Medium b	−0.14 (−0.21−−0.07)***	−0.16 (−0.24−−0.09)***	−0.03 (−0.10–0.04)	−0.05 (−0.13–0.03)
Advantaged b	−0.21 (−0.28−−0.13)***	−0.23 (−0.31−−0.15)***	−0.06 (−0.14–0.02)	−0.08 (−0.16–0.01)
Most advantaged b	−0.20 (−0.30−−0.10)***	−0.20 (−0.31−−0.09)***	−0.02 (−0.12–0.09)	−0.01 (−0.13–0.10)
Educational attainment
Secondary c			−0.12 (−0.18−−0.07)***	−0.12 (−0.18−−0.06)***
Tertiary c			−0.18 (−0.26−−0.11)***	−0.18 (−0.26−−0.11)***
Main occupation in mid-life
High-skill job d			0.01 (−0.03–0.06)	0.01 (−0.04–0.06)
Never worked d			0.12 (−0.08–0.31)	0.12 (−0.08–0.31)
Making ends meet in later-life
Easily e			−1.10 (−1.19−−1.02)***	−1.10 (−1.19−−1.02)***
Fairly easily e			−0.98 (−1.06−−0.90)***	−0.98 (−1.06−−0.90)***
With some difficulty e			−0.71 (−0.79−−0.63)***	−0.71 (−0.79−−0.63)***
Marital transitions x CSC
Continuously single a x Disadvantaged b		0.04 (−0.17–0.24)		0.02 (−0.18–0.21)
Continuously widowed a x Disadvantaged b		0.17 (−0.10–0.45)		0.16 (−0.11–0.43)
Became widowed a x Disadvantaged b		−0.07 (−0.37–0.23)		−0.07 (−0.36–0.23)
Continuously single a x Medium b		0.15 (−0.04–0.34)		0.12 (−0.07–0.30)
Continuously widowed a x Medium b		0.10 (−0.17–0.38)		0.08 (−0.19–0.35)
Became widowed a x Medium b		−0.02 (−0.32–0.28)		−0.02 (−0.32–0.27)
Continuously single a x Advantaged b		0.17 (−0.03–0.37)		0.13 (−0.07–0.33)
Continuously widowed a x Advantaged b		−0.05 (−0.36–0.27)		−0.09 (−0.40–0.22)
Became widowed a x Advantaged b		0.00 (−0.35–0.35)		0.03 (−0.32–0.37)
Continuously single a x Most advantaged b		0.05 (−0.21–0.31)		0.04 (−0.22–0.30)
Continuously widowed a x Most advantaged b		−0.10 (−0.58–0.37)		−0.15 (−0.62–0.32)
Became widowed a x Most advantaged b		−0.13 (−0.66–0.40)		−0.12 (−0.64–0.40)
BIC	222,098	222,220	221,367	221,490

Notes. Reference categories are:

^acontinuously coupled.

^bmost disadvantaged.

^cprimary.

^dlow-skill job.

^ewith great difficulty.

BIC = Bayesian information criterion.

All models are adjusted for attrition, birth cohort, migrant status, physical activity, number of chronic conditions, disabilities ((I)ADL), and number of living children. Clustering on the country-level was taken into account in the random part of the model.

Age was entered centered at the midpoint of the sample’s age range (73 years) and then divided by 10. Age squared was also included in the model to better reflect non-linear distribution of depressive symptoms across respondents’ age.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Women

In line with our first hypothesis, Model 1 showed that, compared to women who were continuously coupled, women who were continuously widowed and those who became widowed at some point during follow-up had higher levels of depressive symptoms (Table 3 and Figure 1). Furthermore, the coefficient for those who became widowed was larger than the coefficient for continuously widowed respondents. Continuously single women also reported higher levels of depressive symptoms compared to women who were continuously coupled.

Figure 1.

Depressive symptoms by marital transitions among men and women.

Model 1 also tested the second hypothesis on linkages between CSC and levels of later-life depressive symptoms. All categories of CSC ranging from disadvantaged CSC to the most advantaged CSC were significantly associated with lower levels of depressive symptoms compared to respondents reporting with the most disadvantaged CSC (see Figure 2). The size of the coefficients (i.e., smaller coefficients for more advantaged CSC) seemed to indicate a social gradient across categories in a way that a more advantaged upbringing was related to lower levels of depressive symptoms, which supported this hypothesis. We then added interaction terms between marital transitions and CSC in Model 2 to test whether CSC moderates the link between widowhood and higher levels of depressive symptoms (Hypothesis 3). Because none of the interactions were significant, this hypothesis was not confirmed for women

Figure 2.

Depressive symptoms by childhood socioeconomic circumstances among men and women.

After adjusting the main effect model for adulthood SES (Model 3), the coefficients for marital transitions—even though less so for those who became widowed compared to continuously coupled women—were slightly smaller but remained significant. Only the previously observable social gradient across CSC categories was no longer visible, although the coefficients were still significant. Model 4, which included the marital transition by CSC interaction terms and controlled for adulthood SES, yielded no significant moderation effects.

Men

Similar to the results for women, men who were continuously widowed and those who became widowed during follow-up reported higher levels of depressive symptoms compared to continuously coupled men in Model 1 (Table 4 and Figure 1). Our first hypothesis was therefore also supported for men. Furthermore, continuously single men also showed higher levels of depressive symptoms compared to continuously coupled men, and the size of the coefficient for this group was comparable to the size of the coefficient for men who became widowed during follow-up.

CSC was also associated with depressive symptoms in Model 1. Three out of four categories of CSC (i.e., medium, advantaged, and most advantaged compared to most disadvantaged CSC) were significantly associated with depressive symptoms indicating that a more advantaged upbringing was associated with lower levels of depressive symptoms. Thus, Hypothesis 2 was partially supported for men. Interaction terms between marital transitions and CSC were again not significant in Model 2, which did not support our expectation that adverse CSC would aggravate the link between widowhood and depressive symptoms among men (Hypothesis 3).

However, when we added adulthood SES into the main effects model (Model 3), CSC was no longer significantly associated with depressive symptoms in contrast to the previous findings for women. Associations between marital transitions and depressive symptoms remained significant and the size of these coefficients was only slightly reduced. Model 4, which included interaction terms between marital transitions and CSC and controlled for adulthood SES, also yielded no significant moderation effects.

Discussion

Our study focused on social disparities in later-life depressive symptoms related to CSC and widowhood, which have independently been linked to increased levels of depression in old age (e.g., Angelini et al., 2019; Kamiya et al., 2013; von Arx et al., 2019; Yan et al., 2011). Yet we also examined whether the exposure to these experiences—one in earlier and one in later decades of the life course—accentuated levels of depressive symptoms in old age.

Becoming widowed in old age is an age-normative, yet critical and stressful transition that is linked to a range of detrimental mental and physical health outcomes (e.g., Bennett & Soulsby, 2012; Das, 2013; Infurna et al., 2017). Accordingly, our findings revealed elevated levels of depressive symptoms among both men and women who were continuously widowed and those who became widowed compared to continuously married respondents (e.g., Sasson & Umberson, 2014; Zhang et al., 2016). This is likely related to the emotional stress associated with bereavement, as well as the host of additional stressors triggered by the death of a spouse (e.g., loss of emotional and instrumental support or declining social networks; Manning & Brown, 2011). Nevertheless, characteristics of the marital relationship, such as prior relationship quality, and circumstances surrounding the loss, such as expectancy of the loss in case of terminal illness or long-term caregiving pressure, can mitigate or exacerbate the psychological toll associated with widowhood (e.g., Carr & Utz, 2020; Wong & Waite, 2016; Schaan, 2013). Yet examining these relational factors was beyond the scope of our study.

Prior studies have further shown that certain sociodemographic subgroups are particularly likely to experience widowhood, such as women (Carr & Utz, 2020) and low-SES individuals (Kung, 2020). However, it is unclear whether the distribution of those who experience widowhood also varies by older adults’ exposure to adverse CSC. Descriptively, we observed a clear CSC-related gradient in the share of respondents who became widowed during follow-up and those who were continuously widowed. This could be indicative of the presence of social causation and social selection processes (i.e., wealthier and healthier individuals are less prone to experience widowhood; Umberson & Thomeer, 2020), which may have been fueled by the early launch of chains of disadvantage and economic adversity across the life course (Dannefer, 2020; O’Rand, 2016). Furthermore, the foregone benefits associated with marriage (Carr & Utz; 2020; LaPierre, 2009) among widowed compared to continuously married individuals could magnify these group differences further.

Our models also showed that, in line with prior studies using other databases than SHARE (e.g., Kamiya et al., 2013; Tani et al., 2016), the exposure to adverse CSC was associated with higher levels of depressive symptoms for both genders in old age. This is likely the case because CSC can impact mental health directly and irreversibly through the deprivation of material resources during sensitive developmental periods (Carr, 2019), and indirectly through the early launch of chains of adversity and stress proliferation across the life course (O’Rand, 2016; Pearlin, 2020). For example, one could speculate that malnutrition in childhood hinders age-appropriate cognitive development and increase one’s reactivity to stress (e.g., Shonkoff et al., 2012), which may lead to less school involvement, subsequent risky behaviors, worse job prospects and a lower adulthood SES, and finally to higher levels of later-life depressive symptoms. However, after adjusting our models for adulthood SES, the link between CSC and depressive symptoms faded among men (cf. von Arx et al., 2019). This could suggest that, overall, men in their role as main economic breadwinners—at least for the cohorts of participants included in SHARE who grew up in earlier decades of the twentieth century—may have still been able to partly mitigate the impact of CSC by securing financial resources through the investment in educational attainment and labor market participation in adulthood compared to women (Ervin et al., 2021). It will therefore be crucial to re-examine gender dynamics in these linkages for future cohorts of older adults with starkly different rates of female enrollment in post-secondary education, labor market participation, and financial independence.

Based on prior research on CSC-related differences in older adults social, emotional, physical, and economic resources (e.g., Carr, 2019), which could mitigate their ability to cope with critical life events (van den Berg, 2010), we expected the link between widowhood and depressive symptoms to be amplified among older adults with exposure to adverse CSC. There was no evidence of such interactions contrary to our expectations. Although somewhat surprising, this finding is in line with research by Vandecasteele (2011; using cross-national European data) indicating that union dissolution among younger cohorts of respondents had a poverty-triggering effect across all educational levels and social classes compared to other family-life transitions (e.g., childbirth). Because marital loss is such a universally impactful turning point in one’s life that affects individuals across the social strata similarly (Das, 2013; van den Berg, 2010; Zhang et al., 2016), mental health disparities due to the exposure to adverse CSC may not have widened further. Our null finding could indicate a positive outlook on health promotion among older adults in the sense that individuals raised under more adverse socioeconomic conditions do not seem to accumulate an increased burden of depressive symptoms related to widowhood.

One possible explanation for this finding is that lower-SES individuals may be less vulnerable to life event-related declines in mental health because they are more likely to experience economic deprivation and related secondary stressors already before a critical life event, potentially mollifying the impact of such later-life transitions (Recksiedler & Stawski, 2019; Chen & Miller, 2012). Alternatively, severe emotional and mental health fluctuations associated with widowhood may be happening on a faster time-scale than the biennial assessment schedule of SHARE allows examining. As such, there may very well have been substantial differences in the dynamics of depressive symptoms proximal to the loss that do not result in durable shifts in mental health (Luhmann et al., 2012).

Limitations

Our study has several limitations. First, our analytic sample was restricted to respondents who had completed the retrospective SHARELIFE module in Wave 3 or 7. It is possible that our sample is biased toward healthier individuals who were able to participate in SHARE until at least Wave 3. If that were the case, our results may underestimate the impact of CSC on the link between widowhood and depressive symptoms because respondents in poor health, who also tend to be most likely to have been exposed to adverse CSC (Ferraro et al., 2016; Tani et al., 2016), would be underrepresented in the analyses. Nevertheless, this bias was limited by adjusting for attrition in our analyses, and by excluding participants who died or dropped out during the follow-up in a set of sensitivity analyses (see appendix).

Second, we focused on differences in levels of depressive symptoms associated with marital transitions and CSC. Previous research showed that depressive symptoms exhibit complex dynamics before, during, and after the transition (Infurna et al., 2017; Lin et al., 2019). We differentiated between respondents who became widowed during the observation period and those who were continuously widowed, which served as a rough proxy for the more proximal vs. more long-term effect of widowhood. As one would expect, we observed larger increases in the levels of depressive symptoms among those who became widowed compared to continuously widowed respondents—particularly among women.

Third and because our analyses were based on pooled cross-national data, the reported associations could vary across countries because of, for instance, the potential influence of cultural differences in the bereavement process or regional variation in the exposure to or influence of CSC. Even though cross-national comparisons were beyond the scope of our study, other recent studies using SHARE have examined variations across regional contexts both in the adaptation to widowhood (Schaan, 2013; Schmitz, 2021) and in associations between CSC and trajectories of self-rated health (Sieber et al., 2020).

Conclusion

Our study contributes to the literature on identifying proximal and long-term factors associated with depressive symptoms in later-life, which is particularly relevant and timely because depression is the leading cause of disability worldwide. Compared to prior studies focusing on single dimensions of SES, we used a comprehensive and multifaceted measure of CSC and different indicators of adulthood SES. We are further confident that our carefully conducted sensitivity analyses warrant the trustworthiness of the results and that the strength of our study (e.g., using large-scale, cross-national longitudinal data, validated study indicators, and analytical methods that go beyond cross-sectional group comparisons) outweighs its limitations. Exposures to widowhood and adverse CSC were both linked to depressive symptoms, which highlights the need for targeted intervention and prevention efforts combating the adverse effects of both factors for promoting later-life mental health. For example, implementing stronger security nets on the policy-level that disrupt CSC-related trajectories (e.g., in terms of early learning and school success; Shonkoff & Philipps., 2000), or facilitating access to social support and skill trainings that mitigate possible ripple effects of widowhood (i.e., loneliness; Fiske et al., 2009; Schmitz, 2021), could lower rates of later-life depressive symptoms.

Exposure to adverse CSC, however, did not aggravate older adults’ psychological distress in the presence of widowhood, which we expected due to scarring effects of adverse CSC during sensitive periods and processes of cumulative adversity across the life course. We therefore conclude that later-life widowhood is such an impactful and life-altering transition in the life course that it affected the mental health of individuals across the social strata similarly, despite prior processes of social differentiation related to CSC. Yet, this interesting null finding also leaves at least two promising avenues for future theory building and empirical studies. First, unpacking the dynamic interplay between CSC-related stressors and experiences across different life domains on the micro-level (Ferraro, 2011), such as marital history, workplace characteristics, or intergenerational ties, could further inform our understanding of whether and how exposures to CSC and widowhood jointly shape later-life mental health. Processes on the micro-level could further be modified by macro-level influences on CSC- and transition-related stressors across the life course (e.g., social institutions or socio-historical events; Sieber et al., 2020). Second, future studies should identify protective factors (e.g., social capital and psychological skills) that may offset or buffered stress-related responses to major life events, such as widowhood in the context of exposure to adverse CSC.