Widowhood and Depression in a Cross-National Perspective: Evidence from the United States, Europe, Korea, and China

Abstract

Objectives

We explore marital and depression trajectories over time for men and women, and distinguish between mood and somatic depression across contexts.

Method

We use longitudinal data from 2002 to 2013 from the United States, England, Europe, Korea, and China to explore depression among individuals married at baseline and follow their trajectories into widowhood with married as the reference group. We use random effects models to estimate these trajectories using the Center for Epidemiologic Studies Depression Scale (CES-D) or EURO-D scales for men and women.

Results

Depression peaks within the first year of widowhood for men and women, but women recover to levels comparable to married counterparts in all countries. Men sustain high levels of depression even 6–10 years post-widowhood everywhere except Europe. Widowed women have higher somatic depression compared to men, who have higher mood depression. Family plays differential roles in mediating depression across countries.

Discussion

Our research shows the complex global relationship between widowhood and depression. Studies that do not compare depression trajectories over time may make incorrect inferences about the persistence of depression by gender and country. Interventions should target different components of depression: mood-related symptoms for men and somatic-related symptoms for women for most effective recovery.

Introduction

Researchers in several disciplines—particularly psychiatry and gerontology—have focused on the links between widowhood and depression. Demographers have elucidated the role of gender in differential experiences of widowhood after spousal loss and well-being (Sasson & Umberson, 2014). Besides gender differences, mechanisms to explain the link between widowhood and depression indicate the importance of duration of widowhood (Perrig-Chiello, Spahni, Höpflinger, & Carr, 2015) and the role of social ties in mediating the negative effects of losing one’s spouse, particularly in non-Western contexts (Jeon, Jang, Kim, & Cho, 2013). Despite the attention, this topic has received, the nature and meaning of relationships between widowhood and depression over time and context remain unclear, largely owing to two factors: first, a focus on cross-sectional rather than longitudinal data thus limiting conclusions to simply experiencing depression rather than examining depression trajectories post-widowhood; and second, studies that focus on the experience of the West rather than comparative work to understand how aging, widowhood, and depression are related globally.

Our focus in this article is on marital transitions, gender, and depression in a variety of countries with two main lines of investigation. First, we explore whether men and women experience depression post-widowhood differently compared to each other and married counterparts. Using the dual process model (DPM) as an explanatory framework, we hypothesize that women have higher levels of depression immediately following widowhood compared to men, and recover to levels on par with married counterparts, while men do not. Second, we explore whether distinguishing depression into somatic and mood components allows us to determine not only the ways in which depression is manifest, but how it differs cross-nationally. This line of investigation is a first across international contexts by gender.

Previous Research on Widowhood and Depression

Spousal loss is associated with a variety of adverse health outcomes including decreased physical well-being, increased mortality risk (Elwert & Christakis, 2008), and poor cognitive and functional health (Vable, Subramanian, Rist, & Glymour, 2015) compared to married counterparts. A salient underlying feature is the interplay with poor mental health, largely depression, one of the most immediate responses to spousal loss.

Effect of Duration of Widowhood on Depression

Research suggests depression is strongest within the first 6 months of spousal loss (Vable et al., 2015), within the first year (Parkes, 1998), or beyond 2 years (Zisook & Shuchter, 1993). Others find that individuals are resilient in the face of spousal loss and face no depressive symptoms up to 18 months post-widowhood (Bonanno et al., 2002), thus demonstrating the wide range of duration effects with no clear consensus.

Gender and the Duration of Depression

Gender differences in depression could be explained by men being widowed later in life and spending fewer years as widowers compared to women who may engage in anticipatory planning and socializing (G. Lee, Willetts, & Seccombe, 1998). For women, if their husband was in good health prior to death, presumably pointing to low chances of dying, his sudden death can mean a longer duration of depression (Carr, House, Wortman, Nesse, & Kessler, 2001). Detrimental effects can persist years after widowhood, and men and women, regardless of remarriage, report poorer physical and mental health than married persons (Perrig-Chiello et al., 2015). It is clear the duration of widowhood matters a great deal, with depression effects varying from diminishing a few years after spousal loss to persisting for life.

Gender Differences in Widowhood and Depression

The role of gender in understanding the link between widowhood and depression is mixed and dependent on study design. Some cross-sectional studies find that men are more likely to be depressed (van Grootheest, Beekman, Broese van Groenou, & Deeg, 1999), while others find women are worse off (Perrig-Chiello et al., 2015). Longitudinal studies find support for women being more depressed (Burns, Browning, & Kendig, 2015), or no gender difference (Schaan, 2013). Longitudinal studies illuminate patterns over time: In the short-term, both men and women are depressed, while in the long run, differences emerge (Lee et al., 1998). Women are significantly more likely to be selected into early widowhood, and men are more likely to remarry following widowhood (Sasson & Umberson, 2014). Noting these differences, it is opportune timing for a cross-national comparison.

Various bereavement theories may be applicable in explaining gender differences in depression post-widowhood. Most relevant are appraisal and coping theories (Lazarus & Folkman, 1984) such as emotion-focused coping and problem-focused coping. Emotion-focused coping refers to behavioral and cognitive responses which enable individuals to manage emotional consequences of stressors (Billings & Moos, 1981), such as spousal loss. Women may be better equipped to navigate emotions oscillating between positive and negative affect (Bisconti, Bergeman, & Boker, 2004) by relying on existing support networks. However, while women deal with grief more openly, they tend to resume household management sooner than men, and may be unable to handle grief sufficiently using emotion-focused coping (Carr & Utz, 2001). This could explain elevated depression levels following widowhood. Problem-focused coping on the other hand attempts to analyze and eliminate sources of stress, and is largely attributed to men (de Ridder, 2000). This is a particularly weak explanation of gender differences in bereavement, given that the main stressor—spousal loss—is fixed rather than fixable. For a thorough review of theories related to gender differences in dealing with spousal loss, refer to Stroebe’s in-depth analysis (2001).

We argue that differential experiences in coping with spousal loss and subsequent different durations of depression can be explained by the DPM (Stroebe, 2001; Stroebe & Schut, 2010). Rather than formally test this model, we use it as a heuristic framework to explain different emerging patterns in our analyses. The DPM identifies two sources of stress comprising the loss-restoration orientation: the direct stress that comes from spousal loss (loss dimension) such as appraisal of loss or grief work, and the secondary stress that comes from coping with spousal loss (restoration dimension), such as restructuring one’s life following widowhood (Stroebe, 2001). This is an appealing theory since it explains that individuals can oscillate between these two components, a more realistic explanation of bereavement. More importantly, gender and duration dimensions are easily incorporated in this theory: Women appear to be more loss-oriented following bereavement and are able to express distress at their loss, while men are more restoration-oriented and tend to actively engage with practical issues associated with loss (Stroebe, 2001). The authors predict that over time, a positive coping strategy would enable a focus on more restoration-oriented tasks compared to loss-oriented tasks (Stroebe & Schut, 2010). We hypothesize that men and women will experience depression following widowhood differently: Widowed women will have higher levels of depression immediately following widowhood compared to men—possibly due to loss-oriented tasks—but over time recover to depression levels comparable to married counterparts sooner than widowed men, possibly due to a switch to restoration-oriented tasks.

Cross-National Comparisons for Widowhood and Depression

Context matters in complex ways when studying widowhood and depression, and can lead to varied results depending on country-specific norms like language, culture, and family expectations (Ploubidis & Grundy, 2009). Mechanisms linking widowhood and health may depend on gender and local norms (Carr & Bodnar-Deren, 2009), potentially resulting in different patterns of depression across countries. Widowed men in Europe (Van de Velde, Bracke, & Levecque, 2010) and England (Bennett, Smith, & Hughes, 2005) tend to have higher depression scores than women, with the English study finding that men have a harder time coping in the short- and long-term following widowhood. Having positive or negative relationships with children and other family members was strongly associated with lower or increased depression respectively for widowed men in England (Stafford, McMunn, Zaninotto, & Nazroo, 2011), which implies the importance of relationship quality, not simply social integration. No gender differences were found in China, although social support from children moderated the level of depression in the country (Li, Liang, Toler, & Gu, 2005). Indeed, bereavement may affect loneliness which then impacts other depressive symptoms (Fried et al., 2015), thus signaling the importance of social or family ties in navigating spousal loss and depression. This may be particularly true in Asia, where familial support and filial piety is traditionally expected, with some evidence that changes in expectations are underway, particularly in China (Li et al., 2005). The role of family is not negligible; co-residence with adult children could alleviate the pain and financial stress of spousal loss and lead to lower depression scores as was found among Korean women (Do & Malhotra, 2012), while living alone has been found to be associated with increased depressive symptoms and suicidal ideation among Korean men (Jeon et al., 2013). There is some evidence for depression post-widowhood for women in Asia compared to men, but no studies that follow trajectories to look at duration effects, a specific aim of this paper.

Distinguishing Depression into Somatic and Mood Affects

Studies that distinguish the Center for Epidemiologic Studies Depression Scale (CES-D) components split them into four pieces using the 11 or 20-item scale—somatic complaints, interpersonal problems, depressed mood, and positive mood—with some level of correlation between the four components (Kohout, Berkman, Evans, & Cornoni-Huntley, 1993). Men and women report lower positive affect after widowhood, but changes in depressive symptomology in the long-term are reported for women alone (Burns et al., 2015). Somatic symptoms are particularly severe for those in poor health prior to spousal loss (Utz, Caserta, & Lund, 2011) and can be related to the longer duration of depressive symptoms. Somatic components of depression are especially sensitive to age-related decline in functional and physical health, more so than mood-related symptoms and may be over-reported among the elderly (Fonda & Herzog, 2001). Mood-related and anxiety disorders are common for widowed individuals (Onrust & Cuijpers, 2006), particularly in the first year following widowhood. The failure to disentangle somatic and mood symptoms may conceal factors specific to the pathology and treatment. The progression of illness is also affected depending on type of depression: Individuals with severe somatic symptoms have a higher risk of suicide, mortality due to other causes, and lower functional health (Kapfhammer, 2006). Cultural factors may impact the meaning attached to somatic disorders: In China, mood disorders are presented through somatic symptoms, which allows access to mental health services with reduced stigma (Lim et al., 2011). We thus distinguish depression into somatic and mood symptoms to test how these aspects of depression may affect widowhood differently across both duration of widowhood and international context.

Data and Methods

We use the longitudinal Health and Retirement Study (HRS) from 2002 to 2012 for data from the United States and HRS-Family studies from England, Europe, Korea, and China for our analyses. Specifically: the English Longitudinal Study of Ageing (ELSA) from 2002 to 2012; the Survey of Health, Ageing and Retirement in Europe (SHARE) from 2004 to 2013; the Korean Longitudinal Study of Aging (KLoSA) from 2006 to 2012; and the China Health and Retirement Study (CHARLS) from 2011–2013. We selected these specific waves to represent comparable time frames while recognizing that country-specific factors are likely to be unique. The European data spans 11 countries from Scandinavia to the Mediterranean; however, not all countries are included in all the waves. We include countries in the North (Denmark, Sweden), South (Italy, Spain, Greece), and Continental Europe (Austria, Belgium, France, Germany, Netherlands, Switzerland)—demarcations with distinct patterns for health and aging (Borsch-Supan et al., 2005). These studies are comprehensive longitudinal surveys of representative samples of adults over 50, designed to be comparable across countries. We use the University of Southern California’s Gateway to Global Aging Data for harmonized datasets with identically defined variables for cross-country analysis and supplement these with raw data files for year of spousal death in countries where this variable is missing in the harmonized data. For further information on specific sampling and methodology used, please refer to the Gateway to Global Aging Data guide (J. Lee, 2015).

Variables and Sample

All analyses are weighted and adjusted for sample design. We rescaled person-level weights in each country at each wave to represent the probability of response conditional on inclusion at baseline (Heeringa, West, & Berglund, 2013). Our final sample sizes consist of married individuals age 55 and above at baseline: 6,637 from the United States; 2,740 from England; 5,811 from Europe; 3,813 from Korea; and 7,834 from China.

Dependent variable

Depression is measured in two ways: The CES-D scale for the HRS, ELSA, KLoSA, and CHARLS, and the EURO-D scale for SHARE. While the scales of both are different—12 and eight items respectively—they are valid comparable measures (Crimmins, Kim, & Solé-Auró, 2010). The HRS and ELSA contain the same items in their CES-D measurement about the respondent’s feelings over the week prior to the interview: depression, everything is an effort, sleep is restless, felt alone, felt sad, could not get going, whether the respondent felt happy and enjoyed life. The CHARLS and KLoSA surveys contain a slightly modified eight-item CES-D scale since some items were omitted in those countries. In CHARLS, we substituted “feeling afraid” for “feeling sad,” and “hopeful about the future” instead of “felt like you enjoy life.” In KLoSA, we substituted “had trouble concentrating” for “felt like everything was an effort.” The CES-D has been validated in Asian settings, thus cultural differences in translating the CES-D questionnaire and in reporting depressive symptoms are not a concern in Korea (Mackinnon, McCallum, Andrews, & Anderson, 1998) or China (Chin, Choi, Chan, & Wong, 2015). We distinguish the depression scales into somatic and mood affects, which have been parsed out in recent studies in a similar manner (Carleton et al., 2013) to gain a nuanced understanding of depression pathways. In this study, “somatic” affects include whether respondent felt everything was an effort, had restless sleep, had difficulty getting going, or trouble concentrating (Korea), while “mood” affects refers to whether the respondent felt depressed, alone, sad, happy, enjoyed life, fearful (China), and hopeful about the future (China).

Independent variables

The main independent variable is marital transition, thus we restrict our sample to all individuals who reported being married at baseline. We follow them through the respective analytical periods in each country to determine whether they remain married or experience widowhood. Marital status and year of widowhood (as applicable) is recorded at each wave, and allows us to longitudinally record marital transitions. We exclude individuals who transition to divorce and/or remarriage as that requires a different set of assumptions and theories. Since previous research has found that effects on depression are stark in the beginning and diminish over time, we categorize our marital transition variable with five levels: married (reference group), widowed for less than a year (11 months or less), widowed for a year (12 months), widowed for 2 years (13–24 months), widowed for 3–5 years (25–60 months), and widowed for 6–10 years (61+ months). To keep units consistent across countries, we measure death in years rather than months.

Control variables

We measure age in groups to estimate the effect of the independent variable when moving between categories: 55–64, 65–74, and 75 and above. We measure education in a similar categorical manner to keep different mandatory schooling levels in international cohorts consistent: <11 years, 12 years, college and above. Not all surveys ask comparable questions about social ties; therefore, we use number of living children as a proxy. Finally, we include a log measure of household assets to determine the relationship between economic or financial loss following widowhood on mental health. These assets are comparable across countries; refer to the Gateway to Global Aging Data guide (J. Lee, 2015) for country-specific variable details.

Statistical Methods

We use random effects regression models to estimate the effect of marital trajectories on depression for each country and component of depression. We chose random effects models after conducting Hausman tests between fixed and random effects estimators and assume that individual error terms are not correlated with predictors. This allows the inclusion of time invariant variables like gender as explanatory variables which are central to our investigation. We include an interaction term between gender and marital transition to estimate trajectories separately for men and women. The model is specified below, where $Y_{it}$ is a measure of depression for individual $i$ at time $t$; ${{\rm X}}_{it}$ is a vector of time-varying variables (marital transition, age, household assets, number of children) with $\beta$ coefficient; $\propto$is a function of time invariant variables; $u_{it}$ is the between-person error term; and ${\epsilon }_{it}$ is the within-person error term.

$$Y_{it}= \beta {{\rm X}}_{it}+ \propto + u_{it}+ {\epsilon }_{it}$$

Results

Descriptive Statistics

The first panel for each country refers to individuals who were continuously married and the second panel refers to individuals who were married at baseline but widowed at some point during the follow-up period. Table 1 shows that across all countries, most individuals who transitioned to widowhood were in the 65–74 age group at baseline, and a majority of these were women. In terms of socio-economic status (SES), individuals with less than 11 years of schooling in England, Europe, and China and who had lower baseline household assets on average made up a large proportion of those who transitioned to widowhood.

Table 1.

Sample Descriptive Characteristics at Baseline by Country and Subsequent Marital Transition (2002–2013)

Variable	United States			England			Europe			Korea			China
	CM	TW	t-test	CM	TW	t-test	CM	TW	t-test	CM	TW	t-test	CM	TW	t-test
Age (%)			−20.0*			−13.07*			−15.8*			−11.5*			−10.1*
55–64	62.2	32.0		62.5	32.7		58.3	29.1		66.6	32.2		63.6	34.2
65–74	30.7	42.3		31.9	43.1		33.5	46.0		27.2	44.9		28.5	35.1
75+	7.2	25.8		5.6	24.3		8.2	24.9		6.2	22.9		7.9	30.7
Women (%)	43.5	71.4	−18.1*	65.6	48.5	−7.3*	46.3	69.5	−10.5*	43.2	82.5	−12.4*	45.8	67.2	−5.5*
Education (%)			7.8*			5.9*			5.9*			8.4*			2.5
<11 years	19.0	26.4		49.2	65.6		62.2	72.8		12.8	32.4		49.2	61.7
12 years	30.3	38.8		16.1	12.0		8.8	6.7		76.6	64.1		44.7	34.5
College and above	50.7	34.9		34.7	22.4		29.0	20.6		10.6	3.5		6.0	3.8
Household assets (mean)	545.8 (13.8)	448.9 (23.5)	4.3*	305.8 (11.2)	205.6 (11.8)	3.4*	386.4 (16.0)	281.1 (37.9)	3.1*	14.9 (0.5)	8.5 (1.4)	3.6	215.7 (22.1)	98.2 (18.1)	0.9
Number of children (mean)	3.3 (0.03)	3.5 (0.07)	−3.6*	2.4 (0.03)	2.6 (0.08)	−2.1*	2.3 (0.02)	2.5 (0.09)	−1.7*	3.1 (0.02)	3.8 (0.10)	−6.4	2.9 (0.02)	3.4 (0.1)	−5.1*
Region (%)
North							4.9	5.2	−0.8
Central							52.4	47.6
South							42.7	47.3
N	5,334	1,303		2,357	383		5,226	585		3,521	292		7,629	204
Baseline year	2002	2002		2002	2002		2004	2004		2006	2006		2010	2010
Final year	2012	2012		2012	2012		2010	2010		2012	2012		2012	2012
Total waves	6	6		6	6		3	3		4	4		2	2

Note: Two-sample t-tests are significant at p < .05 if marked by (*). CM = continuously married (married from baseline through the follow-up period), TW = transitioned to widowhood (married at baseline and became widowed during the follow-up period). Std. errors in parentheses. Region is specified only for the European SHARE sample.

Women comprise an overwhelming proportion of those who transition to widowhood, as shown in Table 2. Comparing England and the United States, in the 10-year period between baseline and follow-up close to 10% of men in both countries experience widowhood compared to 25% of women in the United States and 20% of women in England. The proportions are half that in Europe over the 6-year analytic period, with 5% of men and 12% of women becoming widowed. Even with a relatively short analytic period in Korea, about 12% of women experience widowhood, compared to only 2% of men in that country, while 2% of men and 4% of women experience widowhood in China. Differences in mean durations of widowhood between widowed men and women show that women have significantly longer durations of widowhood in the United States and England. These differences are not significant in the other countries.

Table 2.

Proportion of Men and Women Who Remained Married or Transitioned to Widowhood From Baseline to Follow-Up by Country and Mean Duration of Widowhood at Follow-Up for Those Who Experience Widowhood (2002–2013)

Country/region	Men	Women	Total	t-test
United States
Continuously married	90.6	74.9	83.0	−18.1*
Transitioned to widowhood	9.4	25.1	17.0
Mean duration of widowhood (SE)	4.1 (0.2)	4.4 (0.1)		−2.1*
N	3,232	3,347	6,637
England
Continuously married	90.4	80.6	85.7	−7.3*
Transitioned to widowhood	9.6	19.4	14.3
Mean duration of widowhood (SE)	1.7 (0.1)	2.4 (0.1)		−3.3*
N	1,395	1,345	2,740
Europe
Continuously married	95.1	88.0	91.7	−10.5*
Transitioned to widowhood	4.9	12.0	8.3
Mean duration of widowhood (SE)	3.6 (0.3)	3.4 (0.2)		−1.1
N	2,903	2,908	5,811
Korea
Continuously married	98.0	88.5	93.6	−12.4*
Transitioned to widowhood	2.1	11.5	6.4
Mean duration of widowhood (SE)	2.4 (0.3)	2.6 (0.1)		−0.9
N	2,010	1,803	3,813
China
Continuously married	98.3	96.1	97.3	−5.5*
Transitioned to widowhood	1.7	3.9	2.7
Mean duration of widowhood (SE)	0.6 (0.1)	0.6 (0.0)		0.4
N	4,169	3,665	7,834

Note: Two-sample t-tests are significant at p < .001 if marked by (*).

While we see that men and women transition into widowhood differently, it is important to ascertain whether depression scores actually change from baseline to follow-up for the sample. Table 3 parses overall, somatic, and depression scores by marital transition between baseline and follow-up. While the event of widowhood can occur at any wave between baseline and follow-up, we only include measures of depression at baseline and follow-up to maintain consistency across countries, where intervening time periods can be as short as 2 years (China) to as long as 10 years (United States, England). There are two noteworthy patterns. First, across all counties, individuals who transition to widowhood have higher baseline depression scores than those who remain married in the follow-up period. Second, mood depression scores drive change in the United States, England, and Europe, while somatic depression drives change in Korea and China.

Table 3.

Change in Depression Scores from Baseline to Follow-Up by Country and Marital Transition (2002–2013)

	Baseline			Follow-up
	CM	TW	t-test	CM	TW	t-test
United States (N = 6,637)
CES-D overall	0.98 (0.03)	1.25 (0.05)	−5.7*	1.00 (0.03)	1.86 (0.07)	−14.1*
CES-D somatic (0–3)	0.56 (0.01)	0.68 (0.03)	−4.7	0.59 (0.01)	0.79 (0.03)	−6.0*
CES-D mood (0–5)	0.42 (0.02)	0.57 (0.03)	−5.0	0.41 (0.02)	1.07 (0.05)	−16.8*
England (N = 2,740)
CES-D overall	1.06 (0.03)	1.57 (0.10)	−5.5*	1.07 (0.04)	2.11 (0.11)	−11.2*
CES-D somatic (0–3)	0.64 (0.02)	0.94 (0.05)	−5.8*	0.64 (0.02)	0.89 (0.05)	−5.1*
CES-D mood (0–5)	0.42 (0.02)	0.63 (0.06)	−3.9*	0.43 (0.02)	1.22 (0.07)	−13.5*
Europe (N = 5,811)
EURO-D overall	2.22 (0.04)	2.75 (0.15)	−5.0*	2.54 (0.05)	3.60 (0.15)	−−9.6*
EURO-D somatic	0.95 (0.02)	1.19 (0.07)	−3.8*	1.17 (0.02)	1.50 (0.09)	−6.1*
EURO-D mood	1.27 (0.03)	1.56 (0.09)	−4.8*	1.37 (0.03)	2.10 (0.10)	−9.9*
Korea (N = 3,813)
CES-D overall	2.67 (0.05)	3.49 (0.17)	−5.0*	3.58 (0.05)	5.07 (0.19)	−7.7*
CES-D somatic (0–3)	0.97 (0.02)	1.47 (0.09)	−5.5*	1.53 (0.03)	2.35 (0.10)	−7.9*
CES-D mood (0–5)	1.70 (0.03)	2.02 (0.09)	−3.4*	2.05 (0.03)	2.73 (0.10)	−6.4*
China (N = 7,833)
CES-D overall	3.58 (0.04)	4.24 (0.23)	−2.7*	3.17 (0.04)	3.75 (0.22)	−3.1*
CES-D somatic (0–3)	2.24 (0.03)	2.56 (0.14)	−2.18	1.77 (0.03)	1.93 (0.14)	−1.2
CES-D mood (0–5)	1.35 (0.02)	1.68 (0.12)	−3.0*	1.40 (0.02)	1.83 (0.11)	−5.6*

Note: Std. errors in parentheses. CM = continuously married (married from baseline through the follow-up period), TW = transitioned to widowhood (married at baseline and became widowed during the follow-up period).

Two-sample t-tests are significant at p < .05 if marked by (*).

In the United States and England, individuals who are continuously married have roughly the same depression scores over time. In Europe and Korea, those who are continuously married also see an increase in depression scores at follow-up, but the magnitude of increase is much higher for those who transition to widowhood. Conversely, in China, there is an overall decrease in depression scores over time for all individuals, save for a slight increase in mood score for those transitioning to widowhood. We present analyses stratified by gender in the Supplementary Appendices (Supplementary Appendix 1a for men and Supplementary Appendix 1b for women).

Table 4 presents predicted depression scores across all countries for men and women. These scores are obtained after linear regression estimates are obtained. For tables detailing our models for each depression component, refer to Supplementary Appendix 2a (overall depression score), Supplementary Appendix 2b (somatic depression score), and Supplementary Appendix 2c (mood depression score).

Table 4.

Predicted Depression Scores (SE) by Gender, Country, and Marital Transition (2002–2013)

	Men			Women
	Overall	Somatic	Mood	Overall	Somatic	Mood
United States
Married	1.05 (0.02)	0.62 (0.01)	0.41 (0.01)	1.37 (0.02)	0.75 (0.01)	0.62 (0.01)
Widowed <1 years	2.07 (0.09)	0.77 (0.05)	1.30 (0.06)	2.48 (0.06)	0.91 (0.03)	1.58 (0.04)
Widowed 1 year	1.99 (0.09)	0.65 (0.05)	1.36 (0.06)	2.26 (0.06)	0.78 (0.03)	1.49 (0.04)
Widowed 2 years	1.60 (0.10)	0.63 (0.06)	1.96 (0.07)	1.78 (0.07)	0.74 (0.04)	1.05 (0.05)
Widowed 3–5 years	1.51 (0.08)	0.72 (0.04)	0.80 (0.05)	1.46 (0.05)	0.67 (0.03)	0.81 (0.03)
Widowed 6–10 years	1.30 (0.12)	0.74 (0.06)	0.56 (0.08)	1.39 (0.07)	0.67 (0.04)	0.74 (0.05)
England
Married	1.01 (0.03)	0.64 (0.01)	0.37 (0.02)	1.46 (0.03)	0.85 (0.02)	0.60 (0.02)
Widowed <1 years	2.57 (0.25)	0.96 (0.14)	1.54 (0.16)	3.37 (0.17)	1.32 (0.09)	2.06 (0.11)
Widowed 1 year	2.07 (0.15)	0.78 (0.08)	1.26 (0.09)	2.07 (0.11)	0.92 (0.06)	1.16 (0.07)
Widowed 2 years	1.63 (0.31)	0.58 (0.17)	1.10 (0.21)	2.54 (0.22)	1.09 (0.06)	1.46 (0.14)
Widowed 3–5 years	1.57 (0.29)	0.59 (0.17)	0.81 (0.20)	1.74 (0.15)	0.85 (0.08)	0.89 (0.09)
Widowed 6–10 years	1.72 (0.67)	0.77 (0.38)	0.96 (0.43)	1.52 (0.27)	0.77 (0.15)	0.79 (0.18)
Europe
Married	1.77 (0.02)	0.82 (0.01)	0.93 (0.01)	2.57 (0.02)	1.14 (0.01)	1.41 (0.01)
Widowed <1 years	3.68 (0.24)	1.51 (0.14)	2.16 (0.17)	3.86 (0.17)	1.67 (0.09)	2.25 (0.12)
Widowed 1 year	3.08 (0.18	1.24 (0.10)	1.84 (0.13)	3.80 (0.12)	1.69 (0.07)	2.16 (0.08)
Widowed 2 years	2.63 (0.18)	1.18 (0.10)	1.34 (0.13)	3.17 (0.14)	1.42 (0.08)	1.83 (0.10)
Widowed 3–5 years	1.91 (0.19)	0.85 (0.11)	1.06 (0.13)	2.56 (0.13)	1.05 (0.07)	1.60 (0.08)
Widowed 6–10 years	1.72 (0.25)	0.85 (0.14)	0.93 (0.17)	2.36 (0.16)	0.93 (0.09)	1.39 (0.11)
Korea
Married	3.36 (0.04)	1.40 (0.02)	1.95 (0.02)	3.95 (0.05)	1.73 (0.03)	2.21 (0.03)
Widowed <1 years	4.96 (0.49)	2.40 (0.27)	2.56 (0.27)	5.49 (0.25)	2.50 (0.14)	3.00 (0.14)
Widowed 1 year	4.97 (0.51)	2.26 (0.28)	2.69 (0.28)	5.29 (0.26)	2.30 (0.14)	3.01 (0.15)
Widowed 2 years	4.49 (0.56)	2.00 (0.31)	2.50 (0.31)	3.86 (0.31)	1.70 (0.17)	2.17 (0.17)
Widowed 3–5 years	4.32 (0.51)	1.91 (0.28)	2.38 (0.28)	3.85 (0.24)	1.66 (0.13)	2.22 (0.14)
China
Married	3.10 (0.04)	1.85 (0.03)	1.26 (0.02)	3.91 (0.05)	2.38 (0.03)	1.54 (0.03)
Widowed <1 years	4.12 (0.71)	1.94 (0.42)	2.18 (0.31)	4.05 (0.49)	2.63 (0.29)	1.42 (0.23)
Widowed 1 year	3.48 (0.38)	1.79 (0.25)	1.79 (0.25)	3.87 (0.57)	2.19 (0.35)	1.66 (0.21)

United States

Married men in the United States have low overall predicted depression scores (CES-D of 1.05), while those newly widowed (within 1 year of the survey) have the highest depression score at 2.07. Depression decreases over time, however, does not reach the same low level as married men even 6–10 years post-spousal loss. Women, however, have higher depression scores at all marital transition levels compared to men (peak of 2.48 immediately following widowhood), but reach the same levels of depression as married women (CES-D of 1.39) 6–10 years post-widowhood. Women have slightly higher somatic and mood depression scores compared to men up to 2 years post-widowhood, after which men have higher somatic scores and women have higher mood scores in the period after 3 years post-widowhood.

England

Overall, the pattern for predicted depression scores in England is similar to that in the United States: Depression scores for married men and women peak immediately following widowhood (CES-D of 2.57 for men and 3.37 for women) but remain elevated even at 6–10 years for widowed men compared to married men, while scores for widowed women reach levels comparable to married women by that time period. Women in England have higher somatic depression scores compared to men, while men have higher mood depression scores at 1 year and 6–10 years post-widowhood.

Europe

Unlike the United States and England, men and women in Europe have comparable depression trajectories, with scores highest immediately following widowhood (EURO-D of 3.68 for both), and subsequently becoming lower than married counterparts by 6–10 years after widowhood. Also unique to Europe: Women have higher somatic and mood depression scores compared to men, whether married or widowed.

Korea

In Korea, overall depression scores for men remain at their peak for a whole year following widowhood (CES-D of 4.97) and decrease slightly over time, differing from other countries where the peak diminishes after the first time period. However, levels at 3–5 years following widowhood for men are still on average one point higher than married men at 4.32 compared to 3.36. Korean widowed women reach the depression scores of married women by the 6–10-year period, unlike their male counterparts. While widowed women in Korea have elevated levels of somatic and mood depression compared to widowed men at or before 1 year of widowhood, men have higher scores after 2 years.

China

In China, men have slightly higher overall depression scores at 4.12 compared to 4.05 for women immediately following widowhood. The scores remain high for men at the 1-year mark, while they decrease for women in the same time period, with scores even lower than married women. Somatic depression scores are higher for women compared to men, while mood scores are higher for men.

Discussion

Our study is the first to distinguish between marital and depression trajectories by gender in a cross-national context with an empirical distinction between somatic and mood depression. We build on seminal work from the United States (Carr & Bodnar-Deren, 2009; Sasson & Umberson, 2014) that analyzes gender differences in widowhood and depression, and add a comparative frame of analysis with international populations previously studied in China (Li et al., 2005) and Korea (Jeon et al., 2013). Our descriptive findings are in line with work that finds women are more likely to transition to widowhood compared to men (Tiedt, 2010) and some evidence that men are more likely to remain depressed compared to women in the long-term (G. Lee & DeMaris, 2007; G. Lee et al., 1998).

Our analysis on depression and marital transitions unearth differing patterns by gender and context. Bivariate results suggest that pre-widowhood depression scores are higher for widowed individuals than for individuals who remain married. We find regional patterns with respect to types of depression in our analysis of scores between baseline and follow-up. Somatic depression symptoms seem to increase more in Korea, while mood depression seems to drive overall depression in the West. While these results are not entirely surprising given that depression in Asia seems to be somaticized (Simon, VonKorff, Piccinelli, Fullerton, & Ormel, 1999), it points to previous work that finds it is perhaps more acceptable to have somatic symptoms than mood-related symptoms in a setting where depression is highly stigmatized (Simon, Gater, Kisely, & Piccinelli, 1996) although determining that is beyond the scope of this paper. China is a puzzling case within the Asian sample, where all individuals see a decrease in depression. While East Asia has experienced much demographic change, Korea has seen faster increases in female labor force participation, later marriage and lower fertility, and decreased importance on caregiving for elderly parents compared to China (Raymo, Park, Xie, & Yeung, 2015), which could explain the different trajectories within both countries. Co-residence with children is common in China and could explain better psychological well-being and lower depression (Wang, Chen, & Han, 2014). However, regression analyses for overall depression scores actually show that having a higher number of living children is related to an increase in depression in Europe, Korea, and China, which contradicts bivariate results for China. A recent study in Japan found that widowed parents who received emotional support from their children had lower psychological morale than those who believed in the notions of filial piety prevalent in this culture (Takagi & Saito, 2013), which could explain our results. Measuring relationship quality is out of the purview of this current study, but is something we would like to explore once those data are made available.

Our analyses also shed light on gender differences in depression trajectories, a first in a comparative context. Across all countries except for the European sample, we find that men who transition to widowhood have higher depression scores over time compared to their married counterparts, while women recover to the same levels as married women. This is consistent with the DPM and our hypothesis that stated men are more likely to remain depressed, while women are more likely to recover after elevated levels of depression in the short-term. This is also consistent with research in the United States (G. Lee & DeMaris, 2007) and recent evidence from Korea (Jang et al., 2009) that may signal the importance of social connectedness as an important means of coping for women compared to men, something that can be described as a restoration-oriented task in the DPM. Our European results on men and women recovering from depression peaks to levels comparable to married counterparts are in contrast with research from the Netherlands that finds men continue to have high levels of depression many years post-widowhood (van Grootheest et al., 1999). While we find evidence for increased depression among women that is in line with other work (Li et al., 2005), we find that elevated depression is not sustained over time, something other studies have missed by not examining trajectories longitudinally. We also do not find any evidence on resilience in the face of widowhood for men or women; rather we see a recovery for women across all countries which could point to another aspect of coping.

Another interesting finding is that widowed women have higher somatic mood scores than men, while men have higher mood scores across different contexts. This contradicts the DPM in a way, since it would be expected that women’s elevated depression scores compared to men’s would be a result of loss-oriented tasks that allow time for grief work (Wijngaards-de Meij et al., 2008), which would indicate mood-related depression rather than somatic. Women’s recovery to levels comparable to married counterparts is in line with restoration-oriented tasks like becoming active in social organizations or groups, something that men do not do as successfully (Stroebe & Schut, 2010). Previous research has found that somatic symptoms tend to be over-reported among the elderly and could be linked to age-related decline in functional and physical health (Fonda & Herzog, 2001), but why this affects men and women differentially is still a puzzle. It could be that women have higher health care utilization than men, resulting in more somatic diagnoses, something that needs to be explored further.

There are some limitations in this study that must be addressed. First, we include modified CES-D items for China and Korea based on what questions were included in those surveys. These have not been validated, and may bias the results if used inaccurately. Second, the time frame for CES-D scores are different in Korea (year prior to the interview) compared to other countries where the standard is 2 weeks prior to the interview, and may lead to conclusions that are actually not comparable across countries. Finally, our small sample sizes for China, Korea, and widowed men may restrict significance levels in regression estimates. Our research shows the complex relationship between widowhood and depression, not only in addressing whether mood or somatic affects take precedence for men and women, but also in how these relationships differ across countries.

Supplementary Material

Supplementary material is available at The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences online.

Funding

This work was supported in part by a National Institute on Aging training grant to the Population Studies Center at the University of Michigan (T32 AG000221).