Abstract
Purpose:
We examine widowhood effects on mortality across gender and race-ethnicity, with attention to variation in the mediating role of economic resources.
Methods:
Data were drawn from the Health and Retirement Study (1992-2016). The analytic sample included 34,777 respondents aged 51 and older who contributed 208,470 person-period records. Discrete-time hazard models were estimated to predict the odds of death among white men, black men, Hispanic men, white women, black women, and Hispanic women separately. Karlson–Holm–Breen analysis was conducted to examine the mediating role of economic resources across groups.
Results:
Across all gender and racial-ethnic subgroups, widowhood effects on mortality were largest for Hispanic men. Black women and Hispanic women also suffered stronger effects of widowhood on mortality than white women. For both men and women, economic resources were an important pathway through which widowhood increased mortality risk for whites and blacks but not for Hispanics.
Conclusions:
Findings highlight that gender and race-ethnicity intersect with widowhood status to disadvantage some groups more than others. It is important to explore the complex pathways that contribute to the higher mortality risk of racial-ethnic minorities, especially Hispanic men, following widowhood so that effective interventions can be implemented to reduce those risks
Keywords: mortality, widowhood, gender, race-ethnicity
Introduction
Married people enjoy better health and greater longevity than those who are not married [1–3]. Among unmarried people, the widowed show particularly the worst health and elevated mortality risk [3]. The prevalence of widowhood is significantly higher among blacks and women compared to whites and men, pointing to substantial group differences in exposure to loss of a spouse [4]. Most studies that explore the effect of widowhood on mortality have focused on white populations or simply controlled for race-ethnicity in analyses [5,6]. A few recent studies have begun to explore distinguishable effects of widowhood on mortality between whites and blacks but Hispanics are typically not distinguished in analyses [7]. Yet, there may be important differences in marital ties across racial-ethnic groups that shape mortality risk following widowhood. For example, widowhood may work through different pathways to shape mortality risk across racial-ethnic groups given disparities in economic resources that may help in coping with widowhood. Processes that differ by race-ethnicity may further depend on gender. Many studies find gender differences in marital and widowhood experiences and recent work emphasizes that race-ethnicity and gender intersect to shape socioeconomic conditions and life outcomes [8]. Yet, few studies have examined whether the effects of widowhood on mortality vary across gender and racial-ethnic groups and whether socioeconomic resources play the same or distinct roles in linking widowhood and mortality across groups. Mortality disparities associated with widowhood are especially important in the context of increasingly diverse aging populations.
Evidence on Widowhood and Mortality Across Gender and Race-Ethnicity.
Past longitudinal studies consistently show that widowhood is associated with poor health and increased mortality risk [2,9–11, 12]. Although a comparative analysis of race and ethnicity is largely absent from this literature, gender comparisons have received much attention and yield mixed results. The preponderance of the evidence suggests that the adverse effect of widowhood on mortality risk is greater for men than women [13–16]. This is reflected in a meta-analysis of 2,263,888 subjects from 15 prospective cohort studies suggesting that the widowhood effect on mortality is significant for men but not for women [17]. In contrast, data from the Israel Longitudinal Mortality Study found no gender differences in relative risk of mortality after spousal loss [18]. A longitudinal study of elderly married couples in the United States from Medicare databases found that the death of a spouse increased all-cause mortality of both men and women, although wives’ causes of death seemed to be more important for husbands’ mortality than husbands’ causes of death for wives’ mortality [19].
Evidence on variation in widowhood effects on mortality by race-ethnicity is also limited and inconsistent. An analysis of Medicare data suggests that widowhood increases mortality risk for white men and women, but not for black men and women [7]. Data from the National Longitudinal Mortality Study (NLMS) suggests a widowhood/mortality link for white men and women and black men aged 45-64 and black and white women over 65, but not for black women 45-65 or white men over 65 [20]. A survey cohort analysis of National Health Interview Surveys (NHIS) found a widening mortality gap between married and widowed people only among non-Hispanic white women but not among non-Hispanic white men or non-Hispanic black men/women [21]. Still other studies found no racial difference in mortality risk following widowhood [15,18]. Inconsistencies across studies, particularly concerning race and gender, may reflect differences in sample composition, measures of mortality, and methods of analysis.
Few studies have examined the link of widowhood with mortality for Hispanics compared with white and black populations. Data from the Hispanic Established Populations for the Epidemiologic Studies of the Elderly (Hispanic EPESE) 1993-2000 suggests an increased risk of death following widowhood for Hispanic men but not Hispanic women [22]. We know of no studies that examine whether widowhood effects on mortality are similar or different for Hispanics in comparison to their white and black counterparts.
Economic Resources: Linking Widowhood to Mortality across Race-Ethnicity and Gender.
One of the most frequently proposed mechanisms through which widowhood increases mortality risk is change in economic resources [9,10,13,23], a process that may differ by race-ethnicity and gender. Widowhood increases economic vulnerability and financial strain, especially for women [13,30] and racial-ethnic minorities [24]. Loss of income is sometimes cited as a reason for widowhood’s negative effects on health among women [13, 22]. Given that Hispanic and black spouses are socioeconomically disadvantaged relative to white spouses before widowhood [24] and women experience a greater financial decline following widowhood than do men, economic resources may be a more significant factor for mortality risk of women and racial-ethnic minorities following widowhood.
In this study, we estimate the effects of widowhood on mortality risk at the intersection of gender and race-ethnicity. We pay close attention to potential variation in the mediating role of economic resources through which widowhood shapes mortality across gender and racial-ethnic groups. The importance of this study is highlighted by the long-documented adverse effects of widowhood on health and longevity, the greater prevalence of widowhood among blacks and women than whites and men, and persistent racial-ethnic disparities in mortality and access to economic resources.
Data
We analyzed data from the Health and Retirement Study (HRS) (1992-2016). The 1992 wave of the HRS surveyed a national sample of noninstitutionalized adults born in 1931-41 (aged 51-61 in 1992) and their spouses (HRS, 2017). Participants have been interviewed, by telephone or in person, approximately every two years since then, and several other cohorts have since been added to the original HRS sample. The HRS has high response rates (81-89%) in each wave, and provides a unique opportunity to address the current research question because of its large sample size, long-term follow-up of mortality and marital status, and high-quality measures of economic resources and other key sociodemographic variables such as age, gender, race/ethnicity and education. The survey oversampled blacks and Hispanics and collects detailed information on physical, economic, and family conditions as well as mortality approximately every two years. All analyses were adjusted for the complex survey design using the STATA SVY command [25]. We excluded 7,276 respondents from the analysis due to missing data on mortality, widowhood status or other key variables. The final analytic sample included 34,777 respondents aged 51 and older who contributed 208,470 person-period records.
Measures
Mortality.
The dependent variable is the mortality status between waves. The HRS followed the mortality status of respondents during each follow up wave.
Widowhood and other marital status.
Our key independent variable is whether the respondent was widowed (time-varying) at the interview. We also controlled for whether the respondent was remarried at the last follow-up wave (time-invariant) and whether the respondent had never married by the last follow-up wave (time-invariant).
Economic resources.
We measured economic resources with two time-varying variables: total household income and net household wealth. Total household income included respondent’s and spouse’s income from all sources such as earnings, pensions and annuities, Supplemental Security Income and Social Security Disability, Social Security retirement, other government transfers, unemployment and workers’ compensation, household capital income, and other income for the last calendar year. Net household wealth was measured as the total value of household assets minus household debts. We used the RAND HRS version of household income and wealth data, which included consistently imputed missing values across waves. Because household income and wealth had zero or negative values, we further adjusted household income by adding a constant of $100 and adjusted wealth by adding a year-specific constant (depending on the minimum value of wealth in that specific year) in addition to $100 to all respondents so that all wealth and income values were transformed to positive. We then divided the imputed income and wealth by the square root of household size and took the natural logs of the values [11].
Other covariates.
We controlled for age at the baseline survey (in years, time-invariant) and education (less than high school [reference], high school graduate, some college, and college graduates, time-invariant). We also controlled for whether the respondent was born in the South (including East South Central States, West South Central States and South Atlantic States). The analyses were stratified by gender (men and women) and race/ethnicity (non-Hispanic white [hereafter “white”], non-Hispanic black [hereafter “black”], and Hispanics).
Statistical Methods
We estimated discrete-time hazard models to analyze the mortality data. Specifically, we created person-period record files and ran binary logit models. A respondent contributed an observation for each wave at which they were interviewed, up to death or right censoring (i.e., loss to follow-up). The discrete-time hazard model is specified as:
where h(tij) indicates the discrete mortality hazard (i.e., conditional probability) for individual i at wave j. represents the set of intercepts for the 12 periods (two-year intervals) of HRS between 1992-2016, one per period. Xi indicates the vector of time-invariant covariates and Zi(j−1) indicates the vector of time-varying covariates including widowhood status, which were measured one wave prior to mortality status. B1 and B2 are corresponding coefficient vectors. We estimated two models: Model 1 controlled for basic sociodemographic covariates, and Model 2 added economic resource variables (i.e., household income and wealth) to test if economic resources explained the association between widowhood and mortality. We conducted formal mediation testing using the Karlson–Holm–Breen (KHB) method to examine whether economic resources had significant mediating effects. The KHB method is useful for decomposing the total effect into the direct and indirect effects in non-linear probability models such as logistic models [26]. To better understand potential variation by gender and race/ethnicity, we stratified all analyses by six gender and racial-ethnic groups: white men, black men, Hispanic men, white women, black women, and Hispanic women. We conducted Wald tests to determine whether individual regression coefficients were statistically different across gender and racial/ethnic subgroups [27].
Results
Table 1 shows descriptive statistics of all analyzed variables for the total sample as well as by gender and race-ethnicity. The results show that a higher proportion of black men (7.08% vs. 6.24%, p < .001) died across survey waves compared to their white counterparts although black women and white women were not significantly different from each other in terms of proportion who died. In contrast, a lower proportion of both Hispanic men (4.47% vs. 6.24%, p < .001) and Hispanic women (3.65% vs. 6.15%, p < .001) died across waves than their white counterparts. Moreover, a higher proportion of both black men (13.34% vs. 10.01%, p < .001) and black women (32.84% vs. 30.98%, p < .001) were widowed than their white counterparts while a lower proportion of both Hispanic men (7.30% vs. 10.01%, p < .001) and Hispanic women (22.95 vs. 30.98%, p < .001) were widowed compared to their white counterparts. Hispanic men (2.89% vs. 3.70%, p < .001) and Hispanic women (2.60% vs. 2.99%, p < .05) as well as black women (1.92% vs. 2.99%, p < .001) were less likely to remarry than their white counterparts although black men were not significantly different from white men in terms of proportion remarried. Both black men (9.09% vs. 4.47%, p < .001) and black women (11.29% vs. 3.70%, p < .001) as well as Hispanic women (6.28% vs. 3.70%, p < .001) were more likely to never marry than their white counterparts although Hispanic men were not significantly different from white men in terms of proportion of never marrying. Blacks and Hispanics, both men and women, were younger, had lower levels of education, lower income and wealth compared to their white counterparts. For both men and women, blacks were much more likely to be born in the South while Hispanics were less likely to be born in the South than their white counterparts.
Table 1.
Weighted Percentages and Means (Standard Deviations), HRS 1992-2016a
| Total Sample N of respondents=34,777 N of person-periods=208,470 |
||||||
|---|---|---|---|---|---|---|
| Died | 6.08% | Born in South | 31.15% | |||
| Widowed | 21.00% | Education | ||||
| Remarried | 3.22% | Less than high school | 20.00% | |||
| Never married | 4.82% | High school graduate | 34.29% | |||
| Female | 53.12% | Some college | 22.72% | |||
| Race-ethnicity | College graduate | 22.99% | ||||
| Non-Hispanic White | 82.23% | Age | 59.320 (10.262) | |||
| Non-Hispanic Black | 10.46% | Income (logged) | 10.159 (1.145) | |||
| Hispanics | 7.31% | Wealth (logged) | 14.096 (0.853) | |||
| White Men N of respondents=10,941 N of person-periods=65,941 |
Black Men N of respondents=2,517 N of person-periods=12,136 |
Hispanic Men N of respondents=1,725 N of person-periods=8,515 |
||||
| Died | 6.24% | 7.08% | *** | 4.47% | *** | |
| Widowed | 10.01% | 13.34% | *** | 7.30% | *** | |
| Remarried | 3.70% | 3.94% | 2.89% | *** | ||
| Never married | 4.47% | 9.09% | *** | 4.18% | ||
| Born in South | 26.28% | 71.12% | *** | 23.11% | *** | |
| Education | ||||||
| Less than high school | 14.06% | 34.64% | *** | 49.75% | *** | |
| High school graduate | 32.43% | 29.68% | *** | 22.65% | *** | |
| Some college | 22.77% | 2.22% | *** | 17.31% | *** | |
| College graduate | 30.75% | 13.47% | *** | 10.30% | *** | |
| Age | 58.515 | (9.376) | 57.705 | (8.644)*** | 56.705 | (8.363)*** |
| Income (logged) | 10.468 | (0.970) | 9.812 | (1.190)*** | 9.460 | (1.423)*** |
| Wealth (logged) | 14.125 | (0.834) | 13.975 | (0.860)*** | 13.874 | (0.859)*** |
| White Women N of respondents=13,573 N of person-periods=88,895 |
Black Women N of respondents=3,793 N of person-periods=20,747 |
Hispanic Women N of respondents=2,228 N of person-periods=12,236 |
||||
| Died | 6.15% | 6.30% | 3.65% | *** | ||
| Widowed | 30.98% | 32.84% | *** | 22.95% | *** | |
| Remarried | 2.99% | 1.92% | *** | 2.60% | * | |
| Never married | 3.70% | 11.29% | *** | 6.28% | *** | |
| Born in South | 27.11% | 73.47% | *** | 20.88% | *** | |
| Education | ||||||
| Less than high school | 16.04% | 34.38% | *** | 58.39% | *** | |
| High school graduate | 39.07% | 30.26% | *** | 21.65% | *** | |
| Some college | 24.01% | 22.29% | *** | 13.97% | *** | |
| College graduate | 20.87% | 13.08% | *** | 5.99% | *** | |
| Age | 60.619 | (11.172) | 58.774 | (10.081)*** | 57.953 | (9.909)*** |
| Income (logged) | 10.163 | (1.011) | 9.403 | (1.199)*** | 9.192 | (1.478)*** |
| Wealth (logged) | 14.136 | (0.847) | 13.956 | (0.881)*** | 13.907 | (0.861)*** |
Two-tailed t-tests comparing with white counterparts:
p < 0.001,
p < 0.01,
p < 0.05
Statistics are based on person-period files.
Table 2 shows the adjusted odds ratios of death from the discrete-time hazard models for the total sample, suggesting that those who were widowed had significantly higher odds of death between survey waves than those who did not experience widowhood (OR = 1.398, 95% CI = 1.329, 1.471). After controlling for income and wealth as additional covariates in Model 2, the effect of widowhood on mortality reduced but remained significant (OR = 1.332, 95% CI = 1.266, 1.402). Both income (OR = 0.815, 95% CI = 0.796, 0.834) and wealth (OR = 0.886, 95% CI = 0.831, 0.945) were negatively related to mortality.
Table 2.
Adjusted Odds Ratios of Mortality from Discrete-Time Hazard Models for the Total Sample
| Model 1 | Model 2 | |
|---|---|---|
| Widowed | 1.398*** (1.329 - 1.471) |
1.332*** (1.266 - 1.402) |
| Remarried | 0.617*** (0.536 - 0.710) |
0.659*** (0.572 - 0.758) |
| Never married | 1.418*** (1.257 - 1.600) |
1.265*** (1.124 - 1.423) |
| Age | 1.103*** (1.101 - 1.106) |
1.102*** (1.100 - 1.105) |
| Female | 0.634*** (0.609 - 0.660) |
0.610*** (0.586 - 0.636) |
| Race-ethnicity (0=non-Hispanic white) | ||
| Non-Hispanic black | 0.999 (0.921 - 1.084) |
0.889** (0.819 - 0.966) |
| Hispanic | 0.655*** (0.555 - 0.774) |
0.533*** (0.444 - 0.641) |
| Born in South | 1.186*** (1.124 - 1.251) |
1.165*** (1.104 - 1.230) |
| Education (0=less than high school) | ||
| High school graduate | 0.784*** (0.749 - 0.821) |
0.850*** (0.810 - 0.891) |
| Some college | 0.654*** (0.613 - 0.698) |
0.748*** (0.699 - 0.800) |
| College graduate | 0.482*** (0.445 - 0.523) |
0.614*** (0.564 - 0.668) |
| Income (logged) | 0.815*** (0.796 - 0.834) |
|
| Wealth (logged) | 0.886*** (0.831 - 0.945) |
|
| N of respondents | 34,777 | 34,777 |
| N of person-periods | 208,470 | 208,470 |
p < 0.001,
p < 0.01,
p < 0.05.
95% confidence intervals in parentheses.
Table 3 shows the adjusted odds ratios of death from the discrete-time hazard models by race-ethnicity for men. Results of Model 1 in Table 3 suggest that white men who were widowed had significantly higher odds of death between survey waves than white men who did not experience widowhood (OR = 1.439, 95% CI = 1.305, 1.588). The mortality difference by widowhood status also occurred among black men (OR = 1.542, 95% CI = 1.250, 1.903) and, more noticeably, among Hispanic men (OR = 2.205, 95% CI = 1.609, 3.022). A comparison of the point estimate of odds ratios between Models 1 and 2 in Table 3 suggests that adding economic resources to the model, particularly income, reduced the effect of widowhood on mortality for white men and black men and, to a lesser extent, for Hispanic men; and income was more strongly associated with mortality among white men (OR = 0.782, 95% CI = 0.746, 0.819) than among black men (OR = 0.832, 95% CI = 0.792, 0.874) and Hispanic men (OR = 0.928, 95% CI = 0.870, 0.989) (Model 2 of Table 3).
Table 3.
Adjusted Odds Ratios of Mortality from Discrete-Time Hazard Models for Men by Race/ethnicity
| White Men N of respondents=10,941 N of person-periods=65,941 |
Black Men N of respondents=2,517 N of person-periods=12,136 |
Hispanic Men N of respondents=1,725 N of person-periods=8,515 |
||||
|---|---|---|---|---|---|---|
| Model 1 | Model 2 | Model 1 | Model 2 | Model 1 | Model 2 | |
| Widowed | 1.439*** (1.305 - 1.588) |
1.377*** (1.250 - 1.516) |
1.542*** (1.250 - 1.903) |
1.430** (1.154 - 1.773) |
2.205***
a (1.609 - 3.022) |
2.134***
a (1.554 - 2.930) |
| Remarried | 0.579*** (0.471 - 0.712) |
0.615*** (0.501 - 0.753) |
0.526** (0.353 - 0.783) |
0.571** (0.388 - 0.839) |
0.503 (0.234 - 1.083) |
0.528 (0.239 - 1.164) |
| Never married | 1.349** (1.116 - 1.632) |
1.144 (0.941 - 1.390) |
1.663** (1.158 - 2.391) |
1.421 (0.962 - 2.100) |
1.306 (0.611 - 2.790) |
1.234 (0.563 - 2.707) |
| Age | 1.104*** (1.100 - 1.108) |
1.102*** (1.097 - 1.106) |
1.075***
a (1.063 - 1.087) |
1.075***
a (1.063 - 1.087) |
1.098*** (1.086 - 1.111) |
1.098*** (1.085 - 1.110) |
| Born in South | 1.239*** (1.150 - 1.335) |
1.197*** (1.110 - 1.292) |
1.259* (1.034 - 1.532) |
1.252* (1.027 - 1.527) |
1.666**
b (1.177 - 2.358) |
1.686**
b (1.195 - 2.381) |
| Education (0=less than high school) | ||||||
| High school graduate | 0.776*** (0.731 - 0.822) |
0.832*** (0.783 - 0.884) |
0.828 (0.657 - 1.045) |
0.882 (0.703 - 1.107) |
1.024 (0.741 - 1.416) |
1.065 (0.780 - 1.454) |
| Some college | 0.702*** (0.639 - 0.771) |
0.793*** (0.719 - 0.874) |
0.552*** (0.405 - 0.754) |
0.612**
b (0.451 - 0.830) |
0.673 (0.445 - 1.016) |
0.713 (0.481 - 1.055) |
| College graduate | 0.464*** (0.420 - 0.511) |
0.593*** (0.539 - 0.653) |
0.545* (0.339 - 0.876) |
0.654 (0.408 - 1.049) |
0.618 (0.346 - 1.105) |
0.648 (0.348 - 1.207) |
| Income (logged) | 0.782*** (0.746 - 0.819) |
0.832***
b (0.792 - 0.874) |
0.928*
a (0.870 - 0.989) |
|||
| Wealth (logged) | 0.967 (0.845 - 1.107) |
1.109 (0.848 - 1.450) |
1.171 (0.806 - 1.703) |
|||
p < 0.001,
p < 0.01,
p < 0.05.
95% confidence intervals in parentheses.
Wald tests: significantly different from white men at the level of p < 0.05
Wald tests: marginally significantly different from white men at the level of p < 0.1
Table 4 shows the adjusted odds ratios for mortality from the discrete-time hazard models by race-ethnicity for women. Results of Model 1 in Table 4 suggest that being widowed was associated with significantly higher odds of death for women across all racial-ethnic groups, and the point estimate of mortality difference by widowhood status was greater among black women (OR = 1.525, 95% CI = 1.294, 1.798) and Hispanic women (OR = 1.551, 95% CI = 1.163, 2.069) than white women (OR = 1.292, 95% CI = 1.199, 1.393). A comparison of the point estimate of odds ratios between Models 1 and 2 in Table 4 suggests that adding economic resources, particularly income and wealth, reduced the effect of widowhood on mortality for both white and black women, but not for Hispanic women. Moreover, income and wealth were more strongly associated with mortality among white women than black women and Hispanic women (Model 2 of Table 4). Figure 1 summarizes the adjusted mortality odds ratios of widowhood by gender and race/ethnicity based on results from Model 2 of Tables 3 and 4. Figure 1 shows that widowhood is associated with the highest point estimate of mortality odds ratio among Hispanic men and lowest among white women.
Table 4.
Adjusted Odds Ratios of Mortality from Discrete-Time Hazard Models for Women by Race/ethnicity
| White Women N of respondents=13,573 N of person-periods=88,895 |
Black Women N of respondents=3,793 N of person-periods=20,747 |
Hispanic Women N of respondents=2,228 N of person-periods=12,236 |
||||
|---|---|---|---|---|---|---|
| Model 1 | Model 2 | Model 1 | Model 2 | Model 1 | Model 2 | |
| Widowed | 1.292*** (1.199 - 1.393) |
1.216*** (1.123 - 1.316) |
1.525***
b (1.294 - 1.798) |
1.460***
a (1.240 - 1.719) |
1.551** (1.163 - 2.069) |
1.605**
b (1.194 - 2.156) |
| Remarried | 0.698** (0.556 - 0.876) |
0.768* (0.609 - 0.968) |
0.925 (0.466 - 1.836) |
1.006 (0.509 - 1.990) |
0.351* (0.136 - 0.904) |
0.340* (0.132 - 0.881) |
| Never married | 1.517*** (1.291 - 1.782) |
1.362*** (1.164 - 1.593) |
1.308* (1.021 - 1.674) |
1.215 (0.953 - 1.548) |
1.027 (0.560 - 1.882) |
1.095 (0.593 - 2.024) |
| Age | 1.112*** (1.107 - 1.116) |
1.110*** (1.106 - 1.115) |
1.080***
a (1.073 - 1.088) |
1.080***
a (1.072 - 1.088) |
1.106*** (1.089 - 1.124) |
1.108*** (1.090 - 1.126) |
| Born in South | 1.113** (1.038 - 1.193) |
1.087* (1.012 - 1.166) |
1.136 (0.937 - 1.378) |
1.140 (0.940 - 1.383) |
1.331** (1.078 - 1.644) |
1.319**
b (1.072 - 1.622) |
| Education (0=less than high school) | ||||||
| High school graduate | 0.761*** (0.700 - 0.826) |
0.843*** (0.771 - 0.921) |
0.798** (0.689 - 0.923) |
0.839* (0.727 - 0.969) |
0.826 (0.582 - 1.172) |
0.814 (0.564 - 1.175) |
| Some college | 0.615*** (0.551 - 0.686) |
0.735*** (0.653 - 0.827) |
0.584*** (0.466 - 0.731) |
0.642*** (0.512 - 0.804) |
0.626 (0.368 - 1.064) |
0.609 (0.349 - 1.065) |
| College graduate | 0.474*** (0.429 - 0.524) |
0.642*** (0.578 - 0.714) |
0.648**
a (0.502 - 0.838) |
0.757 (0.570 - 1.006) |
1.176 a (0.519 - 2.665) |
1.145 (0.504 - 2.601) |
| Income (logged) | 0.782*** (0.757 - 0.808) |
0.868***
a (0.823 - 0.916) |
1.061 a (0.946 - 1.190) |
|||
| Wealth (logged) | 0.785*** (0.697 - 0.884) |
1.050 a (0.827 - 1.332) |
0.808 (0.546 - 1.195) |
|||
p < 0.001,
p < 0.01,
p < 0.05.
95% confidence intervals in parentheses.
Wald tests: significantly different from white women at the level of p < 0.05
Wald tests: marginally significantly different from white women at the level of p < 0.1
Figure 1.
Adjusted Mortality Odds Ratios of Widowhood with 95% Confidence Intervals (Shown as Error Bars) by Gender and Race/ethnicity (Based on Results from Model 2 of Tables 3 and 4)
We conducted formal mediation tests for economic resources and the results are shown in Table 5. These results suggest significant indirect effects of widowhood on mortality through economic resources for the total sample as well as for whites and blacks but not for Hispanics (both men and women). Specifically, the total effect (i.e., regression coefficient) of widowhood on mortality for the total sample was 0.346, of which about 17% (0.059) was through economic resources (both income and wealth, indicated in results from Table 2). The total effect of widowhood for white men was 0.366, of which about 13% (0.047) was through economic resources (mainly income, indicated in results from Table 3). The total effect of widowhood for black men was 0.439, of which about 18% (0.081) was through economic resources (mainly income, indicated in results from Table 3). Economic resources (both income and wealth, indicated in results from Table 4) explained 26% (i.e., 0.068/0.264) of the widowhood effect on mortality for white women and 13% (0.054/0.432) for black women. The indirect effect of widowhood on mortality risk via economic resources was not significant for either Hispanic men or Hispanic women.
Table 5.
KHB Mediation Analysis for Economic Resources in Widowhood Effects on Mortality
| Total Sample |
||||||
|---|---|---|---|---|---|---|
| Total effect | 0.346 | *** | ||||
| Direct effect | 0.287 | *** | ||||
| Indirect effect | 0.059 | *** | ||||
| White men |
Black men |
Hispanic men |
||||
| Total effect | 0.366 | *** | 0.439 | *** | 0.790 | *** |
| Direct effect | 0.320 | *** | 0.358 | ** | 0.758 | *** |
| Indirect effect | 0.047 | 0.081 | ** | 0.032 | ||
| White women |
Black women |
Hispanic women |
||||
| Total effect | 0.264 | *** | 0.432 | *** | 0.440 | ** |
| Direct effect | 0.196 | *** | 0.379 | *** | 0.473 | ** |
| Indirect effect | 0.068 | *** | 0.054 | *** | −0.033 | |
p < 0.001,
p < 0.01,
p < 0.05
Discussion
There were more than 15 million widowed persons in the United States in 2017, and nearly 12 million were women [28]. Researchers have long recognized widowhood as one of the most traumatic life events people experience in terms of its impact on health and mortality with varying effects by gender [18,29,30]. Evidence on potential racial-ethnic variation in widowhood and mortality has been much more limited. This is important because blacks become widowed at younger ages than whites and, proportionately, more Blacks are widowed at all ages [4]. By age 65-74, 24.3% of blacks are widowed, compared to only 14.8% of whites [28]. Thus, the widowed are much more likely to be female and/or black than to be male and/or white. The higher prevalence of widowhood among blacks and women compared to whites and men leads to “double jeopardy” of disadvantaged groups, which may increase mortality risk following widowhood and contribute to mortality disparities. Moreover, studies on widowhood typically do not compare non-Hispanic whites and blacks to Hispanics. Although the proportion of widowhood is relatively lower among Hispanics than whites in the United States [31], the economic disadvantage of Hispanic populations along with the potentially more salient meaning of marriage attached to Hispanic familism values may make them more vulnerable to the consequences of widowhood.
Our results support this hypothesis showing that widowhood is associated with increased mortality risk across all gender and racial-ethnic subgroups but the effect tends to be stronger for racial-ethnic minority men and women than their white counterparts. This finding is in contrast to an earlier study by Elwert and Christakis (2006) suggesting that the widowhood effect on mortality was stronger among whites than blacks [7]. Notably, Hispanics were not distinguished in the Elwert and Christakis study, and their sample was restricted to ages above 65 [7]. It is likely that their findings concerning race reflect selection bias associated with the older age of their sample compared with that used for the present study: we analyzed an HRS sample of ages above 50. Our sensitivity analysis with a sample restricted to those aged above 65 at the last survey suggested that the widowhood effect on mortality, especially among black men and black women, was smaller among those ages 65+ (results shown in Appendix A) than in the total analytic sample of ages 50+ (results shown in Tables 3 and 4). Results in Appendix A suggest that the effect of widowhood on mortality was not significant among black men ages 65 and above after all covariates were controlled, which is indeed consistent with the Elwert and Christakis’ finding [7]. Given that blacks become widowed at much younger ages than whites [4], it is critical to include a wider age range in order to capture racial differences in widowhood effects which may start and peak much earlier than age 65 for black Americans.
Although Hispanics represent a growing segment of the older US population, and Hispanics experience greater socioeconomic disadvantage compared to non-Hispanic whites, prior studies have largely neglected to consider the potentially distinguishable effect of widowhood on mortality among Hispanics compared to other racial-ethnic groups. We provide some of the first evidence that the effect of widowhood on increased mortality risk is greater for Hispanic men than for all other gender and racial-ethnic subgroups: the mortality risk of Hispanic men who were widowed was more than twice that of Hispanic men who did not experience widowhood. Previous studies suggest that Hispanic families are characterized by cultural values of “familism” that emphasize strong commitment to marital and family life—in ways that are distinct from non-Hispanic white or black populations [31–33], potentially making Hispanics more vulnerable in suffering consequences of the loss of a spouse. This may be especially true for older Hispanic men who are highly likely to be in a traditional marriage where husbands receive health benefits from their wives’ caregiving and emotional support [22,34,35]. Future studies should investigate the unique life experiences of widowhood among racial-ethnic minority men and women, in particular Hispanic men, that may result in elevated mortality risk.
A frequently proposed mechanism through which widowhood shapes mortality risk is change in economic resources following the loss of a spouse [9,10,13,23]. Our results support this hypothesis among both whites and blacks, men and women, but not among Hispanic men or Hispanic women. Increased economic vulnerability and financial strain following widowhood are often suggested as undermining health and longevity of women [13,22,24,36]. Indeed, we found that economic resources played a more important role in explaining the widowhood effect on mortality risk for white women than other groups—this might be related to the relatively higher earnings of white husbands compared with racial-ethnic minority husbands. Our results further suggest that loss of economic resources explains some of the widowhood effect on mortality for white men and black men as well as for their women counterparts. With women’s increased labor force participation, their income becomes increasingly important for their family, and loss of this income may contribute to their husband’s mortality risk following widowhood. Surprisingly, we found that economic resources are less important for mortality following widowhood among Hispanic men or women. Although widowhood leads to economic hardship for all racial-ethnic groups [24], for Hispanics, economic hardship following widowhood does not seem to be a key factor contributing to increased mortality risk. This is consistent with the long-documented “Hispanic Paradox” which reflects a weaker association between socioeconomic status and mortality among Hispanic populations than other racial-ethnic groups [37]. Future studies should investigate other social and psychological burdens of widowhood (e.g., loss of social support, emotional stress) that may explain widowhood effects on mortality among Hispanics.
This study has several limitations. First, the analytic sample was relatively small for some racial-ethnic minority groups such as widowed Hispanic men (unweighted N=201), which may explain some of the non-significant findings. Indeed, the Hispanic population is highly heterogeneous. The relatively small size of Hispanics in our sample further limits our ability to distinguish among Hispanic subgroups. Future studies should analyze other datasets with a larger Hispanic sample to further investigate the widowhood effects across Hispanic subgroups. Moreover, we are not able to examine a full range of potential mediators linking widowhood to mortality due to data limitations. Future studies should continue to investigate other potential pathways (e.g., loss of social support) through which widowhood shapes the risk of mortality in potentially different ways for men and women and across diverse racial and ethnic groups
Conclusions
This is one of the first nationally-representative longitudinal studies to examine widowhood effects on mortality at the intersection of gender and race-ethnicity. We provide the first evidence that Hispanic men are at greater risk for mortality following widowhood than all other gender and racial-ethnic subgroups. Moreover, for non-Hispanic whites and blacks, loss of economic resources following widowhood is a pathway through which widowhood increases mortality risk. Yet, this is not a key explanatory pathway for Hispanics. Our findings highlight that gender and race-ethnicity intersect with widowhood status to disadvantage some groups more than others by increasing mortality risk [38–43]. It is important to further explore the complex pathways that might contribute to higher mortality risk of racial-ethnic minorities, especially Hispanic men, following widowhood so that effective interventions can be implemented to reduce those risks.
Acknowledgements
This research was supported by the U.S. National Institute on Aging, Grants R01AG054624, R01AG061118, and K01AG043417 and by the U.S. Eunice Kennedy Shriver National Institute of Child Health and Human Development, Grant P2CHD042849.
List of abbreviations:
- HRS
Health and Retirement Study
- NLMS
National Longitudinal Mortality Study
- NHIS
National Health Interview Surveys
- Hispanic EPESE
Hispanic Established Populations for the Epidemiologic Studies of the Elderly
- KHB
Karlson–Holm–Breen
Appendix A
Table S1.
Adjusted Odds Ratios of Mortality from Discrete-Time Hazard Models by Gender and Race/ethnicity, Ages 65+
| White men N of respondents=7,560 N of person-periods=51,384 |
Black men N of respondents=1,286 N of person-periods=7,869 |
Hispanic men N of respondents=786 N of person-periods=5,181 |
White women N of respondents=9,516 N of person-periods=68,690 |
Black women N of respondents=1,898 N of person-periods=13,341 |
Hispanic women N of respondents=980 N of person-periods=7,258 |
|
|---|---|---|---|---|---|---|
| Widowed | 1 237*** (1.097 - 1.396) |
1.287 (0.948 - 1.748) |
2.075*** (1.521 - 2.830) |
1.166*** (1.079 - 1.260) |
1.256* (1.059 - 1.490) |
1.399* (1.004 - 1.950) |
| Remarried | 0.555*** (0.425 - 0.725) |
0.636 (0.396 - 1.020) |
0.441 (0.176 - 1.104) |
0.796 (0.613 - 1.033) |
0.826 (0.291 - 2.341) |
0.345 (0.110 - 1.079) |
| Never married | 1.437** (1.139 - 1.815) |
1.093 (0.614 - 1.944) |
0.948 (0.335 - 2.681) |
1.498*** (1.231 - 1.822) |
1.246 (0.808 - 1.921) |
1.536 (0.736 - 3.207) |
| Age | 1.130*** (1.126 - 1.134) |
1.108*** (1.093 - 1.124) |
1.113*** (1.099 - 1.127) |
1.127*** (1.122 - 1.132) |
1.117*** (1.108 - 1.125) |
1.125*** (1.101 - 1.149) |
| Born in South | 1.174*** (1.076 - 1.281) |
1.092 (0.846 - 1.408) |
1.498* (1.083 - 2.073) |
1.121** (1.041 - 1.208) |
0.884 (0.700 - 1.117) |
1.308* (1.068 - 1.602) |
| Education (0=less than high school) | ||||||
| High school graduate | 0.873** (0.797 - 0.957) |
0.846 (0.616 - 1.162) |
1.047 (0.737 - 1.487) |
0.862*** (0.792 - 0.938) |
0.708*** (0.590 - 0.850) |
0.826 (0.512 - 1.332) |
| Some college | 0.805*** (0.721 - 0.898) |
0.554* (0.352 - 0.872) |
0.864 (0.484 - 1.540) |
0.763*** (0.683 - 0.852) |
0.593*** (0.442 - 0.795) |
0.814 (0.435 - 1.521) |
| College graduate | 0.640*** (0.574 - 0.713) |
0.594* (0.397 - 0.890) |
0.889 (0.380 - 2.081) |
0.734*** (0.650 - 0.830) |
0.893 (0.608 - 1.312) |
1.306 (0.575 - 2.965) |
| Income (logged) | 0.776*** (0.735 - 0.819) |
0.771*** (0.711 - 0.836) |
0.898* (0.826 - 0.977) |
0.796*** (0.762 - 0.832) |
0.847*** (0.777 - 0.924) |
1.105 (0.955 - 1.278) |
| Wealth (logged) | 0.923 (0.797 - 1.070) |
1.353 (0.962 - 1.901) |
1.131 (0.738 - 1.734) |
0.757*** (0.670 - 0.857) |
0.890 (0.689 - 1.149) |
0.724 (0.446 - 1.176) |
p < 0.001,
p < 0.01,
p < 0.05.
95% confidence intervals in parentheses.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Contributor Information
Hui Liu, Department of Sociology, Michigan State University, 509 E Circle Dr. 316 Berkey Hall, East Lansing, MI 48823.
Debra Umberson, Department of Sociology and Population Research Center, The University of Texas at Austin, Austin, Texas 78712.
Minle Xu, Population Research Center, The University of Texas at Austin, Austin, Texas 78712.
References
- 1.Waite LJ. Does marriage matter?. Demography 1995;32(4):483–507. [PubMed] [Google Scholar]
- 2.Hughes ME, Waite LJ. Marital biography and health at mid-life. J Health Soc Behav 2009;50(3):344–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Liu H, Umberson DJ. The times they are a changin’: Marital status and health differentials from 1972 to 2003. J Health Soc Behav 2008;49(3):239–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Umberson DJ, Olson JS, Crosnoe R, Liu H, Pudrovska T, Donnelly R. Death of family members as an overlooked source of racial disadvantage in the United States. Proceedings of the National Academy of Sciences 2017;114(5):915–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Koball HL, Moiduddin E, Henderson J, Goesling B, Besculides M. What Do We Know about the Link between Marriage and Health?. J Fam Issues 2010;31:1019–1040. [Google Scholar]
- 6.Wood RG, Goesling B, Avellar S. The effects of marriage on health: A synthesis of recent research evidence. Princeton, NJ: Mathematica Policy Research, Inc. 2007. [Google Scholar]
- 7.Elwert F, Christakis NA. Widowhood and race. Am Sociol Rev 2006;71(1):16–41. [Google Scholar]
- 8.Greenman E, Xie Y. Double jeopardy? The interaction of gender and race on earnings in the United States. Soc Forces 2008;86(3):1217–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Prigerson HG, Maciejewski PK, Rosenheck RA. Preliminary explorations of the harmful interactive effects of widowhood and marital harmony on health, health service use, and health care Costs1. The Gerontologist. 2000. June 1;40(3):349–57. [DOI] [PubMed] [Google Scholar]
- 10.Wilcox S, Evenson KR, Aragaki A, Wassertheil-Smoller S, Mouton CP, Loevinger BL. The effects of widowhood on physical and mental health, health behaviors, and health outcomes: The Women’s Health Initiative. Health Psychol 2003;22(5):513. [DOI] [PubMed] [Google Scholar]
- 11.Liu H, Zhang Z, Choi S, Langa KM. Marital status and dementia: Evidence from the Health and Retirement Study, The Journals of Gerontology: Series B 2019; gbz087; 10.1093/geronb/gbz087 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Sullivan AR, Fenelon A. Patterns of widowhood mortality. J Gerontol B-Psychol 2013;69(1):53–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Lillard LA, Waite LJ. ‘Til death do us part: Marital disruption and mortality. Am J Sociol 1995;100(5):1131–56. [Google Scholar]
- 14.Hu Y, Goldman N. Mortality differentials by marital status: an international comparison. Demography 1990;27(2):233–50. [PubMed] [Google Scholar]
- 15.Schaefer C, Quesenberry CP Jr, Wi S. Mortality following conjugal bereavement and the effects of a shared environment. Am J Epidemiol 1995;141(12):1142–52. [DOI] [PubMed] [Google Scholar]
- 16.Martikainen P, Valkonen T. Mortality after the death of a spouse: rates and causes of death in a large Finnish cohort. Am J Public Health 1996;86(8_Pt_1):1087–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Moon JR, Kondo N, Glymour MM, Subramanian SV. Widowhood and mortality: a meta-analysis. PloS one 2011;6(8):e23465. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Manor O, Eisenbach Z. Mortality after spousal loss: are there socio-demographic differences?. Soc Sci Med 2003;56(2):405–13. [DOI] [PubMed] [Google Scholar]
- 19.Elwert F, Christakis NA. The effect of widowhood on mortality by the causes of death of both spouses. Am J Public Health 2008;98(11):2092–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Johnson NJ, Backlund E, Sorlie PD, Loveless CA. Marital status and mortality: the national longitudinal mortality study. Ann Epidemiol 2000;10(4):224–38. [DOI] [PubMed] [Google Scholar]
- 21.Liu H Till death do us part: marital status and U.S. mortality trends, 1986–2000. J Marriage Fam 2009;71(5):1158–1173. [Google Scholar]
- 22.Stimpson JP, Kuo YF, Ray LA, Raji MA, Peek MK. Risk of mortality related to widowhood in older Mexican Americans. Ann Epidemiol 2007;17(4):313–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Luoma JB, Pearson JL. Suicide and marital status in the United States, 1991–1996: is widowhood a risk factor?. Am J Public Health 2002;92(9):1518–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Angel JL, Jimenez MA, Angel RJ. The economic consequences of widowhood for older minority women. The Gerontologist 2007;47(2):224–34. [DOI] [PubMed] [Google Scholar]
- 25.StataCorp. Stata Statistical Software: Release 15. 2017. College Station, TX: StataCorp LLC. [Google Scholar]
- 26.Kohler U, Karlson KB, Holm A. Comparing coefficients of nested nonlinear probability models. The Stata Journal 2011;11(3):420–38. [Google Scholar]
- 27.Agresti A, Finlay B. Statistical methods for the social sciences. Upper Saddle River, N.J: Pearson Prentice Hall; 2009. [Google Scholar]
- 28.U.S. Census Bureau. America’s Families and Living Arrangements: 2017. 2017. https://www.census.gov/data/tables/2017/demo/families/cps-2017.html. [Accessed 04 Nov 2019]
- 29.Hagedoorn M, Van Yperen NW, Coyne JC, van Jaarsveld CH, Ranchor AV, van Sonderen E, Sanderman R. Does marriage protect older people from distress? The role of equity and recency of bereavement. Psychol Aging 2006;21(3):611–620. [DOI] [PubMed] [Google Scholar]
- 30.Williams DR, Takeuchi DT, Adair RK. Marital status and psychiatric disorders among blacks and whites. J Health Soc Behav 1992;33(2):140. [PubMed] [Google Scholar]
- 31.Oropesa RS, Landale NS. The future of marriage and Hispanics. J Marriage Fam 2004;66(4): 901–20. [Google Scholar]
- 32.Vega WA. The study of Latino families In: Zambrana R, editor. Understanding Latino families: Scholarship, policy, and practice. Thousand Oaks, CA: Sage; 1995. pp. 3–17. [Google Scholar]
- 33.Landale NS, Oropesa RS, Bradatan C. Hispanic Families in the United States: Family Structure and Process in an Era of Family Change In: Tienda M, Mitchell F, editors. Hispanics and the Future of America. Washington DC: National Academies Press (US) 2006. pp. 138–178. [Google Scholar]
- 34.Umberson D Gender, marital status and the social control of health behavior. Soc Sci Med 1992;34(8):907–17. [DOI] [PubMed] [Google Scholar]
- 35.Revenson TA, Griva K, Luszczynska A, Morrison V, Panagopoulou E, Vilchinsky N, Hagedoorn M. Gender and caregiving: The costs of caregiving for women In: Revenson TA, Griva K, Luszczynska A, Morrison V, Panagopoulou E, Vilchinsky N, Hagedoorn M, editors Caregiving in the illness context. London: Palgrave Pivot; 2016. pp. 48–63. [Google Scholar]
- 36.Utz RL. Economic and practical adjustments to late life spousal loss In: Carr D, Nesse RM, Wortman CB, editors. Spousal Bereavement in Late Life. New York: Springer Publishing Company; 2006. pp. 167–192. [Google Scholar]
- 37.Turra CM, Goldman N. Socioeconomic differences in mortality among US adults: Insights into the Hispanic paradox. J Gerontol B-Psychol 2007;62(3):S184–92. [DOI] [PubMed] [Google Scholar]
- 38.Crosnoe R, Elder GH Jr. From childhood to the later years: Pathways of human development. Res Aging. 2004;26(6):623–54. [Google Scholar]
- 39.Palloni A Childhood health and the reproduction of inequalities. Demography. 2006;43(4): 587–615. [DOI] [PubMed] [Google Scholar]
- 40.Pearlin LI, Schieman S, Fazio EM, Meersman SC. Stress, health, and the life course: Some conceptual perspectives. J Health Soc Behav 2005;46(2):205–19. [DOI] [PubMed] [Google Scholar]
- 41.George LK. Sociological perspectives on life transitions. Annu Rev Sociol 1993; 19(1):353–73. [Google Scholar]
- 42.Schnittker J, McLeod JD. The social psychology of health disparities. Annu Rev Sociol 2005;31:75–103. [Google Scholar]
- 43.House JS. Understanding social factors and inequalities in health: 20th century progress and 21st century prospects. J Health Soc Behav 2002:125–42. [PubMed] [Google Scholar]