https://orcid.org/0000-0001-9471-6373
Abstract
Women tend to ruminate more than men, and are generally more hypervigilant to the emotions of others in order to maintain positive social ties. Thus, compared to men, women may ruminate more when their social partners have greater life stresses. However, the literature on stressful events typically focuses on individuals’ experiences and perceptions of stressors experienced by specific social ties such as spousal partners and adult children. The purpose of this study was to examine links between perceptions of a broad array of family and nonfamily social partner stresses and daily rumination among older men and women.
Adults aged 65 and older (N = 293, 55% women) completed baseline assessments of family and nonfamily life stressors and 5–6 consecutive nightly assessments regarding rumination, interpersonal tensions, worries, and support provision.
Multilevel structural equation models revealed that perceptions of greater family and nonfamily life stressors were associated with greater rumination. The links between family stress and rumination varied by gender: family stress was related to greater rumination among women and not men. Moreover, among women, family and nonfamily stress–rumination links were accounted for by greater daily worries about others, and among men, the nonfamily stress–rumination link was due to greater interpersonal tensions as well as daily worries.
These findings may be due in part to gender role socialization and women’s greater kin-keeping and investment in family ties.
Rumination, which involves perseverative thoughts regarding symptoms and causes of distress, is considered a key factor in the development of depression and anxiety (Nolen-Hoeksema et al., 2008; Ruscio et al., 2015) and is associated with poor psychological well-being (Brosschot et al., 2006; Lyubomirsky et al., 2015; Newman & Nezlek, 2019). In particular, individuals who ruminate about stressful events are more likely to experience adverse mental health consequences than those who do not ruminate about these events (McIntosh et al., 2010; Zawadski, 2015). The literature on stressful events has typically focused on the stresses experienced by individuals and not perceptions of the stresses their close social partners have experienced. Yet, we know from studies of specific social ties (e.g., grown child, parent, spouse) that perceptions of social partners’ problems predict poorer well-being (Birditt et al., 2016a; Huo et al., 2019; Kim et al., 2017; Wang et al., 2020). Additionally, the majority of research has focused on specific family ties and has not considered stressors among the full array of family and nonfamily ties. Problems experienced by family may be more stressful than nonfamily due to greater feelings of obligation to provide support. Further, problems experienced by family members may be more likely to be issues that are experienced by both individuals and their social partners compared to nonfamily problems (e.g., death of a relative; financial problems).
The links between social partner stresses and rumination may vary by gender and age. Women tend to ruminate more than men (Hilt et al., 2010; Johnson & Whisman, 2013; Nolen-Hoeksema & Aldao, 2011; Nolen-Hoeksema & Jackson, 2001) and are generally more hypervigilant to the emotions of others in order to maintain positive social ties (Gilligan, 1982; Helgeson, 1994). As such, compared to men, women may be more negatively affected by their social partners’ life stresses. Further, these links may be smaller among the oldest-old compared to young-old adults due to age-related declines in negative affective experiences and increased tendencies to attend to and remember more positive than negative stimuli (Carstensen, 2021; Charles, 2010; Isaacowitz, 2022). The purpose of the present study was to examine associations between perceptions of family and nonfamily life stressors and daily rumination among older men and women, and whether those links vary by gender and age.
Theoretical Background
This study draws on three complementary theoretical frameworks: (1) Anderson’s female vulnerability to network events hypothesis (Anderson et al., 2022), (2) Nolen-Hoeksema’s theory of rumination (Nolen-Hoeksema & Jackson, 2001), and (3) the strength and vulnerability integration model (Charles, 2010).
Anderson et al. (2022) developed what they refer to as the female vulnerability to network events hypothesis. This hypothesis suggests that women are more vulnerable than men to stressful events experienced by close social ties due to gender role socialization, whereby women are socialized to be relationship-oriented and provide empathy and support to social partners when they are in need. This is similar to Kessler and McLeod’s (1984) cost of caring hypothesis, which suggests that women are more emotionally involved in the lives of others than are men. Similarly, Kiecolt-Glaser and Newton (2001) proposed that women have more relationally interdependent self-representations: relative to men, women tend to view themselves in terms of their connections with others and are more invested in maintaining relationships. Women often provide more emotional support compared with men (Kahn et al., 2011).
Because women are more vulnerable to network events (Anderson et al., 2022), they may be more likely to ruminate when their close social partners experience problems. Indeed, according to Nolen-Hoeksema’s theory of rumination (Nolen-Hoeksema & Jackson, 2001), one reason women are more likely than men to ruminate is due to their greater feelings of responsibility for the emotional tone of relationships. Thus, we predicted that women would ruminate more than men, and that because women are more interpersonally focused, women would ruminate more when they perceived others as having more life stress.
It is particularly important to examine how close partner problems affect rumination among older adults. Socioemotional selectivity theory (Carstensen, 2021) suggests that, as individuals age, they become more invested in maintaining emotionally close relationships and show reductions in negative affective experiences and increases in positive affective experiences. The strength and vulnerability integration model (Charles, 2010) further suggests, however, that these age-related improvements are less likely to occur when individuals are under chronic stress. These age-related improvements may continue across older adulthood but the findings are inconsistent. For instance, oldest-old adults show greater reactivity to interpersonal tensions but less reactivity to tensions that are avoided compared to younger adults (e.g., young-old and middle-aged adults; Birditt, 2014; Birditt et al., 2020). Thus, based on the literature and theories, we predicted that oldest-old adults would be less negatively affected by stress than young-old adults (showing a smaller association among social partner stress and rumination) but that this age difference would be smaller among individuals reporting high levels of chronic stress (i.e., when stress cannot be avoided).
This study considers individuals’ reports of stresses experienced by their family and nonfamily social partners and implications of these social partners’ stresses for daily rumination. Life stresses experienced by one’s close social partners can include problems with health, finances, and death, among others. We then ask whether the association between social partner stress and rumination is accounted for by daily interpersonal tensions, daily support provision, and daily interpersonal worries. We predict that greater perceptions of family and nonfamily stress will be associated with greater daily rumination, and that these links will be accounted for by three daily processes (i.e., daily interpersonal stresses, daily support exchange, and daily interpersonal worries); additionally, we predict that these links will be stronger among family than nonfamily ties.
Links Between Life Stressors and Rumination
Life stressors are often associated with greater rumination. For instance, Nolen-Hoeksema and colleagues (1999) found a bidirectional relationship between chronic stress and rumination: chronic strain (e.g., lack of affirmation in close relationships, role burden) predicted more rumination over time, and increased rumination predicted chronic strain over time. Further, Ruscio et al. (2015) used ecological momentary assessments (EMAs) to assess rumination regarding stressful life events and found that people with diagnoses of major depressive disorder and/or generalized anxiety disorder were more likely to ruminate about stressful events than individuals with no diagnosed psychopathology. King and Delongis (2014) found that paramedics’ greater work stress was associated with their greater rumination.
Studies have found links between social partner stress and well-being. For example, parents who report that their adult children have problems (e.g., emotional or financial problems) report more interpersonal tensions and poorer daily well-being (Birditt et al., 2016a; Huo et al., 2019). Midlife adults who worry about or support parents who experience life crises or disability report poorer relationship quality, poorer sleep, and diminished well-being (Kim et al., 2017; Wang et al., 2020). Comparisons of men and women indicate that women often report providing more support (Fingerman et al., 2020) but the well-being of both men and women is similarly affected by problems of their children and their parents (Huo et al., 2019; Wang et al., 2020).
Research on marriage has also found important cross-partner links between stress and well-being. Birditt et al. (2016b) found that wife reports of greater chronic stress predicted increased blood pressure among husbands. Monin et al. (2010) found that older married couples were reactive to their spouses’ pain: individuals who watched their spouse complete a painful task or discussed their spouse’s suffering showed increased heart rate and blood pressure compared to exposure to a stranger’s suffering. Wilson et al. (2023) found that spousal disclosure of upsetting events with greater emotional intensity predicted greater pro-inflammatory gene expression. Anderson et al. (2022) examined life event stresses of panel members and members of their household ages 25–64 in the UK Household Longitudinal Survey and links with mental health. They found that greater perceptions of one’s own life stress were associated with poorer well-being among both men and women. Further, they found that household member stress was associated with a 50% increase in depressive symptoms among women compared to men. Thus, the present research suggests that close social partner problems are associated with well-being, especially among women. There is a lack of research, however, on the links between family and nonfamily social partner problems and rumination.
The Present Study
The majority of research on life stressors and rumination has examined perceptions of one’s own life stress and/or the problems of close others, such as adult children and spouses. This study moves beyond the previous literature by considering individuals’ perceptions of life stresses experienced by a fuller array of close friends or relatives and nonclose social partners. We examine perceived family and nonfamily social partners’ stresses associated with daily rumination, and the potential mechanisms accounting for those effects, including interpersonal tensions, relationship worries, and support exchanges among retired adults aged 65 and older. Because individuals may also be experiencing the same stress as their social partners, we control for individual reports of their own stress in our analyses. Covariates for the study also included demographics, and contextual factors associated with stress and rumination: age, gender, Hispanic ethnicity, social network size, and previous day rumination (Birditt, 2014; Gerich, 2014; Johnson & Whisman, 2013; King & DeLongis, 2014; Russell & Taylor, 2009). We addressed three research questions:
(1) Are perceptions of family and nonfamily social partner life stress associated with greater daily rumination?
(2) Do daily interpersonal tensions, worries, and support provision account for the effects of family and nonfamily life stress on rumination?
(3) Do the links vary by gender and age?
We hypothesized that individuals who report greater perceived family and nonfamily life stresses will ruminate more in daily life. We expected that greater perceived family life stress would be more highly associated with rumination than nonfamily life stress. We expected that these associations between perceived social partner life stresses and rumination would be due to greater interpersonal tensions, worries, and support provision with family members. We also hypothesized that the associations would be greater among women than among men. Finally, we hypothesized that oldest-old respondents would show smaller associations among perceived social partner life stress and rumination than young-old adults.
Method
Participants
Participants were drawn from the Daily Experiences and Well-being Study (DEWS), which sampled individuals aged 65 and older who worked fewer than 20 hr a week from the Austin, Texas area (see Birditt et al., 2019, for a description of the sample and study design). Individuals were contacted using landline telephone lists; a total of 66% of participants who were contacted agreed to participate. The sample for analysis included 293 of the total sample (N = 333), who completed at least one evening survey (M = 4.00; standard deviation [SD] = 1.24; range = 1–6). This selection process has been used in all prior publications using these data (Birditt et al., 2019; Fingerman et al., 2020; Huo et al., 2019). Post hoc analyses considered whether controlling for the number of evening surveys completed affected the results and are presented in Supplementary Material. Fifty-five percent (55%) of the sample was female and participants ranged in age from 65 to 90 years old (M = 73.73, SD = 6.28). Eighty-three percent of participants were White (15% Hispanic), 15% were Black (4% Hispanic), 1% were American Indian or Alaska Native (50% Hispanic), and 1% were Asian (0% Hispanic); 94% had a high school degree or higher. Respondents reported relatively good health and were living in the community. See Table 1 for the sample description.
Daily Experience and Well-being Study Sample Descriptives by Gender (N = 293)
| Variables | Overall sample (N = 293) | Women (n = 161) | Men (n = 132) | t test | |||
|---|---|---|---|---|---|---|---|
| % | M (SD) | % | M (SD) | % | M (SD) | Sig. | |
| % Hispanic ethnicity | 14.7% | 14.9% | 14.4% | ||||
| Age | 73.73 (6.28) | 73.48 (6.47) | 74.03 (6.05) | ||||
| Social network sizea | 15.39 (6.89) | 17.16 (6.59) | 13.24 (6.66) | *** | |||
| Own life stressb | 1.02 (1.03) | 1.12 (1.06) | 0.90 (0.98) | ||||
| Family life stressb | 1.54 (1.37) | 1.72 (1.42) | 1.31 (1.28) | * | |||
| Nonfamily life stressb | 0.98 (1.19) | 1.16 (1.14) | 0.76 (1.23) | * | |||
| Interpersonal tensionsc | 16.8% | 19.8% | 13.1% | ** | |||
| Worriesd | 34.6% | 37.9% | 30.4% | ** | |||
| Support exchangese | 1.34 (1.07) | 1.38 (1.05) | 1.30 (1.08) | ||||
| Ruminationf | 1.66 (0.75) | 1.68 (0.78) | 1.64 (0.71) | ||||
| Variables | Overall sample (N = 293) | Women (n = 161) | Men (n = 132) | t test | |||
|---|---|---|---|---|---|---|---|
| % | M (SD) | % | M (SD) | % | M (SD) | Sig. | |
| % Hispanic ethnicity | 14.7% | 14.9% | 14.4% | ||||
| Age | 73.73 (6.28) | 73.48 (6.47) | 74.03 (6.05) | ||||
| Social network sizea | 15.39 (6.89) | 17.16 (6.59) | 13.24 (6.66) | *** | |||
| Own life stressb | 1.02 (1.03) | 1.12 (1.06) | 0.90 (0.98) | ||||
| Family life stressb | 1.54 (1.37) | 1.72 (1.42) | 1.31 (1.28) | * | |||
| Nonfamily life stressb | 0.98 (1.19) | 1.16 (1.14) | 0.76 (1.23) | * | |||
| Interpersonal tensionsc | 16.8% | 19.8% | 13.1% | ** | |||
| Worriesd | 34.6% | 37.9% | 30.4% | ** | |||
| Support exchangese | 1.34 (1.07) | 1.38 (1.05) | 1.30 (1.08) | ||||
| Ruminationf | 1.66 (0.75) | 1.68 (0.78) | 1.64 (0.71) | ||||
Notes: SD = standard deviation; Sig. = significance.
aSocial network size included the total number of network members reported.
bOwn stress (range 0–4) and family and nonfamily stress (range 0–6) included the count of life stressors.
cInterpersonal tensions includes % days with at least one interpersonal tension.
dWorries include % days reported worries.
eSupport exchanges ranged from 0 (None) to 3.
fRumination ranged from 1 (Not at all) to 5 (A great deal).
***p < .001. ** p < .01. * p < .05.
Daily Experience and Well-being Study Sample Descriptives by Gender (N = 293)
| Variables | Overall sample (N = 293) | Women (n = 161) | Men (n = 132) | t test | |||
|---|---|---|---|---|---|---|---|
| % | M (SD) | % | M (SD) | % | M (SD) | Sig. | |
| % Hispanic ethnicity | 14.7% | 14.9% | 14.4% | ||||
| Age | 73.73 (6.28) | 73.48 (6.47) | 74.03 (6.05) | ||||
| Social network sizea | 15.39 (6.89) | 17.16 (6.59) | 13.24 (6.66) | *** | |||
| Own life stressb | 1.02 (1.03) | 1.12 (1.06) | 0.90 (0.98) | ||||
| Family life stressb | 1.54 (1.37) | 1.72 (1.42) | 1.31 (1.28) | * | |||
| Nonfamily life stressb | 0.98 (1.19) | 1.16 (1.14) | 0.76 (1.23) | * | |||
| Interpersonal tensionsc | 16.8% | 19.8% | 13.1% | ** | |||
| Worriesd | 34.6% | 37.9% | 30.4% | ** | |||
| Support exchangese | 1.34 (1.07) | 1.38 (1.05) | 1.30 (1.08) | ||||
| Ruminationf | 1.66 (0.75) | 1.68 (0.78) | 1.64 (0.71) | ||||
| Variables | Overall sample (N = 293) | Women (n = 161) | Men (n = 132) | t test | |||
|---|---|---|---|---|---|---|---|
| % | M (SD) | % | M (SD) | % | M (SD) | Sig. | |
| % Hispanic ethnicity | 14.7% | 14.9% | 14.4% | ||||
| Age | 73.73 (6.28) | 73.48 (6.47) | 74.03 (6.05) | ||||
| Social network sizea | 15.39 (6.89) | 17.16 (6.59) | 13.24 (6.66) | *** | |||
| Own life stressb | 1.02 (1.03) | 1.12 (1.06) | 0.90 (0.98) | ||||
| Family life stressb | 1.54 (1.37) | 1.72 (1.42) | 1.31 (1.28) | * | |||
| Nonfamily life stressb | 0.98 (1.19) | 1.16 (1.14) | 0.76 (1.23) | * | |||
| Interpersonal tensionsc | 16.8% | 19.8% | 13.1% | ** | |||
| Worriesd | 34.6% | 37.9% | 30.4% | ** | |||
| Support exchangese | 1.34 (1.07) | 1.38 (1.05) | 1.30 (1.08) | ||||
| Ruminationf | 1.66 (0.75) | 1.68 (0.78) | 1.64 (0.71) | ||||
Notes: SD = standard deviation; Sig. = significance.
aSocial network size included the total number of network members reported.
bOwn stress (range 0–4) and family and nonfamily stress (range 0–6) included the count of life stressors.
cInterpersonal tensions includes % days with at least one interpersonal tension.
dWorries include % days reported worries.
eSupport exchanges ranged from 0 (None) to 3.
fRumination ranged from 1 (Not at all) to 5 (A great deal).
***p < .001. ** p < .01. * p < .05.
Compared to those included in the sample, those who were less likely to complete an evening survey—and were thus excluded—were older (b = −0.08, p < .01) and identified as Hispanic (b = −1.11, p < .05). While this study uses the same sample as in previous studies (e.g., Birditt et al., 2019, 2020), the measures and research questions are different. Specifically, Birditt et al. (2019) examined links between daily solitude and well-being; and Birditt et al. (2020) examined links between interpersonal tensions and well-being, and whether those links were moderated by age and coping strategy preferences. The current study included assessments of close partner life stress and rumination to test distinct research questions regarding links between family and nonfamily life stress and rumination, and gender and age differences in those links.
Procedure
Participation in the study included a face-to-face baseline interview followed by 5–6 days (three weekdays and two weekend days) of EMA questionnaires every 3 hr from waking until bedtime (approximately six times) on a mobile device. For each day, participants also completed an end-of-day survey at their self-reported typical bedtime, which we refer to as the evening survey (M = 4.00; SD = 1.24; range = 1–6). Respondents were compensated $150 for their participation. Data analyzed for this study were from the baseline survey and the evening surveys.
Measures
Baseline
Social partner life stressors
Participants were asked to report whether their close friends or relatives had experienced the following life stressors in the last year: serious health problem or physical injury, serious psychological/cognitive/emotional problem, drug or alcohol problem, serious financial problem, loss of a close social partner due to death/move/breakup, victim of crime or had legal problems, housing or neighborhood problem. Each item was coded as 1 (yes) or 0 (no). For each life stressor the participant was asked which social network member it occurred with, which included information regarding relationship type. Family member relationships included: spouse/romantic partner, child (step, in-law), sibling (step, in-law), grandchild (step), great-grandchild, niece/nephew, parent (step, in-law), cousin, aunt/uncle. Nonfamily member relationships included: friends, neighbor, pastor/minister, coworker/boss/employee, caregiver. We created two separate sum scores: one for the number of family stressors and one for the number of nonfamily stressors experienced. These items were adapted from a measure of chronic stress used in the Health and Retirement Study (Troxel et al., 2003) and a measure of life stressors used in the Family Exchanges Study (Bangerter et al., 2018; Birditt et al., 2010; Huo et al., 2019). Participants were also asked whether they experienced each of the stressors themselves and this was included as a covariate in the models.
Daily questionnaire
Daily rumination
Each night, participants completed two items, adapted from Kircanski et al. (2015), indicating the extent to which they experienced the following over the day: “Today, I was dwelling on my feelings and problems,” and “Today, I was worried about things that could happen,” with response options from 1 to 5: Not at all, A little, Somewhat, Quite a bit, and A great deal. Items were averaged to create a rumination score (women: α = 0.80; men: α = 0.79).
Daily mechanisms
Interpersonal tensions
Each night, participants were asked “Today, did you have any social interactions that made you feel irritated, hurt or annoyed?” and “Today, did you have any social interactions in which you could have felt irritated, hurt or annoyed but decided not to?” with response options of 0 (no) and 1 (yes). These two items were combined such that 0 represents answering both questions with a “no” and 1 represents answering one or both questions with a “yes.”
Interpersonal worries
Each night, participants were asked “Today, did you think about a relationship problem or worry about someone, even though you did not have contact with that person?” with a response option of 0 (no) and 1 (yes).
Support provision
Each night, participants were asked, “Today, did you provide emotional support to anyone? (Emotional support involves listening to someone’s concerns or being available when they are upset),” “Today, did you provide practical support to anyone? (For instance, fixing something around the house, running an errand, or providing a ride),” and “Today, did you give advice to anyone? (That is, help with a decision or suggestions about things they could do),” with response options of 0 (no) and 1 (yes). These three items were summed to create a support exchange scale that ranged from 0 to 3.
Covariates
Covariates for the study included own life stress, demographics, and contextual factors associated with stress and rumination, such as age, gender, Hispanic ethnicity, social network size, and previous day rumination (Birditt, 2014; Gerich, 2014; Johnson & Whisman, 2013; King & DeLongis, 2014; Russell & Taylor, 2009). Participants were asked whether they had experienced the following life stressors in the last year (1 = yes; 0 = no): serious health problem or physical injury, serious psychological/cognitive/emotional problem, drug or alcohol problem, serious financial problem, loss of a close social partner due to death/move/breakup, victim of crime or had legal problems, housing or neighborhood problem. These were summed to make an own life stress score. Participants’ age was continuous and calculated based on birthdate. Gender was dichotomous (0 = female or 1 = male). Hispanic ethnicity included: 0 (not Hispanic) or 1 (Hispanic). Finally, social network size included the total number of individuals listed in the convoy diagram of social network members in which individuals list their close social partners in three concentric circles ranging in closeness (Kahn & Antonucci, 1980).
We also considered whether to control for education, marital status, race, living alone, work status, and the number of social interactions reported across all 3-hr periods. However, we did not include these variables as covariates because they were not associated with rumination. Age and gender were also not significantly associated with rumination but we kept them as covariates because we considered them as moderators.
Analysis Plan
Multilevel structural equation modeling (MSEM) and multigroup MSEM were conducted using Mplus version 7 (Muthén & Muthén, 2015) with maximum likelihood estimation. Because the data were nested (days within respondents), we needed to separately estimate the between and within effects of associations at the lowest level. This is more accurate than centering variables on the group (person) mean when there is missing data (Preacher et al., 2010). MSEM allows for testing direct and indirect effects simultaneously. All variables were standardized prior to analysis to obtain standardized path coefficients, which are also a measure of standardized effect size, including for indirect effects (Kelley & Preacher, 2012; Lorah, 2018; Preacher & Kelley, 2011).
Models were estimated in four steps. First, the direct effect of family and nonfamily life stress on daily rumination was estimated using MSEM. Second, we tested the three daily mediators—interpersonal tensions, interpersonal worries, and support exchanges—as indirect effects using MSEM. Next, we tested whether the direct effects of family and nonfamily life stress on rumination varied by gender using multigroup MSEM. Finally, we tested the three daily mediators—interpersonal tensions, interpersonal worries, and support exchanges—as indirect effects for men and women using multigroup MSEM.
We also tested age differences in all the models. We examined whether the links between family and nonfamily social partner life stress and rumination varied by age by entering interactions between stress and age. There were no significant interactions. Because age was not a significant moderator, it is not discussed further.
Model fit was assessed by commonly used indices: the chi-square test (χ2), the comparative fit index (CFI), the root mean squared error of approximation (RMSEA), and the standardized root mean squared residual (SRMR). Good fit criteria for a model include a nonsignificant chi-square test (although null is often rejected with large sample sizes), RMSEA smaller than 0.08, CFI above 0.90, and SRMR less than 0.08 (Schermelleh-Engel et al., 2003). We tested multigroup models by comparing model fit of three types of path models: (1) a common model where all paths were allowed to vary between groups (gender); (2) a path-invariant model where all paths were constrained to be equal by gender; and if model fit becomes significantly worse, (3) testing models to see which paths varied significantly by gender. Paths that varied by gender were released and estimated freely.
Results
Descriptives
Average daily rumination was 1.66 (SD = 0.75, range = 1–5; Table 1). The most frequently reported family and nonfamily social partner stressor was a physical health problem or injury (45.1% and 32.8%, respectively). The most frequent types of family members experiencing stressors were children (31%), grandchildren (21.8%), and siblings (15.5%). The most frequent types of nonfamily members were friends (91%), neighbors (3.1%), or pastors (1.9%). The sum score for family social partner stressors and nonfamily social partner stressors ranged from 0 to 6 with means of 1.54 (family; SD = 1.37) and 0.98 (nonfamily; SD = 1.19).
Women reported greater family and nonfamily social partner life stressors compared to men (t = 2.57, p < .01; t = 2.30, p < .05, respectively). Men and women reported similar levels of daily rumination (t = −0.89, p > .05).
Associations Between Perceived Social Partner Life Stressors and Daily Rumination
First, we examined the link between family and nonfamily life stressors and daily rumination (Table 2). Model fit met all criteria: χ2(13, N = 1,091) = 15.058, p = .304, CFI = 0.982, RMSEA = 0.012, SRMR = 0.000 (within); 0.050 (between). Greater family and nonfamily social partner stressors were associated with greater rumination (β = 0.17, standard error [SE] = 0.05, p = .001; β = 0.16, SE = 0.06, p = .008, respectively).
What Daily Processes Account for the Link Between Social Partner Life Stressors and Rumination?
We estimated the MSEM models again with interpersonal tensions, worries, and support as mediators for the link between family and nonfamily social partner life stress and rumination. We tested each mediator separately. We found that interpersonal tensions significantly mediated the association between nonfamily life stressors and rumination (indirect effect: β = 0.09, SE = 0.05, p = .041). Model fit met all criteria: χ2(19, N = 1,091) = 22.83, p = .245, CFI = 0.975, RMSEA = 0.014, SRMR = 0.012 (within); 0.061 (between). Thinking about a relationship problem or worries significantly mediated the association between family and nonfamily life stressors and rumination (indirect effects: β = 0.15, SE = 0.04, p < .001; β = 0.17, SE = 0.05, p = .001, respectively). Model fit met all criteria: χ2(19, N = 1,091) = 18.603, p = .483, CFI = 1.000, RMSEA = 0.000, SRMR = 0.022 (within); 0.051 (between). Support provision was not a significant mediator for family or nonfamily life stressors.
Do These Links Vary by Gender?
The first step to examine the direct effect of family and nonfamily social partner life stressors on rumination by gender was to compare the common model where all paths were allowed to vary by gender to a path-invariant model where all paths were constrained to be equal by gender. The model constraining all paths to be equal by gender fit significantly worse (Δχ2(18) = 47.344, p < .001), indicating that there were differences by gender (Schermelleh-Engel et al., 2003). Subsequent models were conducted to determine which parts of the model needed to be estimated freely by gender by examining the chi-square difference in models. Chi-square difference tests indicated that the within person residual for rumination (Δχ2(1) = 4.362, p < .05), the between-person mean for family life stressors (Δχ2(1) = 6.648, p < .01), the between-person mean for nonfamily life stressors (Δχ2(1) = 4.340, p < .05), the path from ethnicity to rumination (Δχ2(1) = 11.665, p < .001), and the path from family life stressors to rumination (Δχ2(1) = 13.436, p < .001) should be allowed to vary freely by gender. All other model components did not vary by gender. Model fit met all criteria: χ2(33, N = 1,091) = 29.839, p = .625, CFI = 1.000, RMSEA = 0.000, SRMR = 0.005 (within); 0.067 (between).
The results of the multigroup MSEM of direct effects showed that the association between family social partner life stressors and rumination varied by gender. Greater family life stressors were associated with greater daily rumination only among women (β = 0.29, SE = 0.07, p < .001) and not among men. The association between nonfamily social partner life stressors and rumination did not vary significantly by gender, indicating that both men and women who reported greater nonfamily social partner life stressors ruminated more in daily life. Table 2 shows the findings from the MSEM.
Multilevel SEM of the Link Between Family and Nonfamily Life Stress and Daily Rumination Overall and by Gender
| Variables | Overall sample | Women | Men | Group differences | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| β | t-values | β | t-values | β | t-values | Δχ2/1 df | ||||
| Previous day rumination | −0.15 | (−6.35) | *** | −0.15 | (−6.43) | *** | −0.15 | (−6.43) | *** | 0.06n.s. |
| Female | 0.03 | (0.58) | — | — | — | — | ||||
| Age | −0.05 | (−1.07) | −0.04 | (−0.89) | −0.04 | (−0.89) | 0.12n.s. | |||
| Social network size | −0.13 | (−2.46) | * | −0.13 | (−2.54) | * | −0.13 | (−2.54) | * | 0.20n.s. |
| Hispanic | 0.02 | (0.44) | −0.13 | (−1.98) | * | 0.21 | (2.86) | ** | 11.67*** | |
| Own life stress | 0.17 | (3.03) | ** | 0.18 | (3.48) | ** | 0.18 | (3.48) | ** | 2.18n.s. |
| Family life stress | 0.17 | (3.22) | ** | 0.29 | (4.38) | *** | −0.04 | (−0.49) | 13.44*** | |
| Nonfamily life stress | 0.16 | (2.66) | ** | 0.15 | (2.54) | * | 0.15 | (2.54) | * | 1.22n.s. |
| Variables | Overall sample | Women | Men | Group differences | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| β | t-values | β | t-values | β | t-values | Δχ2/1 df | ||||
| Previous day rumination | −0.15 | (−6.35) | *** | −0.15 | (−6.43) | *** | −0.15 | (−6.43) | *** | 0.06n.s. |
| Female | 0.03 | (0.58) | — | — | — | — | ||||
| Age | −0.05 | (−1.07) | −0.04 | (−0.89) | −0.04 | (−0.89) | 0.12n.s. | |||
| Social network size | −0.13 | (−2.46) | * | −0.13 | (−2.54) | * | −0.13 | (−2.54) | * | 0.20n.s. |
| Hispanic | 0.02 | (0.44) | −0.13 | (−1.98) | * | 0.21 | (2.86) | ** | 11.67*** | |
| Own life stress | 0.17 | (3.03) | ** | 0.18 | (3.48) | ** | 0.18 | (3.48) | ** | 2.18n.s. |
| Family life stress | 0.17 | (3.22) | ** | 0.29 | (4.38) | *** | −0.04 | (−0.49) | 13.44*** | |
| Nonfamily life stress | 0.16 | (2.66) | ** | 0.15 | (2.54) | * | 0.15 | (2.54) | * | 1.22n.s. |
Notes: Overall sample model fit: χ2 = 15.06, df = 13, p > .05, CFI = 0.98, RMSEA = 0.01, SRMR = 0.00 (within); 0.05 (between). Multigroup (by gender) model fit: χ2 = 21.49, df = 33, p > .05, CFI = 1.00, RMSEA = 0.00, SRMR = 0.00 (within); 0.07 (between). CFI = comparative fit index; df = degrees of freedom; n.s. = not significant; RMSEA = root mean squared error of approximation; SEM = structural equation modeling; SRMR = standardized root mean squared residual.
***p < .001. **p < .01. **p < .05.
Multilevel SEM of the Link Between Family and Nonfamily Life Stress and Daily Rumination Overall and by Gender
| Variables | Overall sample | Women | Men | Group differences | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| β | t-values | β | t-values | β | t-values | Δχ2/1 df | ||||
| Previous day rumination | −0.15 | (−6.35) | *** | −0.15 | (−6.43) | *** | −0.15 | (−6.43) | *** | 0.06n.s. |
| Female | 0.03 | (0.58) | — | — | — | — | ||||
| Age | −0.05 | (−1.07) | −0.04 | (−0.89) | −0.04 | (−0.89) | 0.12n.s. | |||
| Social network size | −0.13 | (−2.46) | * | −0.13 | (−2.54) | * | −0.13 | (−2.54) | * | 0.20n.s. |
| Hispanic | 0.02 | (0.44) | −0.13 | (−1.98) | * | 0.21 | (2.86) | ** | 11.67*** | |
| Own life stress | 0.17 | (3.03) | ** | 0.18 | (3.48) | ** | 0.18 | (3.48) | ** | 2.18n.s. |
| Family life stress | 0.17 | (3.22) | ** | 0.29 | (4.38) | *** | −0.04 | (−0.49) | 13.44*** | |
| Nonfamily life stress | 0.16 | (2.66) | ** | 0.15 | (2.54) | * | 0.15 | (2.54) | * | 1.22n.s. |
| Variables | Overall sample | Women | Men | Group differences | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| β | t-values | β | t-values | β | t-values | Δχ2/1 df | ||||
| Previous day rumination | −0.15 | (−6.35) | *** | −0.15 | (−6.43) | *** | −0.15 | (−6.43) | *** | 0.06n.s. |
| Female | 0.03 | (0.58) | — | — | — | — | ||||
| Age | −0.05 | (−1.07) | −0.04 | (−0.89) | −0.04 | (−0.89) | 0.12n.s. | |||
| Social network size | −0.13 | (−2.46) | * | −0.13 | (−2.54) | * | −0.13 | (−2.54) | * | 0.20n.s. |
| Hispanic | 0.02 | (0.44) | −0.13 | (−1.98) | * | 0.21 | (2.86) | ** | 11.67*** | |
| Own life stress | 0.17 | (3.03) | ** | 0.18 | (3.48) | ** | 0.18 | (3.48) | ** | 2.18n.s. |
| Family life stress | 0.17 | (3.22) | ** | 0.29 | (4.38) | *** | −0.04 | (−0.49) | 13.44*** | |
| Nonfamily life stress | 0.16 | (2.66) | ** | 0.15 | (2.54) | * | 0.15 | (2.54) | * | 1.22n.s. |
Notes: Overall sample model fit: χ2 = 15.06, df = 13, p > .05, CFI = 0.98, RMSEA = 0.01, SRMR = 0.00 (within); 0.05 (between). Multigroup (by gender) model fit: χ2 = 21.49, df = 33, p > .05, CFI = 1.00, RMSEA = 0.00, SRMR = 0.00 (within); 0.07 (between). CFI = comparative fit index; df = degrees of freedom; n.s. = not significant; RMSEA = root mean squared error of approximation; SEM = structural equation modeling; SRMR = standardized root mean squared residual.
***p < .001. **p < .01. **p < .05.
Next, we tested the mediation models using multigroup MSEM to determine whether the indirect effects varied by gender. Interpersonal tensions significantly mediated the association between nonfamily social partner life stressors and rumination for men only (indirect effect: β = 0.13, SE = 0.06, p = .031; Figure 1). There was no gender difference in the indirect effect of worries for the link between nonfamily stressors and rumination (Figure 2). Thinking about a relationship problem or worries significantly mediated the association between family social partner life stressors and rumination for women only (indirect effect: β = 0.20, SE = 0.06, p < .001; Figure 3).
Multilevel SEM with interpersonal worries as mediator of the link between nonfamily life stress and daily rumination. Model fit: χ2 = 18.60, df = 19, p > .05, CFI = 1.00, RMSEA = 0.00, SRMR = 0.02 (within); 0.05 (between). Values are standardized path coefficients. The path coefficient from nonfamily life stress to rumination above the arrow is from Step 1 (does not include interpersonal worries) and the path coefficient from nonfamily life stress to rumination below the arrow is from the mediation model that does include interpersonal worries. Covariates in the model included previous day rumination, age, gender, Hispanic ethnicity, social network size, family life stress, and own life stress. ***p < .001. **p < .01. CFI = comparative fit index; df = degrees of freedom; RMSEA = root mean squared error of approximation; SEM = structural equation modeling; SRMR = standardized root mean squared residual.
Multilevel SEM with interpersonal tensions as mediator of the link between nonfamily life stress and daily rumination by gender (showing men only). Model fit: χ2 = 34.07, df = 48, p > .05, CFI = 1.00, RMSEA = 0.00, SRMR = 0.02 (within); 0.09 (between). Values are standardized path coefficients. The path coefficient from nonfamily life stress to rumination above the arrow is from Step 1 (does not include interpersonal tensions) and the path coefficient from nonfamily life stress to rumination below the arrow is from the mediation model that does include interpersonal tensions. Covariates in the model included previous day rumination, age, Hispanic ethnicity, social network size, family life stress, and own life stress. **p < .01. *p < .05. CFI = comparative fit index; df = degrees of freedom; RMSEA = root mean squared error of approximation; SEM = structural equation modeling; SRMR = standardized root mean squared residual.
Multilevel SEM with interpersonal worries as mediator of the link between family life stress and daily rumination by gender (showing women only). Model fit: χ2 = 38.32, df = 47, p > .05, CFI = 1.00, RMSEA = 0.00, SRMR = 0.03 (within); 0.08 (between). Values are standardized path coefficients. The path coefficient from family life stress to rumination above the arrow is from Step 1 (does not include interpersonal worries) and the path coefficient from family life stress to rumination below the arrow is from the mediation model that does include interpersonal worries. Covariates in the model included previous day rumination, age, Hispanic ethnicity, social network size, nonfamily life stress, and own life stress. ***p < .001. CFI = comparative fit index; df = degrees of freedom; RMSEA = root mean squared error of approximation; SEM = structural equation modeling; SRMR = standardized root mean squared residual.
Discussion
Individuals are negatively affected by their own experiences of stress as well as the stresses experienced by social partners. The majority of research has focused on life stresses of particular social partners, including spouses and adult children. In the current study, we extended the focus to examine how perceptions of life stresses experienced by a wider array of family and nonfamily predict daily rumination among men and women in older adulthood. It is especially important to examine social partner life stress later in life when there is a greater investment in maintaining emotionally meaningful social relationships (Carstensen, 2021).
Present study findings indicate that among older adults, the links between close partner life stress and daily rumination vary in important ways by gender. Women ruminate more in daily life when they perceive that family and nonfamily social partners have more life stress, whereas men ruminate more when they perceive that nonfamily social partners are experiencing greater life stress, but not when they perceive that family members are experiencing greater life stress. For women, worries about social partners appeared to be an important mechanism by which social partner life stress affected rumination, whereas both worries and interpersonal tensions were mechanisms for men. The present findings underscore the value of considering how perceptions of life stressors experienced by close social partners affect rumination.
While studies suggest that women are more likely to ruminate than men (McIntyre et al., 2008), the older adults in the present study showed that men and women are equally likely to ruminate in daily life, although the life stresses that predict rumination vary by gender. The finding that there was no gender difference in rumination is consistent with previous research suggesting that older adults show fewer gender differences in rumination than younger adults due to age-related declines in rumination among women (Thomsen et al., 2005). Further, it is possible that there are fewer gender differences when examining daily or state-level rumination compared to trait-level rumination.
The current study examined perceptions of older adults’ family and nonfamily social partners’ life stressors. The previous literature has shown that the problems of social partners have implications for well-being (Stone et al., 2011), but there have been no studies to our knowledge that examine the implications of these stressors for daily rumination, particularly by gender. We found that women ruminated more when they perceived that family and nonfamily members had problems, whereas men ruminate more when only nonfamily had problems. This is somewhat inconsistent with theory suggesting that women are more vulnerable to stressful events experienced by their close social partners, and that women are more likely to ruminate due to their greater feelings of responsibility for the emotional tone of relationships (Kiecolt-Glaser & Newton, 2001). That is, men and women appear to ruminate more when social partners have problems but the particular social partners vary. Women may ruminate about a wider array of social partners because they tend to be more emotionally invested in the emotional tone of all their close social ties. Women may have greater contact with family compared to men and may experience greater obligations for kin-keeping than do men (Kahn et al., 2011). Women also tend to be more likely to co-ruminate with social partners regarding stressors (Rose, 2002; Stone et al., 2011), which may lead to a greater array of relationships that cause stress. Overall, this study indicates that it is critical to examine close social partner stress, as it has important implications for the daily lives of older women and men. Furthermore, including nonfamily ties allowed for the identification of previously understudied relationship effects among men.
For women, everyday interpersonal worries accounted for the link between family and nonfamily life stressors and daily rumination. Research has long indicated that life stress proliferates in daily life, and that everyday stress can affect well-being more than life stress (Genet & Siemer, 2012; Gouin et al., 2012; McIntosh et al., 2010). This study indicates that everyday worries about others provide a mechanism explaining the link between close social partner life stressors and rumination. In contrast, both interpersonal tensions and worries accounted for the link between nonfamily stresses and rumination among men. Interestingly, nonfamily members appear to cause irritation and worries among men; future research should examine these mechanisms further. Perhaps because friendships are usually positive and associated with positive affect, the introduction of life stress causes tensions in those ties among men.
In addition, there were no age differences in the associations between family and nonfamily social partner life stress and rumination. According to socioemotional selectivity theory (Carstensen, 2021) and the strength and vulnerability integration model (Charles, 2010), we hypothesized oldest-old adults would be less negatively affected by the life stresses of others (showing smaller associations between social partner stress and rumination) compared to young-old adults. However, we did not find support for this proposition. Feelings of greater investment in relationships may lead social partner problems to be distressing to older adults irrespective of age-related reductions in negative affective experiences. Further, according to the strength and vulnerability integration model, when individuals are less able to use avoidance, there are fewer age-related improvements in negative affective experiences or emotional reactivity. The problems of others may be more difficult to regulate or avoid because individuals have less control over those problems.
Future Research
There are several directions for future research to pursue. The sample used in the present study is a regional sample and participants are fairly well-educated. While ethnically diverse, further examination of the links between social partner life stress and rumination among more representative samples of older adults is needed. Future research should also examine the implications of stress and rumination for daily well-being and physical health. For example, is it worse for one’s health to ruminate about one’s own life stress (over which individuals may have greater control) or about others’ life stress (when giving support to others may be beneficial or detrimental to one’s own health)? Because we did not collect data on the specific life stresses older adults were ruminating about, we also need to conduct more research on the process of rumination in daily life to understand the specific issues about which older adults ruminate. We do not know, for example, whether individuals are ruminating about life stresses or other issues. Unfortunately, we do not have information regarding the severity, the duration, or controllability of the life stresses. These are important aspects of stress to consider as they most likely have implications for rumination.
Another area to pursue is how rumination plays out among families and within close relationships. The data for this study include only the individuals’ perceptions of their close social partners’ life stresses and not their social partners’ reports. We know from our previous research that the stresses experienced by close social partners also have important effects on individual well-being (e.g., Birditt et al., 2016b; Wang et al., 2020). Because we do not have dyadic information, there is no way to know whether individuals are co-ruminating with social partners about stress. For instance, women might be more likely to co-ruminate with their close ties regarding life stresses, whereas men might be more solitary in their rumination. Future work should consider the perspectives and daily experiences of individuals and members of their social networks.
Overall, this study makes an important contribution to the literature by showing that close partner stress among family and nonfamily is associated with greater daily rumination among women, whereas only nonfamily stress is associated with greater rumination among men. Furthermore, these close partner life stresses seem to affect daily rumination via worries (men and women) and interpersonal tensions (men). These findings provide important potential targets of intervention among men and women as rumination is associated with increased depression (Nolen-Hoeksema et al., 2008; Ruscio et al., 2015). This study offers important clues regarding the daily lives of older women and men, and provides several avenues for future research focused on psychological well-being in later life.
Funding
This study was supported by a grant from the National Institute on Aging (NIA), (R01AG046460; Social Networks and Well-being in Late Life: A Study of Daily Mechanisms to K. L. Fingerman, Principal investigator). This research was also supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2CHD042849 to the Population Research Center at The University of Texas at Austin). C. A. Polenick was supported by the NIA (K01AG059829).
Conflict of Interest
None.
Data Availability
This project was not preregistered. Data and code will be made available upon request.
