Isolation or Replenishment? The Case of Partner Network Exclusivity and Partner Loss in Later Life

Abstract

Objectives

People’s partners and spouses often provide a wide range of essential emotional and practical support. As crucial as they may be, a nontrivial segment of the older population appears to limit close discussions to their partner alone, a phenomenon we term “partner network exclusivity.” This network structure could leave people vulnerable to partner losses and subsequent social isolation. The present research has 3 aims: (a) examine the prevalence of partner-exclusive networks among European older adults; (b) consider who is most likely to inhabit such networks; and (c) investigate whether and how individuals in such precarious networks rebalance them in case of partner losses.

Methods

The analysis uses Wave 4 (2011) and Wave 6 (2015) of the Survey of Health, Ageing and Retirement in Europe (SHARE) to perform logistic regression on one’s possession of partner-exclusive networks and the addition of core ties.

Results

More than a quarter of partnered respondents (28.1%) are in partner-exclusive core networks. Men, childless individuals, and those with financial difficulties are most likely to occupy such networks. Individuals in partner exclusivity are especially likely to enlist additional ties upon partner loss. Nevertheless, men and individuals at early old age are relatively unlikely to rebalance their core networks in case of partner death.

Discussion

This study provides new evidence that network replenishment following relationship disruptions is plausible even for those from precarious network settings. Nevertheless, widowhood produces patterns of vulnerability for a subset of older adults in partner-exclusive core networks.

A romantic partner is often the most crucial part of one’s core network in later life, typically offering a wide variety of emotional and instrumental supports (Cornwell & Laumann, 2015, 2018; Iveniuk et al., 2020). Central though their role may be, partners are typically surrounded by other core network members, including friends, neighbors, children, and other relatives. Yet approximately one in six partnered older adults in Europe confine their core network to only one’s partner (Litwin & Stoeckel, 2013), forgoing alternative sources of close relationships and living in a distinct form of network precarity. Indeed, individuals in such a scenario—what we term partner network exclusivity—are at imminent risk of having no available confidants or companions in the event of partner loss. Entering such an isolated status could lead to higher depression, higher perceived loneliness, poorer health, and reduced quality of life (Litwin & Levinsky, 2021; Perry et al., 2018; Vozikaki et al., 2018; Wrzus et al., 2013).

Given this position of precarity, individuals in partner-exclusive core networks should have every reason to recruit more ties upon partner loss. Indeed, the need for support often spurs network replenishment and becomes an opportunity for network growth (Atchley, 1989; Cornwell & Laumann, 2015, 2018; Lamme et al., 1996). The urge to build new relationships could even reflect a deep-seated evolutionary aversion to relationship loss and loneliness (Cacioppo & Cacioppo, 2018; Das, 2021). Overall, older adults often manage to rebalance their networks to maintain similar size and structural composition (Cornwell et al., 2021). Such processes could be vital for those whose partner was the only member of their core network.

Nevertheless, individuals in partner-exclusive core networks face unique challenges when replenishing their connections upon partner loss. Lacking alternative ties in the first place, the restricted nature of such networks could limit one’s acquisition of additional connections (Cornwell & Laumann, 2018; Lamme et al., 1996; Perry et al., 2018). Individuals who emphasize partner support are more likely to lose relationships bridged by the partner after divorce and to receive less support from family after the spouse dies (Iveniuk et al., 2020; Kalmijn & van Groenou, 2005; Perry et al., 2018). These limitations could become barriers to recruiting new core network members. The growing number of childless individuals could be particularly vulnerable to partner loss if they find themselves in partner exclusivity, although such people could resort to other relatives or friends to fill the gap (Fihel et al., 2022; Verdery et al., 2019). In addition, many individuals could have grown accustomed to the relatively restricted, yet more private, partner-only setup over their life course, thus being hesitant to incorporate additional ties in the core network, especially non-kin ones.

The present research aims to extend research on network transitions by showing how network replenishment unfolds in cases of partner loss at older ages, focusing on a potentially vulnerable group of older adults—those in a position of partner network exclusivity. The present research uses the fourth and sixth waves of the Survey of Health, Ageing and Retirement in Europe (SHARE), which offers a dedicated network name generator method with a panel design, allowing us to identify partners’ departures and new additions to the network.

Background

Partners and Loss in Later Life

Partners often have a long history of mutual emotional and practical support through sharing a household, raising children, and facing family concerns. Having a spouse or partner contributes to higher well-being, lower levels of loneliness, and higher life satisfaction in aging (Boger & Huxhold, 2018; Djundeva et al., 2019). Not surprisingly, individuals are likely to position their spouse or partner at the center of the core network as the closest and most meaningful relationship (Cohn-Schwartz et al., 2021; Mejia & Hooker, 2004). Indeed, voluntarily listing one’s partner in one’s core network signifies an intimate and supportive relationship (Liao & Stevens, 1994). Spousal loss is, therefore, a critical life transition, especially at an older age, when networks play vital coping roles. Prior research has shown that partner loss is associated with increased loneliness, poor health, and depressive symptoms (Antonucci et al., 2001; Dahlberg et al., 2022; Grundy et al., 2019).

Social Network Change Following Partner Loss

Several theoretical perspectives propose explanations for how older adults’ networks adapt to partner loss. The social convoy model suggests that individuals adjust their social relationships according to developmental and contextual factors (Antonucci et al., 2010). So, while life transitions and network losses disrupt the existing opportunity structure for interactions, they could simultaneously set the ground for developing alternative ties and network growth (Cornwell & Laumann, 2018; Perry et al., 2018; Wrzus et al., 2013). Similarly, continuity theory suggests that older adults have grown accustomed to their social roles and activities established throughout their life course. Thus, they should be highly motivated to sustain their networks amid life transitions by cultivating new ties or rekindling dormant ones (Atchley, 1989; Donnelly & Hinterlong, 2010).

Supporting these perspectives on adaptation, individuals experiencing widowhood—after a brief period of social withdrawal and mourning—tend to become more engaged in social activities (Donnelly & Hinterlong, 2010; Li, 2007; Utz et al., 2002). Widows also often mobilize certain connections, especially family members, to provide more contact and support (Cornwell & Laumann, 2018; Ferraro, 1984; Kalmijn, 2012; Pahl & Pevalin, 2005). Divorced individuals, for their part, often get support from friends not linked to their expartners and further expand their friendship circles (Albeck & Kaydar, 2002; Kalmijn & Van Groenou, 2005; Pahl & Pevalin, 2005). Repartnering is also prevalent after partner loss (Lamme et al., 1996), contributing a potential new hub to the core network. Drawing together both convoy and continuity perspectives, recent research observes that network losses and additions in later life often cancel each other out, ultimately producing stable or even slightly enlarged personal networks (Cornwell & Laumann, 2015; Cornwell et al., 2021; Roth, 2020; Schwartz & Litwin, 2018). Overall, it is plausible that older adults often rebuild and rebalance their networks following tie losses.

Existing research on partner loss and network replenishment raises several important points of contingency. The first is the reason for partner loss. As an often exogenous shock beyond one’s control, the death of a spouse usually intensifies the support from one’s network (Cornwell & Laumann, 2018). Meanwhile, divorce or relationship breakup could deteriorate individuals’ relationships with families and friends and potentially limit one’s options for additional ties, particularly if such ties were bridged by the expartner (Albeck & Kaydar, 2002; Perry et al., 2018; Wrzus et al., 2013). Children are more likely to blame fathers for the termination of the marriage (Kaufman & Uhlenberg, 1998). And divorce may also estrange family members who are frustrated and judgmental with the transition (Pahl & Pevalin, 2005).

Second, older age could become a barrier to acquiring additional ties. The availability of age peers as potential network members, for instance, lowers as age increases (Mudrazijia et al., 2020). Research also documents that individuals at older ages often tend to have smaller and less emotionally close networks (Ajrouch et al., 2005; Cornwell et al., 2008). Older individuals also face more limitations from physical and cognitive declines, which could limit their capability to sustain and expand their core networks (Cornwell, 2011, 2015; Litwin & Stoeckel, 2013). Although gray divorce (i.e., at age 50+) and subsequent repartnering have become more common in the past few decades (Brown & Lin, 2012; Brown et al., 2019), repartnering is likewise constrained by advanced age (Schimmele & Wu, 2016). Nevertheless, research on network change observes that as age increases, network losses due to mortality are more likely to result in the addition of new ties (Cornwell & Laumann, 2018).

Third, network replenishment processes are highly gendered. Women often have more contact, support, and new ties following widowhood and divorce than men (Dykstra & De Jong Gieveld, 2004; Kalmijn, 2007, 2012; Lamme et al., 1996; Schwartz & Litwin, 2018). This apparent resilience could largely be attributed to a lower reliance on support from their husbands than vice versa, a higher openness to expressing needs for help, the kin-keeping role in the family, greater social participation, and the higher availability of age peers with similar experiences of partner loss (Ivenuik et al., 2020; Verdery et al., 2019). Moreover, loneliness seems to motivate social interactions with friends among women but not men (Das, 2021). Meanwhile, men are more likely to repartner, especially after divorce, possibly for the support from the partner (Brown et al., 2018; Schimmele & Wu, 2016).

The Case of Partner Exclusivity and Network Addition

The present study extends research on network changes and aging by scrutinizing how network rebalancing unfolds for individuals in precarious partner-only core networks. Acquiring additional ties is especially critical when individuals in partner-exclusive networks encounter partner loss. Individuals in such a network structure could focus disproportionately on their partner for interactions and support while having limited social capital elsewhere (Kalmijn, 2012; Litwin & Stoeckel, 2013; Roth, 2020). Losing the partner puts them at risk of social detachment, an identified risk factor for loneliness and poor well-being (Fiori et al., 2006; Litwin & Levinsky, 2021). Moreover, losing one’s partner could compromise the interaction routines with one’s broader network and disrupt coordinated support, primarily when the partner used to play the organizing or bridging role (Cornwell, 2011). Not surprisingly, losing a partner as the only core network member would leave one in urgent need to regain connectedness and access to support.

At the same time, older adults in partner-exclusive close networks could face unique constraints when replenishing their core networks. Some are endogenous—a lack of other established core network members means a deficit in coordinating responsive support and enlisting new network members to counterbalance the loss. For instance, not having a close friend could mean little chance of a network bridge linking the individual to new relationships. In addition, factors leading older adults to inhabit such restricted networks could impede the acquisition of additional core network ties when a partner is lost. For example, lower levels of education, being out of the labor force, poor physical or mental health, functional and cognitive limitations, financial disadvantages, caregiving obligations, and limited social participation could all confine one’s network and limit the formation of additional ties (Burholt et al., 2020; Cornwell, 2015; Fiori et al., 2006).

Indeed, the restricted nature of a partner-exclusive network may have implications for who enters the core network. Understandably, individuals are likely to prioritize family members over friends and other less intimate people when seeking substitutions in their network (Cantor, 1979; Carstensen, 1992, 2006; Litwin & Levinsky, 2021). This could also apply to individuals in partner-exclusive scenarios. Nevertheless, extended family members and friends are also valuable candidates, particularly for childless or kinless individuals (e.g., Fihel, 2022; Verdery et al., 2019). Considering that the lost partner could provide instant and on-site support, individuals may favor more accessible sources in proximity to fulfill such needs (Small & Sukhu, 2016). Meanwhile, with limited options, one may be more open to having supporters even at a distance, particularly family members such as children.

The present study aims to answer several questions to contextualize the processes of partner loss, particularly for older adults in partner-exclusive networks. First, which older adults are likely to inhabit core networks that consist exclusively of a partner rather than incorporating other ties? Second, are individuals in partner-exclusive networks relatively unlikely to replenish their core network after partner loss? In answering these questions, we draw particular attention to the contingency factors (mentioned earlier) emphasized in existing partner loss research—the reason for loss, gender, and age. Finally, who and where are the new ties added by individuals in partner-exclusive networks when they lose that partner?

Data and Methods

Data

We retrieve the analytical sample from Wave 4 (W4; collected in 2011) to Wave 6 (W6; in 2015) of SHARE (Börsch-Supan, 2020; Börsch-Supan et al., 2013; Malter & Börsch-Supan, 2017). SHARE focuses on the aging population in 14 European countries (Austria, Belgium, Czech Republic, Denmark, Estonia, France, Germany, Italy, Poland, Portugal, Slovenia, Spain, Sweden, and Switzerland). The survey employs a country-specific probability sampling approach to ensure population coverage and collect data using computer-assisted personal interviewing (CAPI). SHARE staff conducted face-to-face interviews with laptops in the respondent’s home, using the questionnaire translated into the local language. The sample consists of community-dwelling individuals aged 50 years and older at the time of data collection and their partners in the same household, regardless of age. The analytic sample consists of individuals aged 50 or older at Wave 4 in 2011 who identified a partner in the core network at the time and participated again at Wave 6.

Waves 4 and 6 of SHARE both feature dedicated ego-centric network modules with name generators. Respondents could list up to six individuals with whom they “discussed important things” in the last 12 months and an additional spot for someone they felt was “important for any reason.” This way, one could list up to seven individuals as their core network (see Supplementary Materials, Section 1). SHARE also gathered detailed information about each network member, such as their relationship to the respondent (e.g., partner, child, relative, or friend). SHARE also keeps track of changes in the core network’s members between the two waves, allowing us to identify lost or gained ties.

Variables

The focal variable, partner exclusivity, measures whether one listed their partner as the sole member of the core network at Wave 4 (1 = yes; 0 = otherwise; see Supplementary Materials, Section 2, for a summary of other solo-member core network configurations). We identify the other focal dependent variable, network addition, based on whether respondents listed anyone in their core network at Wave 6 who was not present in Wave 4. For each member listed at Wave 6, the survey asked, “Was W6 SN member [1–7] mentioned before in W4?” If any tie fits the “No = 0” criteria, it is considered a newly added one. The respondent’s relationship with the added tie is also of interest; this variable denotes whether the connection is a new partner, a child, an extended relative, a friend or neighbor, or a professional support provider (clergy, health care practitioner, etc). Geographic distance to the added tie was measured in several categories: 1 = within 5 km, 2 = 5–25 km, and 3 = >25 km. Although not central to our analysis, we also measured one’s average level of contact frequency and emotional closeness with core network members to facilitate a comparison of the support capacity between partner-exclusive and other network settings. One’s contact frequency with core network ties ranges from Daily (= 1) to Never (= 7). Emotional closeness with a core network tie comes in four categories, including 1 = not very close; 2 = somewhat close; 3 = very close; and 4 = extremely close.

Our analysis also considers partner loss, where a partner at Wave 4 is no longer part of the core network at Wave 6 due to death or other reasons. At Wave 6, upon completion of the name generator procedure, the CAPI system compared the names just mentioned to those given at Wave 4. The system then asked, “Last time you mentioned [Name] as [Relationship]. Did you mention him/her again today?” If one refers to a partner and the answer is “Yes,” it means the partner remains in the core network. If the answer was “No,” the system would inquire: “What is the main reason you did not mention [Name] (the partner) this time?” Possible reasons for a partner loss from the network include “[Name] died (44.1% of all partner losses)”; “I/[Name] moved (3.0%)”; “I/[Name] became ill or had a health problem (7.4%)”; “Respondent does not recognize the named person (2.0%)”; “We are no longer close (17.4%)”; and “Other reasons/Unspecified (26.3%).” Respondents can also correct simple recall and technical mistakes at this point, specifying a name as “I forgot, [Name] should have been included” or “Wrong, [Name] WAS mentioned this time,” which are still part of the network. The follow-up question and the match across two waves are an essential buffer against mistaking forgotten ties for lost ones. We have also cross-validated these network change values with marital status transition data from the survey and counted a loss as death if one experienced widowhood (see Supplementary Materials, Section 3, for more details).

We also consider other sociodemographic background variables potentially associated with partner exclusivity and network additions. Respondents’ age at Wave 4 is a continuous variable, while gender labels men as 0 and women as 1. We specified the respondent’s country of residence as a nominal variable. We also measured a range of structural constraints that could promote partner-exclusive networks and limit one’s rebalancing effort; these include lower education, work status, financial strain, caregiving obligations, childlessness, and area of residence (Cornwell & Laumann, 2018; Litwin & Stoeckel, 2013). Education is presented initially as the seven-level International Standard Classification of Education (ISCED)-97 criteria, which we regroup into three categories: low education = 1 (up to lower secondary education or compulsory education), moderate education = 2 (upper secondary and postsecondary nontertiary education), and high education = 3 (such as the first stage of tertiary education and the second stage of tertiary education). We code labor force participation as 1 if one was not employed or self-employed at Wave 4, such as retired, unemployed, permanently sick/disabled, or homemaker. We identify respondents’ household financial strain from the question: “Thinking of your household’s total monthly income, would you say that your household is able to make ends meet?” Possible answers included “with great difficulty,” “with some difficulty,” “fairly easily,” and “easily.” We coded free from financial difficulties as 0, while 1 indicates some hardships. We also flag individuals responsible for providing intensive care to someone in the household (0 = No, 1 = Yes). We build a dichotomous variable for grandparenting, denoting older individuals who looked after their grandchildren without the presence of their parents in the past 12 months as 1. Childless respondents are labeled as 1 and 0 otherwise. Participating in any organized social activities is coded as 1 (0 = otherwise). Living in rural areas, such as small towns or villages, is coded as 1, compared to living in cities, suburbs, and big towns (=0).

Finally, we consider physical and cognitive limitations for their potential impact on network formation and change. Physical limitations focus on difficulties performing activities of daily living (ADL), including dressing, walking across a room, bathing/showering, eating (cutting up food), getting in/out of bed, and using the toilet. Experiencing any difficulty is coded as 1, otherwise, 0. Individuals requiring intensive care are labeled as 1 (otherwise, 0). Individuals rating their health as fair or poor are coded as 1 (good, very good, excellent = 0). Experiencing challenges in cognitive capability is coded as 1 if one shows any deficiency in numeracy, time orientation, word learning, or word recall.

Analytic Strategy

Our analysis starts with a sample of older individuals who participated in both waves. It presents descriptive statistics to examine the prevalence of partners’ presence in core networks as well as the prevalence of partner-exclusive ones.

The second analysis stage refines the sample to older adults with at least a partner in the core network (N = 17,429). It first uses logistic regression to reveal, among this subsample, which older Europeans are most likely to be in partner-exclusive networks. Using this set of partnered individuals, the analysis further examines whether partner loss due to death or other reasons is associated with acquiring new network members. In particular, we consider whether tie acquisition differs between those previously in partner-exclusive networks and those with other ties besides their partner. This binary logistic regression model uses data across baseline and follow-up waves, but the outcome—whether there is a new network member—only takes on a meaningful value at follow-up (i.e., no ties could be newly added at baseline).

The third stage zooms in on individuals starting from a position of partner exclusivity (N = 4,892). We first employ binary logistic regression to examine whether the association between partner loss and network addition is conditional on partner-exclusive participants’ age and gender. As noted earlier, the dependent variable reflects a network change from baseline to follow-up. We further scrutinize the new ties in the core network, comparing their relationship type and geographic distance across reasons for partner loss, age, and gender.

Across each of the analyses, we use list-wise deletion to handle missing data, a common practice in research using SHARE. The highest missingness among the selected variables was 4.2%, while the total dropped cases accounted for 4.7% of the entire sample. We further verified the findings using multiple imputations with chained equations. The study uses weight to accommodate the sampling process and the sample losses between W4 and W6 (see Supplementary Materials, Section 4, for more details about the weight).

Results

The first stage of the analysis presents the prevalence of mentioning a partner in one’s core network and partner exclusivity across the entire sample participating in both waves. As shown in Table 1, 59.3% (17,429/29,416 × 100%) of all the older adults who participated in both waves mentioned having a partner in their core network. More than a quarter of partnered respondents (4,892/17,429 × 100% = 28.1%) listed the partner as the only tie and are thus considered to occupy partner-exclusive networks. The overall prevalence of partner exclusivity among individuals participating in both waves is 16.7% (4,892/29,416 × 100%). By comparison, more than half of partnered older adults mentioned at least one additional member in their core network acquired between the two waves (58.0%). Individuals with other confidants besides their partners report a lower proportion (55.5%), while those in partner exclusivity have a higher proportion (64.5%). At Wave 6, 8.3% of older adults reported no longer having their Wave 4 partner in the core network (3.5% due to death and 4.8% for other reasons). The proportions of partner loss in partner-exclusive networks are similar: 3.2% and 4.8%, respectively.

Table 1.

Descriptive Statistics for Key Variables, Unweighted

	Respondents participated in both waves		Having partner in the core network at W4		Having a partner and other core tie(s) at W4		In partner-only networks at W4
	N = 29,416		N = 17,429		N = 12,537a		N = 4,892a
	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Added any new member in SN	0.64	0.48	0.58	0.49	0.55	0.50	0.64	0.48
Had partner in SN at W4	0.59	0.49	1.00		1.00		1.00
Partner in SN at W4, lost by death	0.03	0.17	0.04	0.19	0.04	0.19	0.03	0.18
Partner in SN at W4, lost for other reasons	0.03	0.17	0.05	0.21	0.05	0.21	0.05	0.21
Age in 2011	65.56	9.06	64.19	8.41	63.93	8.35	64.84	8.54
Gender: Women	0.58	0.49	0.48	0.50	0.53	0.50	0.36	0.48
Education: low (≤high school)	0.40	0.49	0.36	0.48	0.35	0.48	0.40	0.49
Education: moderate (some college)	0.39	0.49	0.41	0.49	0.41	0.49	0.41	0.49
Education: high (college degree)	0.21	0.41	0.23	0.42	0.24	0.43	0.19	0.39
Employed/working	0.28	0.45	0.32	0.47	0.33	0.47	0.28	0.45
Financial strain	0.38	0.49	0.34	0.47	0.31	0.46	0.41	0.49
Self-rated fair or poor health	0.39	0.49	0.36	0.48	0.35	0.48	0.38	0.49
Facing any cognitive challenge	0.66	0.48	0.63	0.48	0.61	0.49	0.68	0.46
Physical limitations: Any ADL	0.09	0.29	0.08	0.26	0.08	0.27	0.07	0.26
Receiving care	0.03	0.17	0.03	0.18	0.03	0.18	0.03	0.16
Depressive symptoms (EURO-D)	2.46	2.19	2.18	2.03	2.25	2.06	2.02	1.96
Any formal social participation	0.35	0.48	0.37	0.48	0.39	0.49	0.32	0.46
Childlessness	0.09	0.29	0.06	0.23	0.05	0.22	0.08	0.26
Caregiving in HH	0.06	0.24	0.07	0.25	0.07	0.26	0.06	0.23
Grandparenting	0.25	0.43	0.24	0.43	0.26	0.44	0.19	0.39
Residence in a rural area	0.61	0.49	0.63	0.48	0.62	0.49	0.65	0.48
Avg. emotional closeness	3.21	0.63	3.32	0.58	3.26	0.55	3.47	0.62
Avg. contact frequency	1.90	0.96	1.64	0.76	1.88	0.76	1.01	0.16

Note: ADL = limitations in activities of daily living; EURO-D = European Depressive symptoms scale; HH = household; SD = standard deviation; SN = social network; W4 = Wave 4.

^aColumns sum to “Having partner in the core network at W4 (N = 17,429).”

The second stage of the analysis focuses on Wave 4 partnered adults (N = 17,429). It uses multivariable analysis to show which of these people are most likely to inhabit a partner-exclusive core network. Table 2 presents the averaged marginal effects (AMEs) from a series of individual social-demographic backgrounds compared to their respective reference groups. Not surprisingly, men are more likely than women to identify their partner as their only core network member (12.7% higher probability, p < .001). Childless individuals are likewise much more likely to have partner-exclusive networks than those with children (11.4%, p < .001). Financial difficulties also play a significant role, predicting a 5.8% higher chance of having a partner-only network setup (p < .01). In comparison, grandparenting care is associated with a lower chance of partner exclusivity by 8.2% (p < .001). Meanwhile, age, education, physical, mental, and cognitive health, urban versus rural location, organized activity participation, and caregiving obligations are not significantly associated with partner network exclusivity.

Table 2.

Factors Associated With Having a Partner-Exclusive Network, in Averaged Marginal Effects

	AME	SE
Age in 2011	0.001	(0.001)
Gender: Ref = men
Women	−0.127***	(0.012)
Education: Ref = low education
Moderate (some college)	0.021	(0.017)
High (college degree)	−0.004	(0.022)
Employment status: Ref = not working
Employed/working	−0.037	(0.020)
Financial strain: Ref = having no difficulty
Any difficulty to make ends meet	0.058**	(0.019)
Self-rated Health: Ref = good or better
Fair or poor health	−0.005	(0.015)
Physical limitations: Ref = no ADL
Any ADL	−0.018	(0.028)
Receiving care	−0.001	0.037
Depressive symptoms (ERUO-D)	−0.006	(0.004)
Formal social participation: Ref = no participation
Any formal social participation	−0.012	(0.017)
Child status: Ref = have any children
Childlessness	0.114***	(0.027)
Residence: Ref = in urban areas
Residence in rural areas	0.030	(0.016)
Cognitive capabilities: Ref = no difficulties
Facing any cognitive challenge	0.025	(0.017)
Caregiving in household: Ref = not giving care
Giving care to someone in the household	−0.047	(0.027)
Grandparenting: Ref = no grandparenting role
Taking care of any grand children	−0.082***	0.015

Notes: The model also controls for countries of residence not listed in the table. ADL = limitations in activities of daily living; AME = averaged marginal effect; EURO-D = European Depressive symptoms scale; SE = standard error.

**p < .01. ***p < .001.

Using the same sample of partnered older adults (N = 17,429), the analysis further examines whether partner loss due to death or other reasons is associated with adding new network members between waves. Multivariable logistic regression compares older individuals in partner-exclusive networks with those with alternative ties in the core network. Figure 1 presents predicted probabilities from this model regressing the presence of any additional ties on possible partner loss scenarios, conditional on partner exclusivity. Results show that partner loss due to death and other reasons is significantly associated with a higher probability of having new ties in the core network. Of note, the predicted probabilities are significantly higher for individuals previously in partner-exclusive networks when the partner is gone for reasons other than death (95% vs 81%). That said, the difference between partner-exclusive and nonexclusive networks is not statistically significant when the loss is due to death. From these findings, it appears that approximately 20% of individuals who had solo-member partners end up having no one in their core networks at the follow-up wave.

Figure 1.

Predicted probabilities of acquiring additional tie in different partner loss scenarios (Round dots: Individuals not in partner-exclusive networks; Triangle dots: Individuals in partner-exclusive networks).

The third stage further restricts the sample to older adults starting from partner-exclusive networks (N = 4,892) and examines whether their network rebalancing from the loss is conditional on age and gender. Results reveal that men and individuals at least advanced ages are relatively less likely to replenish their core network after losing a partner in a partner-exclusive setting. Figure 2A and B present the associations between partner loss and the predicted probabilities of having additional tie(s) among individuals in this subsample, conditional on respondents’ age and gender. The likelihood of listing additional core network members for individuals who did not experience partner loss or who lost their partner for reasons other than death slightly declines as age increases. These two groups are significantly different (probabilities just above 60% and 90% for most of the age range, respectively). Meanwhile, the likelihood of mentioning additional core ties increases with age—from about 0.4 to 0.9—if the loss is due to death.

Figure 2.

(A and B) Age and gender differences in the predicted probabilities of acquiring additional ties after partner losses, among individuals in partner-exclusive networks.

In addition, women are more likely than men to replenish their networks when losing their partner, regardless of the reason. When a partner dies, men are slightly less likely than their counterparts with a stable partner presence in the network to mention a new core network tie. However, losing a partner for reasons other than death boosts men’s probability of network replenishment by about 30% relative to those two groups. The model predicts that after the death of a partner, less than 60% of men will add new connections to their previously partner-exclusive core network, while about 90% of women will do the same.

Table 3 shows a descriptive portrait of people’s relationships with newly acquired ties among those listing partners in their core network. Among older adults initially in partner-exclusive networks, those who lost their partner are more likely to add new core ties across all relationship types, though having additional friends is only marginally significant (p < .1) in the case of a partner’s death. Not surprisingly, children are the most prevalent option (70.1% for a partner’s death vs 54.0% for loss for other reasons), followed by other relatives (27.4%/26.2%), friends (18.0%/23.2%), and other non-kins (16.6%/11.0%). It is noteworthy that new partners are also prevalent if the loss was for reasons other than death (28.7%), higher than but still comparable to relatives (26.2%) and friends (23.2%). A comparison between age and gender groups further reveals differences in the acquisition of new core ties after losing one’s sole connection in the context of partner exclusivity. Women are more likely to enlist their children, regardless of the reason for the loss. They are also more likely to incorporate relatives when the partner’s loss was for reasons other than death. As age increases, including children as new core connections becomes more common, whereas adding relatives or friends becomes less likely. It is noteworthy that the probability of adding relatives is least common among individuals in their 60s. Table 4 further highlights that, though older adults losing partners as their only core tie favor proximate replacements, they are also relatively likely to add new members at farther distances, such as more than 5 km or even beyond 25 km from their residence.

Table 3.

Descriptive Statistics About Older Adults’ Relationship With the Confidants Newly Acquired at Wave 6

	Partner	Children	Relatives	Friends	Other non-kin
Other network members available
All respondents	0.018	0.251	0.219	0.190	0.077
Experienced no partner loss (Ref)	0.008	0.247	0.214	0.182	0.072
Partner died	0.017*	0.315***	0.300***	0.232***	0.146***
Partner lost for other reasons	0.206***	0.294***	0.252*	0.322***	0.124***
In partner exclusivity
All respondents	0.024	0.460	0.197	0.145	0.067
Experienced no partner loss (Ref)	0.011	0.447	0.191	0.139	0.061
Partner died	0.006	0.701***	0.274**	0.178	0.166***
Partner lost for other reasons	0.287***	0.540**	0.262**	0.232***	0.110**
In partner exclusivity, experienced loss
Death, men (Ref)	0.017	0.593	0.186	0.203	0.119
Death, women	0.000	0.765*	0.327	0.163	0.194
For other reasons, men (Ref)	0.331	0.476	0.185	0.218	0.105
For other reasons, women	0.239	0.611*	0.345**	0.248	0.115
Death, 50–59 (Ref)	0.000	0.588	0.412	0.235	0.176
Death, 60–69	0.023	0.750	0.250	0.273	0.045
Death, 70+	0.000	0.698	0.260	0.125*	0.219
For other reasons, 50–59 (Ref)	0.276	0.474	0.355	0.316	0.066
For other reasons, 60–69	0.298	0.548	0.167**	0.214	0.083
For other reasons, 70+	0.286	0.597	0.273	0.169*	0.182*

Notes: Proportion differences across the groups are compared with the Chi-square test.

*p < .05. **p < .01. ***p < .001.

Table 4.

Descriptive Statistics About Older Adults’ Distance to Network Members Newly Acquired at Wave 6

	In 5 km	Between 5and 25 km	Above 25 km
Other network members available
All respondents	0.337	0.186	0.203
Experienced no partner loss (Ref)	0.322	0.180	0.197
Partner died	0.453***	0.215	0.266***
Partner lost for other reasons	0.520***	0.280***	0.268***
In partner exclusivity
All respondents	0.443	0.185	0.230
Experienced no partner loss (Ref)	0.421	0.180	0.223
Partner died	0.650***	0.287***	0.312**
Partner lost for other reasons	0.709***	0.232*	0.312***
In partner exclusivity, experienced loss
Death, men (Ref)	0.593	0.271	0.271
Death, women	0.684	0.296	0.337
For other reasons, men (Ref)	0.669	0.226	0.274
For other reasons, women	0.752	0.239	0.354
Death, 50–59 (Ref)	0.588	0.353	0.235
Death, 60–69	0.682	0.341	0.273
Death, 70+	0.646	0.250	0.344
For other reasons, 50–59 (Ref)	0.737	0.289	0.316
For other reasons, 60–69	0.655	0.179	0.298
For other reasons, 70+	0.740	0.234	0.325

Notes: Proportion differences across the groups are compared with the Chi-square test.

*p < .05. **p < .01. ***p < .001.

Discussion

Cultivating new core network ties enhances health and well-being in later life, a period that often consists of multiple life transitions and network losses (Cornwell & Laumann, 2015). The present research extends the growing literature on network transitions at older ages, particularly the question of how people adapt to network losses by incorporating new ties (Atchley, 1989; Cornwell & Laumann, 2018; Cornwell et al., 2021; Lamme et al., 1996; Perry et al., 2018; Wrzus et al., 2013). A core contribution was to highlight network context before a critical life transition, essentially the baseline from which older adults adjust their networks. Specifically, our analysis turns attention to a unique subset of older adults in partner-exclusive core networks. Indeed, our findings show that over a quarter of older adults with a partner tie in their core network are in a position of partner exclusivity (28.1%). In such network configurations, older adults may be especially vulnerable to partner loss. They seemingly face a grim prospect in network replenishment, as this process could hinge on one’s prior level of connectedness (Cornwell et al., 2021). As such, they appear to be at imminent risk of a depleted core network and social isolation, which threaten mental health, physical health, and well-being (Burholt et al., 2020; Litwin & Levinsky, 2021).

Fortunately, results from the present research essentially contradict the more pessimistic expectations related to partner exclusivity and network precarity. Instead, our observations are consistent with both the convoy and continuity perspectives, highlighting that individuals occupying relatively confined, partner-exclusive networks tend to bring additional ties into their core networks following partner loss. The findings imply resilience and flexibility to the social convoy, where older people adapt within what appears to be a restricted network setting—a configuration developed over their adult life course (Antonucci et al., 2010). And continuity also appears to hold, especially in circumstances where people repartner following partner loss (Atchley, 1989; Donnelly & Hinterlong, 2010). That is, incorporating a new spouse or romantic companion can counterbalance an emergent deficit by preserving a prior social role configuration and lifestyle, even if the sustained network structure may appear precarious.

The apparent resilience of those in partner-exclusive networks implies that structural constraints that promote precarious networks and limit network rebalancing are fewer than we anticipated. It is true that men, the childless and financially strained individuals, are most likely to be found in partner-exclusive core networks, potentially resorting to that partner for all-encompassing support and overlooking other potential confidants. Nevertheless, many other factors identified in previous research as predictors of constricted networks fail to explain who maintains a partner-exclusive network. Such factors include low levels of education, retirement, poor physical and mental health, rural residence, cognitive and functional limitations, caregiving obligations, and inactivity in social groups (Djundeva et al., 2019; Fiori et al., 2006; Litwin & Stoeckel, 2013). More importantly, most of these structural constraints appear not to be substantial barriers to acquiring additional core network ties for individuals starting from partner-exclusive networks, except for social inactivity. One interpretation is that, for many older individuals, maintaining a partner-exclusive core network is essentially a lifestyle preference. With a partner no longer available, many may simply adapt by acquiring additional core network ties.

Regarding who gets added to the core network, an adult child represents the most common source of new connections following partner loss, followed by relatives. These findings suggest that family members responsively shift from the periphery to the core of personal networks, keeping people from becoming socially detached (Litwin & Levinsky, 2021). Moreover, such patterns are particularly prevalent among women, indicating that they rely more on family support rather than having lower family involvement when expanding their network (Schwartz & Litwin, 2018). Nevertheless, it is noteworthy that individuals in partner-exclusive networks are disproportionately childless, meaning that this option is not available to many in such networks.

Reflecting that point, a high proportion of people in partner-exclusive networks also go on to add friends and other non-kin members, comparable to that of relatives (14.5% friends + 6.7% other non-kin members vs 19.7% relatives excluding a new partner and children). Though new additions are not primarily non-kin, the evidence supports the argument that friends, neighbors, and others are valuable sources of diversified support that deserve close attention (Ellwardt et al., 2017). Non-kin support could be especially valuable for individuals starting with relatively restricted networks. Still, their contributions may be largely complementary to family sources and somewhat limited for childless individuals (Fihel et al., 2022).

Finally, while proximate new network members are the most common additions, presumably preferred for their accessibility (Small & Adler, 2019), older adults in partner-exclusive networks adapting to tie loss also often enlist connections beyond their locality. Indeed, the geographical distance to these connections could be a reason why many adults found themselves in partner-exclusive networks in the first place—not until the central member is gone are far-off people elevated to the status of a core network member. Altogether, older adults appear to mobilize all types of relationships—kin and non-kin, distant and close by—to rebalance partner loss in a potentially precarious partner-exclusive network setting.

These facts considered, our initial concern about partner-exclusive networks as a form of network precarity still appears partly justified, though principally for men and those in early old age. Indeed, people holding partner-exclusive core networks are particularly vulnerable to a partner’s death, as they are relatively unlikely to acquire additional core connections and are thus more susceptible to ending up isolated. These findings correspond with other research showing that men typically benefit more from marriage than their wives (Dykstra & De Jong Gierveld, 2004; van den Broek, 2017), maintain fewer close relationships with their children than women, and demonstrate less resilience in their social networks following marital disruptions (e.g., Dykstra & De Jong Gierveld, 2004; Kalmijn, 2007, 2012; Lamme et al., 1996; Schwartz & Litwin, 2018).

One potential limitation of the present research is that the 4-year window between the two waves is potentially short for adjustments to partner disruption. Research on network adjustments to life events in later life could benefit from tracing network changes over a longer time with more waves of data collection. Individuals’ adjustments to marital disruptions are often not linear, and the coping strategies often vary across individuals and situations. For example, contact and support are likely to intensify immediately after a spousal loss, typically from existing relationships; however, such activity usually declines after about 2.5 years, while individuals typically increase other social participation a few years after the transition (Ferraro, 1984; Guiaux et al., 2007). Individuals who have recently experienced partner transitions may not have fully accommodated the change and rebalanced their network by the subsequent wave.

In addition, the current data do not depict broader social connectedness beyond the respondents’ core networks. Though widely used, the type of name-generator technique used in SHARE is potentially limited in representing more peripheral ties, such as casual acquaintances and contacts providing specific forms of support (Cornwell et al., 2021). Various forms of embeddedness with these peripheral ties may endow partner exclusivity with different meanings. For instance, some individuals have no meaningful social relationships aside from their partners. Meanwhile, others with partner-exclusive core networks have contacts in their periphery but are highly selective in identifying core network members and thus only list their partners. Peripheral links could fulfill companionship needs or represent caches of available support, ultimately facilitating social connectedness while slipping by our measurement tool.

Overall, the present research highlights that network replenishment is likely after a partner disappears from older adults’ networks—even when that partner is the sole member of the core network. It provides evidence of rebalancing amidst a structural configuration that makes it seem unlikely. We show that aging individuals holding partner-exclusive networks—networks arguably most prone to destabilization—often end up not in isolation but in a position to rebuild their core networks, primarily with families but often also with friends. Nevertheless, families should consider proactively supporting widowers and individuals who lost their partner to death early on, as they could run short of supportive companions and confidants if they overly relied on the partner. Future studies examining network rebalance should further contextualize the process in other life transitions, pay attention to the variations across groups, and identify the most vulnerable to isolation amidst life transitions and crises.

Funding

None declared.

Conflict of Interest

None declared.

Author Contributions

H. Sun initiated the idea, conducted the literature review and data analysis, and assembled the manuscript. M. H. Schafer collaborated with conceptualization, refined the study framing, helped structure the literature review, supervised the data analysis, and assisted with writing.