Social Networks in Later Life: Similarities and Differences between Sexual-Minority and Heterosexual Older Adults

Abstract

Community-based research suggests that lesbian, gay and bisexual (LGB) older adults are more socially isolated than their heterosexual counterparts. However, little is known about how social networks in late adulthood differ between LGB and heterosexual people at the population level. Using data from the 2015–16 National Social Life, Health and Aging Project (N=3,929), we compare the size, frequency of contact, composition, diversity, and density of core discussion networks as well as family and friend support across sexual orientation groups. We find that LGB people share a few network characteristics with their heterosexual counterparts, including network size, frequency of contact with network members, and proportion of non-spousal kin members in the network. However, their networks are less likely to include an intimate partner, are more likely to include friends, and have lower diversity and density. LGB people also report lower family support but higher friend support than heterosexual people.

INTRODUCTION

Social networks provide instrumental and emotional support, offer moral and behavioral guidance, and create a sense of belonging and meaning (Durkheim 1979; Turner and Turner 2013). It is widely acknowledged that the structural and qualitative characteristics of social networks are relevant to the health and well-being of older adults (Berkman and Glass 2000; House, Landis, and Umberson 1988; Umberson and Montez 2010). Larger networks, higher density networks (in which network members tend to know one another), more frequent interaction with network members, and higher levels of perceived or received support, for instance, are all associated with better mental and physical health outcomes (Cornwell, Laumann, and Schumm 2008; Cornwell and Waite 2009). Although the health implications of social networks are well researched, few studies have examined how social networks vary by sexual orientation using population-representative samples.

Some studies using non-representative, community-based samples have suggested that sexual minorities tend to face more challenges to building and maintaining social relationships over the life course (Erosheva et al. 2016; Fredriksen-Goldsen and Muraco 2010; Grossman, D’Augelli, and Hershberger 2000). For example, legal barriers and social stigma have made it difficult for sexual minorities to establish or nurture spousal and family relationships (Cronin and King 2014; Hsieh and Liu 2021). As individuals age, social networks often become smaller, more family-centered, and less diverse (contain fewer different kinds of relationships) (Carr and Moorman 2011; Cornwell et al. 2008). Thus, we might expect that older sexual minorities are particularly disadvantaged in their social networks compared to heterosexual older adults. But because there has been little population-level data on older sexual minorities, it is unclear whether and how network structure differs between older sexual minority and heterosexual adults and whether older sexual minorities are at significantly higher risk of social isolation compared to their heterosexual counterparts.

The current study is the first to compare egocentric social networks – network ties around an individual – between older sexual minority and heterosexual adults based on a probability sample of non-institutionalized older adults in the United States. Using data from the 2015–16 National Social Life, Health and Aging Project (NSHAP), a nationally representative study, we examine the size, frequency of contact, composition (e.g., proportion of kin vs. friends), diversity, and density of older adults’ core discussion networks—the web of confidants that individuals often rely on as primary sources of support (Burt 1984; Marsden 1987). Additionally, we assess the levels of family and friend support from older adults’ general social networks to assess the overall quality of their networks.

In contrast to most prior research using non-probability samples to study sexual minorities’ social relationships, the present study attends to the subject at the population level and overcomes the issue of oversampling socially active participants (Grossman et al. 2000). Furthermore, unlike prior community-based studies focusing only on sexual minorities, this study examines differences in network characteristics between sexual minority and heterosexual older adults. Therefore, findings can inform program planning in social services and elder care regarding the social needs of sexual minorities in comparison to heterosexual adults at older ages.

LITERATURE REVIEW

Social Networks in Later Life

Individuals experience many transitions in their social roles and activities as they age, and these transitions often alter the structure and quality of their social networks. Many studies have shown that narrowing role sets due to retirement, health deterioration, and bereavement in later life tend to limit older adults’ social lives (Carr and Moorman 2011; Cornwell et al. 2008). Individuals maintain smaller social networks and lower frequency of contact with network members at older ages (Cornwell et al. 2008; Marsden 1987; McPherson, Smith-Lovin, and Brashears 2006). Network composition – whether the network contains higher proportions of family members relative to friends and other non-kin – also changes with age. Reducing the number of roles one assumes as a colleague, friend, partner, and/or caregiver often decreases the diversity of social networks. Compared to younger adults’ networks, older adults’ networks tend to be more kin-centered, more engaged with spouses, children, and perhaps siblings, while being less engaged with friends, coworkers, and other nonrelatives (Carstensen 1992; de Jong Gierveld et al. 2009).

Although declines in the size, frequency of contact, and diversity of social networks suggest a more isolated life and poorer well-being, other aspects of network change may compensate for this loss. Previous research has argued that while networks in late adulthood are smaller and less diverse than those in early and middle adulthood, they have higher density (i.e., network members are more likely to know one another) and better quality (Carstensen 1992; Cornwell et al. 2008; Marsden 1987). Partly because older adults’ networks consist of a higher portion of family ties, the networks are denser, thereby forming close-knit social contexts where informational, emotional, and instrumental resources can be easily pooled and shared. In addition, older adults’ relationships may be more supportive and emotionally rewarding than younger adults’ relationships because people select and trim their networks based on the levels of support and closeness provided by network ties over the life course (Carstensen 1992; Carstensen, Isaacowitz, and Charles 1999). More peripheral, superficial, or conflictual ties are often dropped, whereas more supportive and intimate ties remain in or are added to networks as people age. This could potentially enhance older adults’ access to support and wellbeing. However, it is possible that denser networks are not wholly positive. Some studies note that in circumstances where the needs of care arise, caregiver burden that is primarily shared by a small and dense network may increase stress and negative interaction (Fingerman et al. 2008; Wong and Hsieh 2019).

Differences in Social Networks by Sexual Orientation in Later Life

Social network characteristics may vary systematically across sexual orientation due to stigmatization of sexual minority identities and behaviors. Many older sexual minorities struggle or have struggled in relationships with their families of origin, intimate partners, children from previous heterosexual marriages, coworkers, religious communities, and other community members in ways that older heterosexuals do not (Cronin and King 2014; Goldberg 2007; Hsieh and Liu 2021; Kim and Fredriksen-Goldsen 2016; King 2016). Previous studies have noted that older sexual minorities are more likely to experience estrangement or conflict with biological family members because of their sexuality (Cronin and King 2014; Grossman et al. 2000; Kuyper and Fokkema 2010). This may lead to smaller kin networks, lower frequency of family contact, and poorer family support. Additionally, due to legal restrictions on same-sex marriage and a lack of cultural support for same-sex relationships in the past, older sexual minorities are less likely to live with a marital or intimate partner and are less likely to have children than their heterosexual counterparts (Kim and Fredriksen-Goldsen 2016; Kuyper and Fokkema 2010; MetLife 2010). Because partners and children often play the most important roles in older adults’ social lives through providing company and care (Cantor 1979; Carstensen et al. 1999), older sexual minorities are at greater risk of social isolation, potentially lacking contact and support relative to older heterosexual adults.

Although sexual minorities may have smaller kin networks, they may develop larger friend networks to compensate for the lack of family and partner support (Grossman et al. 2000; MetLife 2010; de Vries and Hoctel 2007). Prior studies based on convenience samples have suggested that sexual minorities often refer to their close friendships as chosen families, who they can turn to not only for socializing, but also for help in the face of crisis (de Vries and Hoctel 2007; Wilkens 2015). For example, the HIV/AIDS epidemic in the 80s and 90s mobilized and strengthened friend support in LGBTQ communities, and friends substituted for families in meeting the daily needs of care and household maintenance (Nardi 1997). Friends, in general, also provide some of the strongest coping resources to buffer the impact of sexuality-based stigma on identity, personhood, and self-worth (Grossman et al. 2000; Nardi 1982; de Vries and Hoctel 2007). However, a recent national study of older adults has noted that despite reporting more support from friends, LGB older adults may not have a larger number of friends than their heterosexual counterparts (Hsieh and Liu 2021). This may be because some of their friendships were lost to the HIV/AIDS epidemic, and thus remaining friendships may not completely make up for the lack of family and partner relationships. Therefore, we expect that the size of LGB older adults’ core networks may be smaller, and that their frequency of contact with core network members may be lower, compared to heterosexual adults. However, other network characteristics may differ by sexual orientation due to fewer spousal and kin ties as well as more friends in the networks. In particular, sexual minorities’ network composition (e.g., proportion of kin vs. friends) may differ from that of heterosexual individuals. Their network diversity (number of different kinds of relationships) may be lower because of a lack of spousal and kin ties. Their network density (whether network members know one another) may also be lower because of having a friend-centered network, which is more likely to include network members who do not know one another than a family-centered network.

Previous research provides limited evidence for the similarity or difference in social networks across sexual orientation groups because it examines only sexual minorities, particularly those who actively participate in community organizations. To address the gap in knowledge, we develop the following hypotheses to guide our comparison of the structural and qualitative characteristics of social networks between older sexual minority and heterosexual adults using data collected at the population level:

Hypothesis 1:

Lesbian, gay, and bisexual (LGB) older adults’ core network size is smaller than that of heterosexual older adults.

Hypothesis 2:

LGB older adults contact their core network members less frequently than heterosexual older adults do.

Hypothesis 3:

LGB older adults’ core networks are less likely than heterosexual older adults’ networks to include an intimate partner, and include fewer kin members but more friends than heterosexual older adults’ networks.

Hypothesis 4:

LGB older adults’ core networks are less diverse than heterosexual older adults’ networks.

Hypothesis 5:

LGB older adults’ core networks are less dense than heterosexual older adults’ networks.

Hypothesis 6:

Considering the quality of social networks, older LGB people have lower family support, but higher friend support than their heterosexual counterparts.

RESEARCH DESIGN AND METHODS

Sample

This study uses data from the 2015–16 National Social Life, Health, and Aging Project (NSHAP), when a sexual identity question was first introduced to its questionnaire. NSHAP is a nationally representative study of social relationships and health among community-dwelling older adults in the U.S. The questionnaire is divided into two parts: the main in-person questionnaire (IPQ), which is conducted by face-to-face interview, and the leave-behind questionnaire (LBQ), which is self-administered and was returned by 84% of the IPQ sample respondents. Social network information is collected from all respondents in the IPQ, but because the sexual identity question is located in the LBQ, our study focuses on respondents who completed the LBQ and had valid responses to key variables and covariates (N=3,929). Among this sample, 3,484 respondents self-identified as heterosexual, 81 respondents self-identified as LGB, and 364 respondents did not answer the sexual identity question. More specifically, the LGB sample is composed of 16 lesbian women (20% of LGB sample); 28 gay men (34%); 16 bisexual women (20%); and 21 bisexual men (26%). All the respondents were at least 50 years old at the time of the survey.

Variables

We examine nine network variables, seven of which measure the structural characteristics of core discussion networks and two of which measure the qualitative characteristics of overall family and friend networks. For structural characteristics, network size indicates the number of people listed in the respondent’s core discussion network. In the NSHAP network module, respondents were asked to name up to five people with whom they discussed things that are important to them. This module provides the basis for all our structural measurements. Frequency of contact measures the total volume of contact with the members of core discussion networks. Respondents were originally asked how often they contacted each network member per year; eight possible responses were available, from “less than once a year” to “every day.” Following Cornwell et al. (2008), we transformed the response to the estimated number of days of contact with each network member and then summed the estimates across all members to get the total volume of contact. Specifically, we coded “less than once a year” as 0.5 days, “once a year” as 1 day, “a couple times a year” as 2 days, “once a month” as 12 days, “once every two weeks” as 26 days, “once a week” as 52 days, “several times a week” as 182 days (3.5 days per week multiplied by 52 weeks in a year), and “every day” as 365 days. Alternative coding schemes yielded consistent results (robustness checks are available upon request). Partner in network is a dummy variable indicating whether the respondent listed a marital, romantic, or sexual partner in the core discussion network. Proportion kin in network indicates the share of network members who are non-spousal relatives. Proportion friends in network indicates the share of network members who are friends. Network diversity is measured by the number of unique kinds of relationships in the core discussion network including: spouse, romantic, or sexual partner; parent; sibling; child or grandchild; other relatives; friend; and community members, professionals, or other non-relatives. Network density measures the likelihood of respondents’ core discussion network members knowing one another. It is expressed as a proportion calculated by taking the number of relationships between network members that actually exist divided by all possible relationships in the network.

The qualitative characteristics of older adults’ overall family and friend networks are measured by family and friend support. Family support is measured by the average score of the following three items, all of which are on a 4-point scale (0=never, 1=hardly ever or rarely, 2=some of the time, 3=often): how often could respondents (1) open up to and (2) rely on their family members, and how often their family members (3) really understand the way they feel (alpha=0.76). Note that family members here do not include marital, romantic or sexual partners. Friend support is measured in the same way as family support, with three items parallel to those assessing family support (alpha=0.80).

Sexual identity indicates whether the respondent self-identified as heterosexual or straight, self-identified as lesbian, gay, or bisexual (LGB), or did not report a sexual identity. Because the LGB sample is small (N=81), to have enough statistical power we do not further separate them into lesbian, gay, and bisexual subgroups (though we discuss exploratory analyses of sexual orientation and gender below). Since older sexual minorities are less likely to openly identify themselves as “gay,” “lesbian,” and “bisexual” due to fear of stigma and discrimination (Institute of Medicine 2011), some sexual minority respondents may choose not to disclose their sexual identity in the survey context. In our unweighted sample, 9% of the respondents (N=364) refused to report an identity. We keep those who did not answer the sexual identity question in the analysis to avoid sample selection bias and to learn about whether they share network and demographic characteristics with LGB or heterosexual people.

In the regression analyses, we control for age (in years), gender (men or women), race/ethnicity (White vs. non-White), and education (high school or less vs. more than high school). We include these demographic factors in the analyses because they are associated with social network characteristics (Cornwell et al. 2009). More detailed answer categories for the race and education variables are presented in Table 1, but they are collapsed into dichotomous controls in the regressions to avoid small cells in the LGB sample.

Table 1.

Descriptive Characteristics of the Sample by Sexual Orientation, N=3,929

	Heterosexual (N=3,484)	LGB (N=81)	Not Reported (N=364)
Age (range: 50–97)	64.0	62.2a	67.4ab
Race/ethnicity
White	80.1%	66.1%a	49.4% a
Black	10.2%	13.3%	16.5%
Hispanic	5.8%	11.2%	27.7% a
Other	3.8%	9.4%	6.5%
Education
<HS	8.5%	5.0%	35.9% ab
HS	23.1%	11.8%	25.2%
Vocational	37.1%	43.3%	23.5% ab
Bachelors or more	31.3%	39.9%	15.3% ab
Gender (Women)	54.2%	43.1%	63.4% ab
Network size (range: 0–5)	3.9	3.7	3.5a
Frequency of contact (range: 1–1,825)	756.1	746.9	785.9
% with any partner in network	60.9%	44.9%	39.7% a
% Kin in network	45.3%	43.6%	48.8%ab
% Friends in network	27.2%	38.1%a	27.9%b
Network diversity (range: 1–5)	2.8	2.3a	2.3a
Network density (range: 0–1)	0.8	0.7	0.8
Family support (range: 0–3)	2.3	2.1a	2.3b
Friend support (range: 0–3)	2.1	2.2	1.9ab

^aindicates significant difference from heterosexual at p<0.05

^bindicates significant difference from LGB at p<0.05

Analytic Plan

We first use descriptive bivariate analyses (ANOVA, chi-squared tests) to examine similarities and differences in demographic and network characteristics across heterosexual individuals, sexual minority individuals, and those who did not report a sexual identity. Then, we use multivariate regression analyses to understand whether differences in social networks persist after controlling for the demographic characteristics described above. For ease of interpretation, we use OLS regression to test differences in network characteristics between sexual orientation groups for network size, frequency of contact with network members, proportion of kin network members, proportion of friend network members, network density, network diversity, family support, and friend support. But because many of these network characteristic measures may not comply with the assumptions of OLS regression, we also check the robustness of the results using tobit regression for censored outcomes (size, diversity, and family and friend support) and generalized linear regression models with logit link and the binomial family for proportional outcomes (proportion kin, proportion friends, and density). All robustness checks produce results that are consistent with findings from the OLS models (robustness checks are not presented but are available upon request). Lastly, we use logistic regression for the analysis of whether respondents report having a marital/romantic/sexual partner in the network.

Our analytic sample varies modestly across analyses primarily because the observation of most dependent variables (except network size) is conditional on having at least one or two network members. For example, the analysis of network density includes the smallest sample size (N=3,626) because network density can only be measured for respondents reporting two or more network members. All analyses are conducted in Stata 16 and include survey weights to generalize to the community-dwelling population of older adults aged 50 and up in 2015–16.

RESULTS

Descriptive Statistics

Table 1 shows that demographic characteristics vary significantly between older adults of different sexual orientations. LGB adults are younger (mean age=62.2) than heterosexual adults (64.0) and those who did not report their sexual identity (67.4). LGB people (66.1%) and people who did not report a sexual identity (49.4%) are less likely to identify as White than heterosexual people (80.1%). Those who did not report a sexual identity are also more likely to identify as women (63.4%) than heterosexual (54.2%) and LGB people (43.1%). On average, LGB adults are the most likely to complete a bachelor’s degree (39.9%) compared to adults with a heterosexual identity (31.3%) and those who did not report a sexual identity (15.3%). These descriptive results not only show anticipated demographic differences between LGB and heterosexual individuals, but also suggest that older adults who did not respond to the sexual orientation question are demographically different from LGB- or heterosexual-identified older adults.

Quite a few social network characteristics vary across heterosexual and LGB individuals and those who did not report a sexual identity in the bivariate analyses. Overall, heterosexual people exhibit some advantage in network size, diversity, and family support in comparison to the other groups. In particular, heterosexual people have significantly larger core discussion networks (3.9) than people who did not report a sexual identity (3.5), although no significant difference in network size is detected between heterosexual and LGB people (3.7). They have the most diverse networks (2.8) compared to both LGB adults (2.3) and people who did not report a sexual identity (2.3). In addition, heterosexual people report higher levels of family support (2.3) than LGB people (2.1) although their support is similar to the level reported by people who did not report a sexual identity (2.3).

Network composition also differs across the three groups. Heterosexual people are more likely to have a partner in their core discussion networks (60.9%) than LGB people (44.9%) and people who did not report a sexual identity (39.7%). However, those who did not report their sexual identity (48.8%) have a higher proportion of kin ties (excluding the spouse) in their networks than both heterosexual (45.3%) and LGB individuals (43.6%). By contrast, LGB people (38.1%) have a higher proportion of friends in their networks than heterosexual people (27.2%) and those who did not report a sexual identity (27.9%). However, LGB people (2.2) only report more friend support compared to people who did not report their sexual identity (1.9); their friend support is similar to the level reported by heterosexual people (2.1). Finally, despite the above differences, the total frequency of contact with network members and network density do not differ significantly between the three groups.

Overall, the descriptive findings demonstrate that heterosexual and LGB-identified people are qualitatively different from those who do not reveal their sexual identity in the survey context. Because the current study focuses on the comparison of LGB and heterosexual older adults, in the following regression analysis we center our discussion on these two groups. However, results for those who did not report their sexual identity remain in the tables for interested readers.

Regression Analyses

The differences in social network characteristics described above largely remain significant when age, gender, race/ethnicity, and education are adjusted for in regression analyses (see a comparison of Panels A and B in Table 2). Inconsistent with Hypotheses 1 and 2, the size of networks is similar across LGB and heterosexual individuals (Model 1), as is the total frequency of contact with network members (Model 2). However, the composition of networks differs for LGB compared to heterosexual individuals. In particular, LGB people have lower odds of having a marital, romantic, or sexual partner in their core discussion networks compared to heterosexual people (Model 3). The proportion of friends in the network is also significantly higher among LGB than heterosexual adults (Model 5). But, to our surprise, the proportion of non-spousal kin in network is equivalent across LGB and heterosexual adults (Model 4). These results together partially support Hypothesis 3.

Table 2.

OLS (Models 1, 2, 4–9) and Logistic (Model 3) Regressions Predicting Network Characteristics Across Sexual Identity Group

Panel A: Unadjusted Regression Coefficients	Model 1: Network Size	Model 2: Frequency of Contact	Model 3: Any Partner in Network (Odds Ratio)	Model 4: Proportion Kin in Network	Model 5: Proportion Friends in Network	Model 6: Network Diversity	Model 7: Network Density	Model 8: Support from Family	Model 9: Support from Friends
Sexual Orientation (ref. heterosexual)
LGB	−0.25 (0.15)	−9.17 (41.69)	0.52* (0.16)	−0.02 (0.04)	0.11** (0.03)	−0.42** (0.14)	−0.08+ (0.04)	−0.27** (0.09)	0.17* (0.07)
Not reported	−0.41*** (0.09)	29.58 (28.46)	0.42*** (0.06)	0.04 (0.03)	0.01 (0.02)	−0.49*** (0.06)	−0.00 (0.02)	−0.05 (0.05)	−0.23*** (0.06)
Constant	3.90*** (0.04)	756.88*** (9.35)	1.56*** (0.08)	0.45*** (0.01)	0.27*** (0.01)	2.76*** (0.03)	0.80*** (0.01)	2.32*** (0.01)	2.08*** (0.02)
Panel B: Adjusted Regression Coefficients	Model 1: Network Size	Model 2: Frequency of Contact	Model 3: Any Partner in Network (Odds Ratio)	Model 4: Proportion Kin in Network	Model 5: Proportion Friends in Network	Model 6: Network Diversity	Model 7: Network Density	Model 8: Support from Family	Model 9: Support from Friends
Sexual Orientation (ref. heterosexual)
LGB	−0.23 (0.16)	−11.54 (42.35)	0.41* (0.15)	−0.00 (0.04)	0.11** (0.03)	−0.42** (0.12)	−0.08* (0.04)	−0.24* (0.09)	0.20** (0.07)
Not reported	−0.24** (0.09)	14.45 (26.61)	0.69* (0.11)	−0.02 (0.03)	0.02 (0.03)	−0.28*** (0.07)	−0.01 (0.02)	−0.01 (0.05)	−0.12* (0.06)
Age	−0.00 (0.00)	−4.42*** (0.82)	0.95*** (0.00)	0.00*** (0.00)	0.00* (0.00)	−0.02*** (0.00)	0.00 (0.00)	0.00* (0.00)	−0.00** (0.00)
Non-White (ref. White)	−0.25*** (0.06)	53.60** (19.40)	0.49*** (0.05)	0.06*** (0.01)	−0.02 (0.01)	−0.27*** (0.05)	0.02 (0.01)	−0.10** (0.03)	−0.24*** (0.03)
More than High School (ref. HS or less)	0.44*** (0.06)	−4.50 (11.67)	1.60*** (0.13)	−0.06*** (0.01)	0.05*** (0.01)	0.23*** (0.04)	−0.03* (0.01)	0.11** (0.03)	0.17*** (0.03)
Women (ref. men)	0.42*** (0.05)	127.48*** (13.67)	0.43*** (0.04)	0.09*** (0.01)	0.04*** (0.01)	0.24*** (0.03)	−0.04** (0.01)	0.18*** (0.02)	0.31*** (0.02)
Constant	3.48*** (0.25)	962.87*** (61.05)	55.70*** (19.14)	0.26*** (0.04)	0.13** (0.04)	3.65*** (0.15)	0.80*** (0.04)	1.95*** (0.09)	2.07*** (0.08)
N	3,929	3,884	3,896	3,896	3,896	3,896	3,626	3,915	3,898

Note:

⁺p<0.1

^*p<0.05

^**p<0.01

^***p<0.001. All models are estimated by OLS regressions, except for Model 3, which is estimated by logistic regression. Results from Model 3 are odds ratios; all other results are unstandardized coefficients. LGB=lesbian, gay, and bisexual.

The number of unique relationship types in the network varies across sexual orientation groups, in line with Hypothesis 4. LGB people have lower network diversity than heterosexual people (Model 6). Further, in support of Hypothesis 5, network density is also lower among LGB people than heterosexual people (Model 7). This finding suggests that network members are less likely to know one another in sexual minorities’ networks. Finally, as predicted in Hypothesis 6, family support and friend support both differ significantly across LGB and heterosexual individuals. Compared to heterosexual adults, LGB adults report lower levels of family support (Model 8) but higher levels of friend support (Model 9).

DISCUSSION

As the first national study using population-based data to examine social networks across sexual orientation groups in late adulthood, we show that LGB-identified older adults share a few characteristics of their core discussion networks with their heterosexual counterparts. In particular, core network size, frequency of contact with core network members, and proportion of non-spousal kin members in the network are similar for LGB and heterosexual individuals. Although previous studies, based primarily on non-probability samples, have suggested overall network disadvantage among LGB people, our results show more similarity in network structure between LGB and heterosexual people. In particular, we find that older LGB adults have a similar number of close ties to rely on and contact their confidants as frequently as older heterosexual adults do. This is likely because friendships compensate for the lack of partner or family support among older LGB adults as noted in prior studies (Grossman et al. 2000; de Vries and Hoctel 2007). In addition, to our surprise, we find that LGB adults have a similar proportion of kin members in their core networks. These findings imply that despite having more difficulty maintaining family relationships due to a nonheterosexual identity (Cronin and King 2014; Kuyper and Fokkema 2010), older LGB adults keep close family ties in their networks like their heterosexual counterparts do.

Despite such similarity, there remain several differences in the structure and quality of core networks across LGB and heterosexual older adults. Notably, LGB adults’ core networks are less likely to include an intimate partner but are more likely to include a friend than heterosexual adults’ networks. As previous studies have indicated, cultural disapproval of same-sex relationships and historical legal restrictions on same-sex marriage have led to lower partnership rates among sexual minorities (Kim and Fredriksen-Goldsen 2016; Kuyper and Fokkema 2010). Lacking an intimate partner in one’s social life may increase the relative weight of friendships in it. Likely due to differences in partnership status and having more friend-centered networks, LGB adults’ core networks also have lower diversity (fewer unique types of relationships) and lower density (fewer members who know one another).

Together, these findings have several implications. Because partners play a central role in socializing, caregiving, and health facilitation in late adulthood (Reczek et al. 2018; Thomeer, Reczek, and Umberson 2015; Umberson, Donnelly, and Pollitt 2018), older LGB people may be at higher risk of loneliness and health decline even with additional friend support (Hsieh and Liu 2021). For example, different kinds of relationships tend to provide different forms of social support, and more diverse networks are linked to better well-being such as reduced loneliness and anxiety in later life (Litwin and Shiovitz-Ezra 2011). Lower diversity in LGB adults’ core networks may imply poorer well-being. Finally, denser networks can provide a closer-knit safety net where informational, emotional, and instrumental resources can be potentially pooled and exchanged, thereby offering stronger support to older adults (Cornwell et al. 2008). Lower density in LGB adults’ core networks may suggest fewer accessible resources, though it is possible that less dense networks come with fewer strains that can accompany high levels of closeness (Fingerman et al. 2008).

Moreover, our analysis shows that older LGB adults have lower family support but higher friend support than older heterosexual adults. These results, along with those discussed above, indicate that although older sexual minorities and heterosexual adults report an equivalent core network size and a similar proportion of kin members in their core networks, sexual minorities still receive (or perceive) lower levels of support from family members than heterosexual adults do. This suggests that sexual orientation may matter more to the quality of family relationships than to the quantity of close family ties. It is also consistent with previous community-based studies showing that sexual minority individuals experience more family conflict than do heterosexual individuals (Cronin and King 2014; Grossman et al. 2000). Fortunately, older LGB adults receive higher levels of friend support than older heterosexual adults, potentially balancing their lower levels of family support. This confirms the important role friendship plays in the late life of sexual minorities as noted in prior research (de Vries and Hoctel 2007). Overall, these findings imply that social services and elder care services that assume the existence of spousal or family support may not work for LGB older adults. But programs that leverage LGB adults’ greater friend ties and friend support would be more successful in promoting their wellbeing.

Finally, we find that older adults who did not report their sexual identity are different from LGB or heterosexual older adults in multiple demographic and network traits. Overall, they are demographically more marginalized—older, less educated, and more likely to identify as Hispanic and as women—compared to the other two groups. Unlike LGB older adults, they report more reliance on family than friends. Although we are uncertain about the extent to which this group contains members of sexual minorities, our findings suggest that missing data on sexual identity in the older population is likely linked to advanced age and poor access to resources.

This study has a few limitations. First, because the sample size for LGB older adults is small, our analysis may not have enough power to detect all the differences between LGB and heterosexual older adults. The sample also limits us from examining differences between sexual minority subgroups. In exploratory analyses, we separate gay/lesbian and bisexual individuals but find few differences between these two sexual minority groups (results available upon request). Additionally, we test how gender, age, race, and education interact with sexual orientation to influence network characteristics. We do not detect statistically significant differences in the relationship between sexual minority status and network characteristics across gender, age, or race, but find some evidence that network characteristics differ between more educated and less educated LGB people. Specifically, more educated LGB adults have lower kin proportions, lower network density, and less family support, but have higher friend proportions in network than their less educated counterparts (results available upon request). These findings suggest that social life may vary by educational attainment within the LGB older adult population. The current study is only a first step to use population-based data to examine network differences by sexual orientation. Future research should fully explore variation across sexual orientation and associated sociodemographic characteristics when larger datasets become available. Moreover, while core discussion networks represent the primary sources of support for many people (Cornwell et al. 2008; Marsden 1987), they do not usually encompass more peripheral social ties such as neighbors and religious community members that may provide additional instrumental or psychosocial resources for older adults (Erosheva et al. 2016; Fischer and Offer 2020). Relatedly, the network module in NSHAP only allowed respondents to nominate up to five people with whom they discuss important things, and thus information about additional confidants, if they exist, is excluded from our analysis. This may be one of the reasons why we find no differences in some network characteristics, such as size frequency of contact, by sexual orientation. It is likely that more variation would appear if respondents were allowed to nominate more than five confidants. Whether the structure or quality of extended social networks differs by sexual orientation needs further exploration in future research.

CONCLUSION

Despite the above limitations, our study advances the knowledge of social networks in late adulthood by comparing older adults of different sexual identities. Our findings highlight that compared to heterosexual older adults, LGB older adults have some network disadvantages, including lack of a partner, lower network diversity and density, and less family support. But they have similar network size, frequency of contact with network members, and proportion of kin members in the network, as well as an important network strength—higher friend support. With marriage equality, we expect that the gap in having a partner in a network will gradually close in the future, and so will network diversity. We also anticipate that the disparity in family and friend support may diminish when non-heterosexual identities become more broadly accepted and culturally supported. Nevertheless, friendship in older age may be an underutilized source of support for heterosexual individuals. In an aging society with declining fertility, older LGB adults’ experiences can inspire all of us to be creative at gathering alternative sources of support.