Relationship Quality and Sexuality: A Latent Profile Analysis of Long-term Heterosexual and LGB Long-term Partnerships

ABSTRACT

Purpose: Drawing on survey data (N = 7,826) collected in the United Kingdom, Australia, and the United States, this paper examines whether sexual orientation is a differentiating factor in explaining relationship quality and maintenance. Previous research suggests that sexual orientation is not significant in determining relationship satisfaction; however, these analyses have used traditional variable driven approaches, which do not provide an holistic view of the relationship by considering the unique combination of characteristics.

Method: In this study, latent profile analyses were used, which is a person-centered approach that allows for identification of different types of long-term relationships.

Results: Data suggested that LGB individuals had marginally higher levels of relationship quality compared to their heterosexual counterparts, and sexual orientation was also associated with differing types of long-term relationships.

Discussion: These findings are interrogated in more detail, in particular, how sexual orientation is associated with types of long-term relationships and how everyday practices are associated with relationship quality.

KEYWORDS:

• Relationship quality
• couple relationships
• sexuality
• sexual orientation
• relationship types
• latent profile analyses

Over the past 20 years, intimacy, personal life, and family relationships have been the focus of critical and contested examination. Social theorizing has suggested that in the context of contemporary self-help culture (Giddens, 1992), couple relationships are analyzed, scrutinized, and ultimately abandoned because of increased reflexivity, and this has led to a weakening of the meaning of commitment (Bauman, 2003; Beck, 2000a). Yet, a large proportion of the population is coupled, and a happy intimate relationship plays a significant role in personal well-being. For example, a recent large-scale study interested in “the new science of happiness” (Kroll, 2014, p. 1) found that “having a satisfying romantic relationship is important for retaining and increasing future life satisfaction” (Dyrdal, Røysamb, Nes, & Vittersø, 2011, p. 947). Longitudinal studies point to the importance of sexual satisfaction in maintaining relationship satisfaction in heterosexual couples (McNulty, Wenner, & Fisher, 2016), for example. Although sexual identity is an important consideration in assessing relationship quality (Heaphy, Smart, & Einarsdottir, 2013; Weeks, Heaphy, & Donovan, 2001), same-sex relationship satisfaction remains underrepresented in research.

Despite common – and culturally dated – stereotypes, research has shown that many lesbians and gay men form long-term cohabiting couples (Bryant & Demian, 1994). It is estimated that 40–60% of gay men and 45–80% of lesbians are currently involved in a committed romantic relationship (Kurdek, 1995). Whilst marriage rates in the UK are in decline for heterosexual couples, cohabitation rates have doubled over the past decade, and the married couple with or without children remains the most common type of relationship unit (ONS, 2018). Similar patterns can also be found in the U.S. (Herek, 2006; Lenhart & Duggan, 2014) and Australia (Qu & Weston, 2011). In the UK civil partnership was introduced in 2005 and same-sex marriage in 2014. In the U.S., same-sex couples gained access to same-sex marriage in 2015 after a Supreme Court decision. In Australia the legal right for same sex couples to marry was passed in Parliament in December 2017. Internationally, same-sex partnership legislation is thus relatively new, however, marriage appears to be a notable feature of the relationship trajectory for the majority of young lesbian and gay couples (Gabb, 2019; Heaphy et al., 2013). The impact of same-sex partnership recognition on relationship duration therefore remains unknown, and comparable heterosexual and same-sex long-term marital data are not yet available.

The enduring appeal of couple relationships is evident across the sexual identity spectrum (Gabb & Fink, 2015). Literature suggests that stressful and discriminating social-cultural contexts do still adversely impact on relationship quality through the embedded obstacles, biases, and stigma that persist in heterosexist cultures (Connolly, 2005). Money worries are an additional stressor for couples (Badgette, 2001), and relationship status (legal union vs cohabitation) is significant here (Solomon, Rothblum, & Balsam, 2005). The adverse impact of income inequality, diminished social support, and attachment insecurity are also likely to be mediated by the extent to which a gay identity is self-accepted (Elizur & Mintzer, 2003).

Legal partnership opportunities therefore appear to have a positive impact on the quality of long-term couple partnerships (King & Bartlett, 2005) as this potential for increased and recognized stability in same-sex relationships can enhance physical and mental health (Degges-White & Marszalek, 2006) and generate wider social acceptance in neighborhoods and families of origin (Duncan & Phillips, 2008). Lower levels of well-being and relationship satisfaction have thus been attributed to the lack of access to formal civil union or marriage (Wilkins, 2015). Indeed same-sex couples not in civil unions are more likely to end their relationships than those who are or heterosexual married couples; however, compared with heterosexual married participants, married and unmarried same-sex couples reported greater relationship quality, compatibility, and intimacy and lower levels of conflict (Balsam, Beauchaine, Rothblum, & Solomon, 2008). More generally, lesbian parents rated their couple relationships more strongly than the heterosexuals (Borneskog, Skoog Svanberg, Lampic, & Sydsjö, 2012), and gay and lesbian partnerships have been found to function better than heterosexual counterparts (Kurdek, 2004).

Sexual identity and relationship quality

Measurable differences between same-sex and heterosexual relationships include the division of domestic labor (Gottman et al., 2003; Solomon et al., 2005), attitudes toward and practices of intimacy (Beals & Peplau, 2001; Gabb, 2019; Kurdek, 1988; Kurdek & Schmitt, 1987), and sexual fidelity (Kurdek, 2003). In lesbian and gay households, the absence of gender difference appears to facilitate greater equality in the sharing of domestic tasks. Equity in divisions of paid work and domestic labor between lesbian mothers and co-parents is a characteristic ideal if not the norm in lesbian-parented families (Perlesz et al., 2010). This runs counter to arguments on the detraditionalisation of intimacy (Giddens, 1992) wherein the absence of traditional gender roles are said to undermine relationship longevity and relationship satisfaction (Bauman, 2003). Data from the Australian 2011 Census indicated that female same-sex couples are nearly twice more likely than heterosexual couples to engage in equal unpaid domestic labor (ABS, 2013). The extent to which patterns of domesticity correspond with rates of relationship satisfaction is contested with claims and counter-claims being put forward. Analyses of Household, Income, and Labor Dynamics in Australia (HILDA) survey data suggest that on all of the measures of health and subjective well-being, LGB individuals fare significantly worse than heterosexual individuals (Wilkins, 2015). Disparities are most apparent during adolescence and early adulthood (Perales, 2014). Whilst differences between LGB and heterosexual couples should not be overstated, and there are many similarities between these groups, HILDA survey data show that LGB couples tend to be slightly less happy with their relationships and indicate that their relationship does not meet their original expectations (Wilkins, 2015). The latter point may, of course, indicate either unrealistically high expectations that are hard to achieve and/or that heterosexual partners, especially women, put up with poor relationship quality because they perceive gender inequalities to be the norm, for example.

In contrast to these studies, other research with lesbians and heterosexual women has found that there are no significant differences of relationship satisfaction, commitment, passion, or intimacy based on sexual orientation (Cusack, Hughes, & Cook, 2012). Research of heterosexual and gay men’s romantic relationships has indicated no differences in the perceived quality of romantic relationships (Wade & Donis, 2007). Across these cohorts, no differences between sexuality-defined types of couple were evident on levels of behaviors or on their contributions to relationship quality (Julien, Chartrand, Simard, Bouthillier, & Bégin, 2003). Extensive comparative studies of same-sex and heterosexual partnerships have demonstrated that whilst differences may be present to a lesser or greater degree, sexuality is not a significant differentiating factor as similar factors predict satisfaction and stability in all these couples (Kurdek, 1988, 1995, 2003; Kurdek & Schmitt, 1987). The processes that regulate relationship functioning can be applied to gay, lesbian, and heterosexual couples (Kurdek, 2004). Despite variabilities, close dyadic relationships regardless of sexual identity work in similar ways (Kurdek, 2006). Parenthood remains a significant and germane factor. Parental stress adversely impacts relationship quality (Walker, Barrett, Wilson, & Chang, 2010), and all new parents experience decline in their relationship quality across the first year of parenthood regardless of sexual orientation (Goldberg, Smith, & Kashy, 2010).

Therefore, findings on the impact and import of sexual orientation on couple relationship experiences are divergent. In the current study, we used our large cross-national dataset to further probe the salience of sexuality, and if, when, or how it differentiates couple experience. We asked the question: “Is sexual orientation associated with a type of long-term relationships based on various relationship qualities, including relationship satisfaction?” The most common form of analysis used in quantitative comparative and cohort research on relationship satisfaction is variable-driven statistical methods. Given the conflicting literature base, we sought to answer the question from a different angle by testing our thesis using a person-centered approach. We hypothesized that significant heterogeneity exists in relationship types as indicated by various relationship qualities such as the division of labor, sex, or faith, and these differences would in turn significantly vary across sexual orientation groups. Thus, our three hypotheses were:

1. Different relationship types exist for those in long-term relationships;

2. Demographic, individual, and relationship variables, including sexual orientation, predict the likelihood of relationship profiles; and

3. Relationship quality and maintenance will be significantly different across relationship profiles.

Latent profile analyses (LPA) allow researchers to identify different groups, in this case relationship types, by analyzing characteristics of the relationship as reported by study participants. It is therefore an emergent approach; deriving types from the dataset rather than imposing these from the outset. The combination of characteristics can then be interpreted to provide a better understanding of relationship types by considering the important characteristics of the relationship, not the individual. We included other important relationship factors known to influence overall relationship satisfaction; that is, sex, money, faith, and chores. Type of long-term relationship (i.e., membership in a unique profile of long-term relationship) was then predicted using characteristics of the respondent (i.e., demographic and relationship characteristics) and these were used to predict reported relationship maintenance and relationship quality scores.

Method

Procedure

To further interrogate the impact of sexuality on relationship experience in our cross-national survey of long-term couple relationships, we devised scales of relationship quality and relationship maintenance (see Chonody, Gabb, Killian, & Dunk-West, 2018; Chonody, Killian, Gabb, & Dunk-West, 2016). Here we analyze findings from these scales alongside data from survey questions. These data were collected in two phases via anonymous surveys using Survey Monkey. In phase one, survey administration occurred throughout the UK in 2011–2012 as part of the Enduring Love? Couple Relationships in the 21^st Century research project funded by the Economic and Social Research Council (RES-062-23-3056). This mixed methods study included items that utilized Likert-type responses as well as open-ended questions to examine how couples understand, experience, and sustain their long-term relationships. In phase two, the survey from this study was replicated in the U.S. and Australia from August to December 2014. In this study we sought to study individuals who are in long-term enduring relationships; however, “long-term” was not specifically delineated. Previous pilot research suggested that the perception of relationship duration is informed by such factors as age, childhood, past relationships, and an imagined future in this relationship (Gabb & Fink, 2015). Similarly, we did not limit a long-term enduring relationship to a traditional monogamous structure. Again, participants determined if her/his relationship qualified as a long-term enduring relationship.

Recruitment in the UK for participation in the study (Phase 1) included news coverage regarding the survey and postings on various online forums, such as community noticeboards, newsletters, and forums about parenting and relationship support. The survey was also distributed in hard copy to hard-to-reach community groups and networks through face-to-face contact. Study recruitment in the U.S. and Australia (Phase 2) was more limited and relied on snowballing methods. The study link was posted on social media platforms (e.g., Twitter, Facebook), and participants were encouraged to share the link with others who might want to participate. The link was also shared through the university networks where the authors work.

Ethics boards in the UK and Australia approved this study prior to data collection as did the Institutional Review Board (IRB) in the U.S. The front page of the online survey included a cover letter detailing participants’ rights, including the right to withdraw at any time, the scope of participation, the focus of the study, and contact details for the researchers. Completion of the survey was considered consent to participate.

Participants and sample

The community sample (N = 7,826) was obtained from the quantitative portion of the UK study (n = 5,683) and the subsequent data collection efforts in the U.S. (n = 1,652) and Australia (n = 491). Data collection efforts in the UK and the U.S. also had participants that indicated that they were from another country (n = 306), and these participants were retained in the current analyses. The participants were mostly 25–54 years of age (72.7%), and the majority was White (91.9%). Nearly half reported being in a long-term relationship from 1–10 years (41.2%) and having on average 1.84 children (SD = 0.93). A majority reported having completed a bachelor’s degree and above (75.7%) and were employed full or part time (72.5%).

Measures

Based on our analysis of the literature, including quantitative surveys and qualitative studies, relationship quality was operationalized as the degree to which a commitment exists, mutual enjoyment (including companionship) is present, and a sense that this person is the right one. In the online survey, a series of statements were presented, and 25 theoretical variables were tested to develop a scale of relationship quality for enduring couples. Utilizing these theoretical variables, items were first checked for their performance by assessing skew and kurtosis and through a review of a correlational analysis. Next, an exploratory factor analysis was conducted for data reduction, and any item that exhibited cross-loadings or a low lambda were removed. In the final step, confirmatory factor analysis was used to establish evidence of the factorial structure (see Chonody et al., 2018). The results of these analyses was the Relationship Quality Scale (RQS), which was found to have strong evidence of factorial validity, and known groups and convergent construct validity were supported (Chonody et al., 2018). The final RQS comprised nine items on perceptions and practices of relationship satisfaction including: “This is the relationship I always dreamed of,” “My partner makes me laugh,” and “We have shared values.” Reliability was excellent with a Cronbach’s α of .891.

Research has found that everyday activities or routine relationship maintenance are integral to relationship satisfaction (Dainton, 2000). It is thus important to assess the degree to which these behaviors are occurring and the impacts these have on relationship quality. We therefore created a strengths-based scale that assesses the behaviors of relationship maintenance utilizing the steps described above for the RQ scale (see Chonody et al., 2016). This Relationship Maintenance Scale (RMS) was developed to measure everyday, routine behaviors that occur in relationships. The RMS also showed good evidence for factorial and known groups validity and had good internal consistency reliability (Cronbach’s α = .78).

We also sought to include other important relationship variables that influence relationship quality and maintenance through statements that probed the relationship work (Chonody et al., 2016) that couples do and the relationship qualities that are cherished, unacknowledged, wished for, and/or expected in long-term relationships. In particular, money, household division of labor, sex, and religiosity have been found to play a role in relationship satisfaction, and items to address these factors were included in the survey. These items all used the same 5-point Likert-type scale (1 = strongly disagree, 5 = strongly agree). In Table 1, these items are listed (numbers 1–8). To garner an overall snapshot of how participants feel overall about their relationship, they were asked: “How happy are you with your partner?” (same Likert-type scale). Lastly, participants were asked: “How would you define a long-term relationship?” whereby participants chose from: length of time together; having children; being a family; shared commitment; being married; or some combination of these. Each facet that they chose was used as a single variable in our analyses (listed in Table 1 [numbers 10–14] by percentage of participants who indicated that this feature was important in their definition). That is, each of the choices for this item was coded as yes/no for the analyses.

Table 1. Items measuring relationship factors as indicators of class within LPA.

			Class 1:	Class 2:	Class 3:	Class 4:
		Full Sample	Traditionalists	De-traditionalists	Working at it	Endurers
Item	Item Description	n = 7,826 (100%)	n = 3692, 47.2%	n = 659, 8.4%	n = 2695, 34.4%	n = 780, 10.0%
1	Sex is an important part of our relationship, M (SD)	3.69 (1.06)	3.93 (0.96)	4.01 (0.93)	3.56 (1.01)	2.87 (1.21)
2	Faith shapes our relationship, M (SD)	2.20 (1.30)	2.18 (1.31)	3.00 (1.47)	2.12 (1.23)	1.92 (1.13)
3	My partner wants to have sex more often than I do, M (SD)	2.83 (1.18)	2.76 (1.12)	2.78 (1.12)	2.93 (1.20)	2.89 (1.40)
4	I think I love my partner more than they love me, M (SD)	2.27 (0.91)	2.07 (0.87)	2.18 (0.94)	2.42 (0.83)	2.70 (1.04)
5	We share our domestic chores fairly, M (SD)	3.48 (1.13)	3.74 (1.05)	3.79 (1.03)	3.31 (1.09)	2.73 (1.22)
6	We share our financial resources fairly, M (SD)	4.21 (1.07)	4.43 (0.91)	4.58 (0.83)	4.07 (1.09)	3.48 (1.36)
7	We argue over money, M (SD)	1.91 (0.98)	1.67 (0.83)	1.77 (0.85)	2.05 (0.97)	2.57 (1.23)
8	Our relationship is mainly about practicalities such as domestic chores and money, M (SD)	2.13 (1.00)	1.76 (0.79)	2.01 (0.94)	2.29 (0.90)	3.27 (1.17)
9	How happy are you with your partner overall?, M (SD)	4.37 (0.84)	5.00 (0.00)	5.00 (0.00)	4.00 (0.00)	2.45 (0.70)
10	Length of time spent together, n (%)	6018 (76.9%)	2687 (44.6%)	563 (9.4%)	2144 (35.6%)	624 (10.4%)
11	Having children, n (%)	779 (10.0%)	10 (1.3%)	410 (52.6%)	264 (33.9%)	95 (12.2%)
12	Being a family, n (%)	1912 (24.4%)	411 (21.5%)	641 (33.5%)	692 (36.2%)	168 (8.8%)
13	Shared commitment, n (%)	5270 (67.3%)	2395 (45.4%)	637 (12.1%)	1822 (34.6%)	416 (7.9%)
14	Being married, n (%)	1448 (18.5%)	327 (22.6%)	529 (36.5%)	458 (31.6%)	134 (9.3%)

Analysis

The aim of our analysis here is to (1) identify profiles of individuals in long-term relationships; (2) examine the association between profile membership and demographic characteristics; and (3) test for differences in relationship profiles on measures of relationship quality and maintenance. To address these aims, we used latent profile analysis (LPA) to empirically derive profiles of long-term relationships as indicated by 14 items (Table 1). LPA is a person-centered analytical approach in which sample data are used to explore the heterogeneity of a sample. LPA models estimate unmeasured profiles or profiles of individuals indirectly through multiple other direct measures completed by a sample (McCutcheon, 1987). LPA analysis is a statistical method within latent variable mixture modeling, capable of estimating profiles with continuous or categorical variables, and has been increasing used in research to explore heterogeneity within samples (Killian, Cimino, Weller, & Hyun Seo, 2019).

Within LPA, the number of latent profiles is tested iteratively beginning with a one-profile solution and increasing the number whilst noting tests of model fit. The current analyses used log likelihood values, Akaike’s information criterion (AIC; Akaike, 1987), the Bayesian information criterion (BIC; Schwartz, 1978), and a BIC value adjusted for sample size. Each of these fit indices compares the likelihood goodness-of-fit with the number of latent profiles within the overall sample (Muthén & Muthén, 2000; Nagin, 1999). Nylund, Asparouhov, and Muthén (2007) have reported that the BIC is the more reliable of these information criteria indices of fit as the AIC has been found to overestimate the number of latent profiles or profiles in the model. For these information criteria, lower values indicate better fit.

The number of profiles is increased to the point where there is a minimal or negligible decrease or even an increase in the BIC and AIC. The Lo-Mendell-Rubin test (LMR; Lo, Mendell, & Rubin, 2001) and Bootstrap Likelihood Ratio test (BLRT; McLachlan & Peel, 2000) compare a latent profile model with k profiles to that of a model with k-1 profiles. The obtained p-value from each test then indicates if the additional profile significantly improves model fit. Nylund et al. (2007) recommend the use of the BLRT when available to compare model fit between those of k and k-1 profiles. An entropy score was calculated per model which indicates how well indicators predict profile membership. Entropy scores closer to 1.0 indicate better profile or profile prediction (Celeux & Soromenho, 1996). Lastly, average posterior profile probabilities (AvePP) were obtained for the final models to ensure scores greater than .70 for each profile, indicating good classification for individuals within a profile.

To further validate the LPA model across countries, the same analytic approach was conducted separately for each subsample from the U.K., the U.S., and Australia. Profile solutions and fit indices were compared across countries. The LPA was completed with Mplus 8.0 (Muthén & Muthén, 2016), which allows for the use of continuous and categorical indicators of profile. We used maximum likelihood estimation with robust standard errors during LPA analyses, which is a full-information estimation method using all available data for all participants to estimate the number and nature of latent profiles in a sample. Other analyses were completed using SPSS 25.0 (Corp, 2017).

Results

Demographic characteristics of LGB sample

Demographic characteristics by sexual orientation and descriptive statistics of the full sample are provided in Tables 2 and 3. Difference among three sexual orientation categories were also calculated. Although many differences in demographic characteristics were significant among sexual orientation categories, many of these differences were of small and negligible effect size (i.e., Cramer’s V and df are lower). Differences were statistically significant due to the statistical power generated by the large sample size. Differences in sexual orientation of note included differences in religious affiliation (χ²[10] = 194.21, p < .001, Cramer’s V = .127, large effect size), relationship status (χ²[12] = 2127.55, p < .001, Cramer’s V = .379, large effect size). Those reporting identifying as heterosexual had a higher proportion identifying as Christian and Muslim with Gay/Lesbian and bi-sexual individuals reporting higher proportions within the Jewish, Buddhist, and none religious groups. Those reporting identifying as Gay/Lesbian and bi-sexual individuals reported more often being a couple not living together and being in a civil partnership. Moderate effect sizes were noted for age (χ²[10] = 140.56, p < .001, Cramer’s V = .097), education level (χ²[8] = 45.61, p < .001, Cramer’s V = .060), and if children are in the household (χ²[3] = 174.34, p < .001, Cramer’s V = .166). In the sample, heterosexual long-term relationships reported having children in the home, reporting comparatively lower levels of educational attainment, and being older.

Table 2. Sexual orientation and demographics.

		Heterosexual	Gay/Lesbian	Bi-sexual
Variable	Full Sample	(n = 6609, 88.3%)	(n = 472, 6.3%)	(n = 407, 5.4%)	Test	Effect size
Gender					x^2 2)-40.82***	Cramer’s V – .074
Male	1459 (19.2%)	1247 (86.8%)	137 (9.5%)	52 (3.6%)
Female	6132 (80.8%)	5341 (88.7%)	332 (5.5%)	351 (5.8%)
Age					χ^2(10)-140.56***	Cramer’s V – .097
16–24	597 (7.8%)	480 (83.6%)	36 (6.3%)	58 (10.1%)
25–34	2091 (27.4%)	1716 (84.2%)	148 (7.3%)	175 (8.6%)
35–44	1942 (25.4%)	1696 (88.8%)	114 (6.0%)	100 (5.2%)
45–54	1519 (19.9%)	1333 (89.3%)	111 (7.4%)	49 (3.3%)
55–64	1084 (14.2%)	1004 (93.6%)	50 (4.7%)	19 (1.8%)
65+	400 (5.2%)	376 (95.7%)	11 (2.8%)	6 (1.5%)
Country					χ^χ(4)χ77.39***	Cramer’s V χ .072
United Kingdom	5683 (72.6%)	4820 (89.0%)	302 (5.6%)	293 (5.4%)
United States	1652 (21.2%)	1436 (89.2%)	103 (6.4%)	71 (4.4%)
Australia	491 (6.3%)	353 (76.2%)	472 (6.3%)	43 (9.3%)
Education level					χ^χ(8)-45.61***	Cramer’s V – .060
No high school diploma	99 (1.5%)	89 (90.8%)	6 (6.1%)	3 (3.1%)
High school diploma/equivalency	305 (4.7%)	272 (90.1%)	13 (4.3%)	17 (5.6%)
Vocational training/some college	1204 (18.4%)	1072 (90.4%)	55 (4.6%)	59 (5.0%)
Professional/bachelor’s degree	2763 (42.3%)	2431 (89.5%)	155 (5.7%)	131 (4.8%)
Master’s/PhD	2164 (33.1%)	1772 (84.3%)	416 (6.5%)	354 (5.5%)
Employment					χ^2(14)-111.87***	Cramer’s V – .093
Part-time work	1749 (26.6%)	1526 (89.1%)	79 (4.6%)	107 (6.3%)
Full-time work	3025 (45.9%)	2573 (86.5%)	255 (8.6%)	148 (5.0%)
Retired	491 (89.1%)	457 (94.8%)	18 (3.7%)	7 (1.5%)
Homemaker/carer	506 (98.8%)	473 (93.8%)	7 (1.4%)	24 (4.8%)
Volunteer	81 (100%)	68 (86.1%)	5 (6.3%)	6 (7.6%)
Full/part-time student	428 (81.7%)	333 (81.6%)	34 (8.3%)	41 (10.0%)
Not employed or working	177 (75.2%)	157 (89.2%)	14 (8.0%)	5 (2.8%)
Disabled	129 (91.1%)	104 (82.5%)	7 (5.6%)	15 (11.9%)
Religious affiliation					x^2(10)-194.21***	Cramer’s V – .127
Christian (Protestant, Catholic)	2884 (46.8%)	2682 (94.0%)	117 (4.1%)	53 (1.9%)
Jewish	104 (1.7%)	88 (86.3%)	12 (11.8%)	2 (2.0%)
Muslim	49 (0.8%)	45 (95.7%)	0 (0.0%)	2 (4.3%)
Buddhist	79 (1.3%)	56 (70.9%)	11 (13.9%)	12 (15.2%)
None	3016 (48.9%)	2481 (84.2%)	236 (8.0%)	230 (7.8%)
Other (Sikh, Hindu)	31 (0.5%)	28 (93.3%)	0 (0.0%)	2 (6.7%)
Parent, child in household (yes)	2878 (44.6%)	2663 (93.7%)	60 (2.1%)	119 (4.2%)	x^2(3)-174.34***	Cramer’s V – .166
Number of children	1.84 (0.93)	1.86 (0.93)	1.61 (0.86)	1.80 (0.93)	F(2, 2849) = 2.20	η^2 = .002
Relationship status					x^−(12)-2127.55***	Cramer’s V – .379
Married	4831 (63.2%)	4513 (95.7%)	53 (1.1%)	152 (3.2%)
Couple- not living together	778 (10.2%)	595 (80.5%)	86 (11.6%)	58 (7.8%)
Living together	1678 (22%)	1292 (79.9%)	158 (9.8%)	167 (10.3%)
Civil partnership	234 (3.1%)	48 (21.8%)	161 (73.2%)	11 (5.0%)
Dating	122 (1.6%)	100 (88.4%)	4 (3.6%)	8 (7.1%)
Number of years in relationship					x^−(10)-249.42***	Cramer’s V – .129
Under 1 year	316 (4.1%)	241 (83.4%)	29 (10.0%)	19 (6.6%)
1–5	1725 (22.3%)	1342 (81.6%)	159 (9.7%)	144 (8.8%)
6–10	1459 (18.9%)	1185 (83.8%)	101 (7.1%)	128 (9.1%)
11–15	1089 (14.1%)	943 (88.5%)	77 (7.2%)	45 (4.2%)
16–20	750 (9.7%)	653 (89.8%)	43 (5.9%)	31 (4.3%)
20+	2381 (30.8%)	2226 (95.7%)	61 (2.6%)	39 (1.7%)
Relationship Quality, total score, M (SD)	37.75 (5.94)	37.68 (6.02)	39.15 (4.91)	37.83 (5.44)	F(2, 6961) = 12.50.28***	η^2 = .004
Relationship Maintenance, total score, M (SD)	30.97 (5.03)	30.86 (5.11)	32.38 (4.10)	31.49 (4.38)	F(2, 6885) = 18.32***	η^2 = .005

Table 3. Demographics and classes.

		Class 1:	Class 2:	Class 3:	Class 4:
		Traditionalists	De-traditionalists	Working at it	Endurers
Variable	Full Sample	(n = 3692, 47.2%)	(n = 659, 8.4%)	(n = 2695, 34.4%)	(n = 780, 10.0%)	Test	Effect size
Gender						x^2(3)-20.62***	Cramer’s V – .052
Male	1459 (19.2%)	686 (47.0%)	165 (11.3%)	473 (32.4%)	135 (9.3%)
Female	6132 (80.8%)	2799 (45.6%)	493 (8.0%)	2204 (35.9%)	636 (10.4%)
Age						x^2(15)-164.35***	Cramer’s V – .085
16–24	597 (7.8%)	330 (55.3%)	41 (6.9%)	180 (30.2%)	46 (7.7%)
25–34	2091 (27.4%)	1075 (51.4%)	114 (6.9%)	712 (34.1%)	160 (7.7%)
35–44	1942 (25.4%)	838 (42.6%)	146 (7.5%)	745 (38.4%)	223 (11.5%)
45–54	1519 (19.9%)	650 (42.8%)	134 (8.8%)	533 (35.1%)	202 (13.3%)
55–64	1084 (14.2%)	451 (41.6%)	115 (10.6%)	402 (37.1%)	116 (10.7%)
65+	400 (5.2%)	169 (42.3%)	79 (19.8%)	120 (30.0%)	32 (8.0%)
Sexual orientation						x^2(6)-37.56***	Cramer’s V – .050
Heterosexual	6609 (88.3%)	2961 (44.8%)	606 (9.2%)	2348 (35.5%)	694 (10.5%)
Gay/lesbian	472 (6.3%)	268 (56.8%)	24 (5.1%)	151 (32.0%)	29 (6.1%)
Bisexual	407 (5.4%)	204 (50.1%)	25 (6.1%)	141 (34.6%)	37 (9.1%)
Country						x^2(6)-91.23***	Cramer’s V – 0.076
United Kingdom	5683 (72.6%)	2752 (48.4%)	407 (7.2%)	1960 (34.5%)	564 (9.9%)
United States	1652 (21.2%)	703 (42.6%)	230 (13.9%)	562 (34.0%)	157 (9.5%)
Australia	491 (6.3%)	237 (48.3%)	22 (4.5%)	173 (35.2%)	59 (12.0%)
Education level						x^2(12)-27.07**	Cramer’s V – 0.037
No high school diploma	99 (1.5%)	42 (42.4%)	17 (17.2%)	29 (29.3%)	11 (11.1%)
High school diploma/equivalency	305 (4.7%)	144 (47.2%)	27 (8.9%)	97 (31.8%)	37 (12.1%)
Vocational training/some college	1204 (18.4%)	523 (43.4%)	117 (9.7%)	422 (35.0%)	142 (11.8%)
Professional/bachelor’s degree	2763 (42.3%)	1274 (46.1%)	211 (7.6%)	1014 (36.7%)	264 (9.6%)
Master’s/PhD	2164 (33.1%)	978 (45.2%)	164 (7.6%)	809 (37.4%)	213 (9.8%)
Employment						x^2(21)-94.43***	Cramer’s V – 0.069
Part-time work	1749 (26.6%)	735 (42.0%)	143 (8.2%)	667 (38.1%)	204 (11.7%)
Full-time work	3025 (45.9%)	1430 (47.3%)	207 (6.8%)	1093 (36.1%)	295 (9.8%)
Retired	491 (89.1%)	228 (46.3%)	81 (16.5%)	145 (29.5%)	37 (7.5%)
Homemaker/carer	506 (98.8%)	215 (42.5%)	51 (10.1%)	182 (36.0%)	58 (11.5%)
Volunteer	81 (100%)	30 (37.0%)	7 (8.6%)	35 (43.2%)	9 (11.1%)
Full/part-time student	428 (81.7%)	214 (50.0%)	33 (7.7%)	143 (33.4%)	38 (8.9%)
Not employed or working	177 (75.2%)	75 (42.4%)	11 (6.2%)	72 (40.7%)	19 (10.7%)
Disabled	129 (91.1%)	63 (48.8%)	20 (15.5%)	32 (24.8%)	14 (10.9%)
Religious affiliation						x^2(15)-137.58***	Cramer’s V – 0.086
Christian (Protestant, Catholic)	2884 (46.8%)	1240 (43.0%)	352 (12.2%)	1005 (34.8%)	287 (10.0%)
Jewish	104 (1.7%)	44 (42.3%)	4 (3.8%)	44 (42.3%)	12 (11.5%)
Muslim	49 (0.8%)	12 (24.5%)	6 (12.2%)	15 (30.6%)	16 (32.7%)
Buddhist	79 (1.3%)	31 (39.2%)	3 (3.8%)	36 (45.6%)	9 (11.4%)
None	3016 (48.9%)	1454 (48.2%)	157 (5.2%)	1111 (36.8%)	294 (9.7%)
Other (Sikh, Hindu)	31 (0.5%)	13 (41.9%)	1 (3.2%)	13 (41.9%)	4 (12.9%)
Parent, child in household (yes)	2878 (44.6%)	1090 (37.9%)	261 (9.1%)	1133 (39.4%)	394 (13.7%)	x^2(3)-130.29***	Cramer’s V – .142
Number of children	1.84 (0.93)	1.78 (0.91)	1.99 (1.03)	1.85 (0.93)	1.90 (0.89)	F(3, 2888) = 4.42**	η^2 = .005
Relationship status						x^2(12)-103.99***	Cramer’s V – 0.067
Married	4831 (63.2%)	2153 (44.6%)	522 (10.8%)	1672 (34.6%)	484 (10.0%)
Couple- not living together	778 (10.2%)	402 (51.7%)	30 (3.9%)	268 (34.4%)	78 (10.0%)
Living together	1678 (22%)	817 (48.7%)	82 (4.9%)	611 (36.4%)	168 (10.0%)
Civil partnership	234 (3.1%)	129 (55.1%)	9 (3.8%)	72 (30.8%)	24 (10.3%)
Dating	122 (1.6%)	55 (45.1%)	7 (5.7%)	40 (32.8%)	29 (16.4%)
Number of years in relationship						x^2(15)-116.44***	Cramer’s V – 0.071
Under 1 year	316 (4.1%)	189 (59.8%)	18 (5.7%)	90 (28.5%)	19 (6.0%)
1–5	1725 (22.3%)	903 (52.3%)	120 (7.0%)	559 (32.4%)	143 (8.3%)
6–10	1459 (18.9%)	716 (49.1%)	99 (6.8%)	495 (33.9%)	149 (10.2%)
11–15	1089 (14.1%)	478 (43.9%)	76 (7.0%)	411 (37.7%)	124 (11.4%)
16–20	750 (9.7%)	311 (41.5%)	63 (8.4%)	290 (38.7%)	86 (11.5%)
20+	2381 (30.8%)	1001 (42.0%)	280 (11.8%)	841 (35.3%)	259 (10.9%)
Relationship Quality, total score, M (SD)	37.75 (5.94)	41.17 (3.08)	41.51 (3.23)	35.87 (3.86)	26.40 (5.95)	F(3, 7097) = 3512.28***	η^2 = .598
Relationship Maintenance, total score, M (SD)	30.97 (5.03)	33.33 (3.36)	34.09 (3.46)	29.69 (3.75)	22.73 (5.44)	F(3, 7014) = 1798.26***	η^2 = .435

Aim 1: LPA analysis

The cross-sectional data were entered into a LPA model using the 14 relationship variables reported by each participant (Table 1). The exploratory LPA model was conducted first with a one-profile solution and then adding a profile after each iteration of the model. The results of the one-, two-, three-, four-, and five-profile solutions are presented in Table 4.

Table 4. LPA modeling results and cross-validation per country.

								Profile size per model, n (%)
Profiles	Loglikelihood	AIC	BIC	Sample-size adjusted BIC	Entropy	LMR p-value	BLRT p-value	1	2	3	4	5
1	−115030.890	230107.780	230267.980	230194.891	-	-	-	7826 (100%)	-	-	-	-
2	−112205.437	224486.875	224751.553	224630.796	.890	<.001	<.001	824 (10.5%)	7002 (89.4%)	-	-	-
3	−110422.220	220950.439	221319.595	221151.172	.856	<.001	<.001	806 (10.3%)	5992 (76.6%)	1028 (13.1%)	-	-
4	−108868.876	217873.752	218347.386	218131.296	.905	<.001	<.001	3692 (47.2%)	659 (8.4%)	2695 (34.4%)	780 (10.0%)	-
5	−108266.149	216698.298	217276.410	217012.653	.887	.991	>.999	2239 (28.6%)	781 (10.0%)	3723 (47.5%)	640 (8.2%)	443 (5.7%)
USA
1	−24738.372	49522.745	49647.169	49574.101	-	-	-	1652 (100%)	-	-	-	-
2	−24094.089	48264.178	48469.748	48349.028	.866	<.001	<.001	1451 (86.3%)	201 (12.2%)	-	-	-
3	−23634.329	47374.658	47661.374	47493.002	.842	<.001	<.001	1159 (70.2%)	181 (11.0%)	312 (18.9%)	-	-
4	−23234.391	46604.782	46972.644	46756.619	.899	<.001	<.001	159 (9.6%)	715 (43.3%)	564 (34.1%)	214 (13.0%)	-
5	−21963.616	44093.231	44542.240	44278.562	.810	<.001	<.001	63 (3.8%)	94 (5.7%)	715 (43.3%)	566 (34.3%)	214 (13.0%)
UK
1	−82960.424	165966.849	166119.689	166046.602	-	-	-	5683 (100%)	-	-	-	-
2	−80902.835	161881.670	162134.189	162013.436	.901	<.001	<.001	5114 (90.0%)	569 (10.0%)	-	-	-
3	−79728.768	159563.535	159915.733	159747.314	.867	<.001	<.001	4464 (78.5%)	666 (11.7%)	553 (9.7%)	-	-
4	−78636.165	157408.331	157860.207	157644.123	.909	.034	<.001	1956 (34.4%)	2764 (48.6%)	564 (9.9%)	399 (7.0%)	-
5	−78567.690	157301.380	157852.934	157589.185	.779	.750	<.001	511 (9.0%)	554 (9.7%)	1300 (22.9%)	2975 (52.3%)	344 (6.1%)
AUS
1	−6766.089	13578.179	13674.697	13601.695	-	-	-	491 (100%)	-	-	-	-
2	−6589.101	13254.202	13413.667	13293.056	.841	.002	<.001	73 (14.9%)	418 (85.1%)	-	-	-
3	−6488.309	13082.617	13305.029	13136.808	.837	.007	<.001	57 (11.6%)	332 (67.6%)	102 (20.8%)	-	-
4	−6431.691	12999.381	13284.739	13068.908	.832	.440	<.001	337 (68.6%)	59 (12.0%)	40 (8.1%)	55 (11.2%)	-
5	−6382.295	12930.591	13278.896	13015.455	.834	.746	<.001	33 (6.7%)	57 (11.6%)	295 (60.1%)	33 (7.7%)	68 (13.8%)

The LPA analyses resulted in a four-profile solution (AIC = 217873.752, BIC = 218347.386, Entropy = .905; BLRT = 2349.728, p < .001). The four-profile solution presented with better model fit than previous models with fewer profiles (Table 4). Additionally, both LMR and BLRT indicated a four-profile solution to be a significant improvement over models with fewer profiles. A five-profile solution did not result in improved model fit with a decrease in entropy scores (Entropy = .887) and no significant improvement in model fit (BLRT = 7786.675, p = .999). The four-profile solution had an AvePP of .949, indicating excellent average patent profile posterior probability for each individual assigned to a profile.

Cross validation by country

LPA modeling was then conducted separately with the subsamples from each country (Table 4). The four-profile solution was supported in each the UK (AIC = 217873.752, BIC = 218347.39, Entropy = .905; BLRT = 2349.728, p < .001) and U.S. subsamples (AIC = 46604.78, BIC = 46972.64, Entropy = .899; BLRT = 23549.00, p < .001). The four-profile model was the most supported in the UK subsample. The addition of the fifth profile was found not to add significantly to model fit to the data (LMR p = .744) and entropy scores (.779) dropped below the conventional standard of .80 as evidence of a well-fitting model. In the U.S. subsample, the four-profile solution had the highest entropy score and demonstrated improvement in model fit over the three-profile solution. While AIC and BIC scores continued to improve in the five-profile solution, demonstrated an arbitrary splitting of one profile from the four-profile solution into two separate profiles to satisfy the modeling of a fifth profile. This splitting in profiles with the additional profile model was deemed to be contrary to interpretability warned against by Muthén and Muthén (2000). The four-profile solution was deemed the best fitting in the U.S. subsample.

The smallest subsample was the 491 participants from Australia (6.3% of the total international sample of 7,826). The Australian subsample appeared to support a three-profile or four-factor solution. Although the entropy score for the three-profile solution was the highest (.837), the marginal decrease in the entropy score for the four-profile solution (.832) and the significant increase in model fit with the additional profile (BLRT = 6488.31, p < .001) lead us to use the four-profile solution within analyses with the total sample.

Profile interpretation

Traditionalists

Profile 1 (47.2%; n = 3,692). Profile 1 included individuals who had the most idealized (and traditional) view of how a long-term relationship is defined (see Figure 1a). That is, they highly ranked having kids, being a family, and being married as characteristics of a long-term relationship. They also reported that sex was an important aspect of their relationship. Individuals in this profile reported sharing domestic chores and financial resources fairly, and they had the lowest scores for arguing over money. Similarly, this profile reported the lowest scores when asked about their relationships only being based on practicalities such as domestic chores and money. This group also reported high levels of happiness with their partner.

Figure 1. Differences on profiles in LPA per indicator.

De-traditionalists

Profile 2 (8.4%; n = 659) reported the highest levels of finding sex as an important part of their relationship and of faith shaping their relationship (see Figure 1b). These individuals reported sharing domestic chores and financial resources the most fairly as well as high levels of happiness with their partners. Members of this profile did not define their relationships within the traditional norms of a long-term relationship. That is, they were least likely to define their relationship in terms of the length of time together, having children, being a family, sharing commitment, and being married. They had the highest rating for the importance of sex and similar to the Traditionalists, they were less likely to describe their relationship as one based on practicalities.

Working at it

Profile 3 (34.4%; n = 2,695) reported moderate amounts of all indicators of profile for relationships (see Figure 1c); however, they were fairly low on faith shaping the relationship. This profile was quite moderate in all ways, and on most indicators, they were just above/below profiles one and two.

Endurers

Profile 4 (10.0%; n = 780) included individuals who reported that sex was the least important part of their relationship and endorsed arguing over money the most (see Figure 1d). This profile most often reported that their relationship was about practicalities such as domestic chores and money. Relatedly, this profile also reported the lowest scores for sharing financial and domestic chores fairly. They had the lowest level of happiness with their partner and the highest score on feeling that they love their partner more than their partner loves them. In terms of how they defined a long-term relationship, this profile endorsed shared commitment at the highest rate.

Aim 2: profiles and demographics

Tests for association between profile membership and demographic characteristics indicated a number of significant results. Age categories were significantly associated with profile membership (X²[15] = 164.35, p < .001, Cramer’s V = .085, large effect size). Profile 1 was proportionately younger than other profiles whereas Profile 2 was older.

Those participants who were coupled yet not living together, living together, or in a civil partnership were more likely to be classified in Profile 1 (χ²[12] = 103.99, p < .001, Cramer’s V = .067, moderate effect size) and less so in Profile 2. Long-term relationships (20 years or more) were more likely to be classified in Profile 2. Similarly, length of the relationship was associated with profile membership (χ²[15] = 116.44, p < .001, Cramer’s V = .071, large effect size). Participants in newer relationships were classified more often in Profile 1 (10 years and newer).

In terms of sexual orientation, a significantly greater proportion of LGB were classified in Profile 1 (χ²[6] = 37.56, p < .001, Cramer’s V = .050, moderate effect size) and a smaller proportion in Profile 2. Alternatively, participants who were parents with children in the household were proportionately classified as groups 2, 3, and 4 (χ²[3] = 130.29, p < .001, Cramer’s V = .142, large effect size). The average number of children, though statistically significant, did not indicate any practical difference between profiles (F(3, 2888) = 4.42, p = .004, η² = .005).

Country of origin was significantly associated with profile membership (χ²[6] = 91.23, p < .001, Cramer’s V = .076, moderate effect size). Those participants from the U.S. were more likely classified into Profiles 1 or 2 when compared to other profiles or countries.

Less represented religious groups in the sample tended to deviate proportionally in terms of profile membership (χ²[15] = 137.58, p < .001, Cramer’s V = .086, large effect size). Those participants identifying as Jewish, Buddhist, and other were more likely to be classified in Profile 3. Muslim participants were less represented in Profile 1, yet greater in Profiles 2 and particularly 4.

Though statistically significant, gender (χ²[3] = 20.62, p < .001, Cramer’s V = .052, small effect size) and education level (χ²[12] = 27.07, p < .001, Cramer’s V = .037, small effect size) were largely unrelated with profile membership.

Aim 3: profiles and relationship quality and maintenance

Relationship quality (RQ) and relationship maintenance (RM) scores significantly differed across profiles. RQ scores from Profiles 1 and 2 were significantly higher than those reported by Profiles 3 and 4 (F[3, 7097] = 3512.28, p < .001, η² = .598). Profile 3 also reported significantly greater scores than Profile 4. Similarly, participants in each profile reported significantly different RM scores (F[3, 7014] = 1798.26, p < .001, η² = .435). Participants in Profile 2 reported the highest RM scores followed by Profiles 1, 3, and then 4. Though gay and lesbian respondents reported greater RQ and RM total scores than heterosexual respondents, the differences were quite small. Identified sexual orientation explained less than 1% of the variation in RQ (F[2, 6961] = 12.50, p < .001, η² = .004) and RM total scores (F[2, 6885] = 18.32, p < .001, η² = .005).

Discussion

Four long-term relationship types were identified through the LPA. For profile one (Traditionalists) and profile two (De-traditionalists), we identified relationships that are marked primarily by happiness with one’s partner, relationship quality, and high levels of relationship maintenance. However, some key differences between these two relationship types were notably evident. For profile 1, participants were likely to be younger, identify as LGB, and less likely to have children or be married. This relationship type was also marked by a more traditional definition of a long-term relationship even though those characteristics were not necessarily present in the profile (e.g., children, marriage). On the other hand, profile 2, which was more likely to be composed of those who were married and in a 20+ year relationship, did not define long-term relationships along these lines. In fact, this profile was least likely to describe an enduring relationship by marriage, children, or shared commitment. These differences may be reflective of the fact that after being in a relationship for a long time, one comes to see its endurance beyond those traditional social markers.

Participants associated with profiles 1 and 2, on face value, run counter to socio-cultural population-level data on increasingly permissive sexual attitudes and sexual experimentation among the younger population, for example (Mercer et al., 2013). Transformations of intimacy (Giddens, 1992) and the critique of monogamy (Kipnis, 2004) and marriage as a “zombie institution” (Beck, 2000b, p. 198) attach non-traditional lifestyles to younger generations of LGB and queer individuals and those in non-marital relationships (i.e., those in profile two), whereas our findings show these individuals being more closely aligned with profile 1, the Traditionalists. There is, however, precedent for such contrariety. Qualitative research on same-sex marriage has shown that for young LGB people, to attain ordinary marriages involves reflexivity and heightened degrees of agency. When conventions have been denied, ordinariness is a political stance as well as a personal goal. LGB couples are not, therefore, unreflexive followers of tradition, they are instead “active (and sometimes highly reflexive) scriptors of convention” (Heaphy et al., 2013, p. 172).

For profile 3 (i.e., Working at it), profile characteristics suggested that these individuals are muddling along in their relationships with a moderate level of happiness, relationship quality, and relationship maintenance. Their relationship remains intact, and they do not appear to define their relationships based solely on practicalities, such as domestic chores. Conversely, profile 4 (i.e., Endurers) had the least amount of happiness in their relationship and the lowest levels of both relationship quality and relationship maintenance. This relationship was marked by practicalities, arguing over money, and disparate divisions of labor and finances. However, shared commitment was highly endorsed as the marker of a long-term relationship, which may be the glue that is holding these relationships together. Profiles 3 and 4 reflect extant research, which points to parenthood and money worries as significant relationship stressors (Walker et al., 2010) that adversely impact on relationship satisfaction.

Limitations

Our results should be contextualized within the limitations of the study. First, we used a convenience sample, and this limits generalizability beyond the scope of the study. However, a large sample from multiple countries and efforts to validate the LPA model across the international sample should increase external validity to individuals in these countries. Future research should seek to include a more representative and international sample, including couples who may be currently seeking relationship support. Second, our sample of LGB participants was relatively small given the larger populations of the countries included in the survey. Thus, further research in this area is needed, and efforts to recruit a larger sample of LGBQ participants are warranted. Similarly, we did not ask bisexual participants to indicate the gender of their partner, thus we cannot assess the extent to which participants were same-gender passing or heterosexual passing. These differences may be important in relationship quality and should be further investigated in future research. Third, we used new measures in our study, which may have influenced the results. Replication using other scales would add to our understanding of the current results. Relatedly, other important relationship variables were not included in this study, such as financial stress, health or mental health problems, infidelity, or child disability, which can have a significant impact on relationship quality and maintenance. Additional studies on how these factors can impact couples are essentially to understanding enduring relationships.

Additional RQS and RMS survey research should also investigate the ways in which couples negotiate relationship boundaries to include non-monogamous agreements and significant others (such as friends, extended family, and children), for example, and how wider practices of intimacy (Jamieson, 1998) beyond the couple (Roseneil & Budgeon, 2004) may influence the way that relationships are experienced and defined. Relatedly, the ways in which identity impacts one’s lived experience necessitates greater investigation, including gender, sexual orientation, and sexual practices.

Sexuality matters

So, what role does sexual orientation play in these profiles and the relationship satisfaction of long-term partnerships? Our results corroborate previous findings (Kurdek, 1988, 1995, 2003; Kurdek & Schmitt, 1987) in that there was no practical difference between LGB and heterosexual people in terms of relationship quality and relationship maintenance. Rather, sexual orientation was relevant when discerning the relationship type, which in turn was related to relationship quality and relationship maintenance. In fact, profile membership explained 44–60% of the variance in relationship maintenance and relationship quality. For the two profiles that were more likely to include those who are LGB (profile 1, Traditionalists and to a lesser extent profile 2, De-traditionalists), the relationship characteristics were marked by high levels of shared domestic labor and financial resources, high levels of happiness with one’s partner, and the importance of sex to the relationship. These two profiles had the most relationship quality and relationship maintenance, indicating that these characteristics are important to the continuance of relationships over time. This is further supported by the associated features of profile 4, Endurers, which were essentially the opposite of all those things working well for profile 1 and 2. Therefore, sexual orientation in-and-of-itself may not play a significant role in relationship quality, but LGB people were more likely to have an egalitarian relationship, which does have an impact on relationship quality. Future research with a LGBQ sample should seek to investigate these profiles and further parse potential meaning of how these relationship variables impact relationship quality and maintenance.

The lack of significance for sexual orientation in our analyses may appear to be counterintuitive to the dominant discourse attached to sexual identity given the evolving and somewhat precarious social and legal status of same-sex couples. As discussed in our literature review, social and historical conditions are important factors in understanding self-reports of relationship satisfaction. Overwhelmingly, research into sexual identity and young people, for example, points to the stigmatization of sexual minority identity categories and their negative impact on identity in relation to victimization in educational contexts (Russell, Ryan, Toomey, Diaz, & Sanchez, 2011), increased risk of suicide (Baams, Grossman, & Russell, 2015), and substance misuse and depression (Lea, de Wit, & Reynolds, 2014), for example. It is also important to recognize that legislative changes in same-sex marriage do little to ameliorate historical stigmatization of LGB people, particularly for those whose intimate relationships fall outside of traditional heteronormative assumptions and monogamous expectations (Sheff, 2005; Toft & Yip, 2018).

Our sample is older. They grew up in more hostile social contexts than contemporary queer youth, and the legacy of this context is not easily left behind. There is some evidence, for example, that young queer people today think about sexual orientation as an identity whereas older people reflect on sexual orientation in behavioral terms (Hart-Brinson, 2016). LGB experiences are also not homogenous, with recent research pointing to the enduring discrimination and marginalization of bisexual people (Taylor, Power, Smith, & Rathbone, 2019), for example. Further research is needed to gain a better understanding of the mechanisms through which dominant negative discourses of heteronormativity and traditional monogamy are mediated in the context of same-sex relationships. For now, whilst there is little empirical work on the intersections of resilience and LGB identities (Kwon, 2013), our findings provide an important foundation upon which to build knowledge about the resilience of people who identify as LGB and the strength of their partnerships.

Our data, for example, point to the ways in which everyday relationship practices are constitutive of relationship quality and couple identity in terms of shared domesticity and the role of practicalities (Gabb & Fink, 2015), so too the role of sexual intimacy as ordinary relationship maintenance behavior (Elliott & Umberson, 2008; Erickson, 2005; Gabb, 2019). Contextualizing participant responses about what they do in relationships, showing how social interactions, and thus, sociality, produce/s social-sexual identity needs further investigation through empirical studies whereby these understandings of the sexual self can be examined (Gabb, 2013; Jackson & Scott, 2010). However, whilst routines may reinforce couple identity, our findings suggest that relationship satisfaction is highest for those who base the relationship on something other than the practicalities associated with a long-term relationship. To understand the nature of long-term partnerships across sexual identities, it is important to recognize the ways in which relationship characteristics influence the overall quality of that relationship. Further empirical work in understanding relationship satisfaction should also investigate the role that reflexivity and detraditionalization play in mediating experiences and household practices.

Additional information

Funding

This work was supported by the Economic and Social Research Council [RES-062-23-3056].